WO2021097618A1

WO2021097618A1 - Point cloud segmentation method and system, and computer storage medium

Info

Publication number: WO2021097618A1
Application number: PCT/CN2019/119222
Authority: WO
Inventors: 李延召
Original assignee: 深圳市大疆创新科技有限公司
Priority date: 2019-11-18
Filing date: 2019-11-18
Publication date: 2021-05-27
Also published as: CN114270400A

Abstract

A point cloud segmentation method and system, and a computer storage medium. Said method comprises: acquiring point cloud data (S110); determining a ground point from a point cloud according to the height of the point cloud in the point cloud data (S120); determining above-ground points from the point cloud according to the ground point (S130); and determining a target point from the above-ground points according to the characteristics of the above-ground points, the target point being a foreground point or a background point. Said method is simple, effective and easy to implement, has high universality and adaptability, and can implement efficient and robust segmentation of different types of point cloud data.

Description

Point cloud segmentation method, system and computer storage medium

Technical field

The embodiments of the present invention relate to the field of information technology, and more specifically, to a point cloud segmentation method, system, and computer storage medium.

Background technique

Point cloud segmentation is to segment point clouds with similar attributes into disjoint sets, so that the segmented point clouds in the same area have similar characteristics. Point cloud segmentation is a key link in point cloud data processing, which can provide important information for subsequent processing, and is the prerequisite and basis for object recognition.

The current point cloud segmentation methods usually include two categories. The first type is a segmentation algorithm based on deep learning. The usual method is to first collect a large amount of target scene data, and then manually label them one by one to train the deep network model as sample data to obtain a point cloud segmentation device. This method is complicated to implement and strongly depends on the scene, the installation method and model of the lidar. The second category is traditional algorithms. The usual approach is to design complex classification logic based on the neighborhood relationship between points and finally realize the segmentation of the point cloud. However, this kind of method is usually only suitable for regular scanning lidars, and lacks a systematic segmentation scheme for the point cloud of complex trajectories and non-repetitive scanning lidars.

Summary of the invention

A series of simplified concepts are introduced in the content of the invention, which will be described in further detail in the detailed implementation section. The inventive content part of the present invention does not mean an attempt to limit the key features and necessary technical features of the claimed technical solution, nor does it mean an attempt to determine the protection scope of the claimed technical solution.

In view of the shortcomings of the prior art, the first aspect of the embodiments of the present invention provides a point cloud segmentation method, including:

Obtain point cloud data;

Divide the horizontal plane of the point cloud space in the point cloud data into multiple grids, and perform relative height filtering in each grid according to the lowest height of the point cloud points in the grid ；

Performing absolute height filtering according to the absolute height of the point cloud point;

Perform plane fitting on the point cloud points after the relative height filtering and the absolute height filtering, and determine the ground point and the above ground point according to whether the distance between the point cloud point and the fitted plane is greater than a threshold;

Extract seed points from the above-ground points according to the characteristics of the above-ground points, and perform regional growth based on the seed points to determine a preliminary target point;

Define a target area according to the preliminary target point, and determine an above-ground point in the target area as a target point;

Clustering the ground points outside the target area to obtain multiple point cloud clusters;

According to the characteristics of the point cloud cluster, a target point cloud cluster in the point cloud cluster is determined, and an above-ground point in the target point cloud cluster is determined as the target point.

A second aspect of the embodiments of the present invention provides a point cloud segmentation method, and the method includes:

Obtain point cloud data;

Determining a ground point from the point cloud according to the height of the point cloud in the point cloud data;

Determining an above-ground point from the point cloud according to the ground point;

A target point is determined from the above-ground points according to the characteristics of the above-ground point, and the target point is a front scenic spot or a background point.

A third aspect of the embodiments of the present invention provides a point cloud segmentation method, and the method includes:

Obtain point cloud data;

Extracting seed points in the point cloud data according to the characteristics of the point cloud points in the point cloud data;

Perform region growth based on the seed point to determine a preliminary target point in the point cloud data.

A fourth aspect of the embodiments of the present invention provides a point cloud segmentation system, and the system includes:

Memory, used to store executable instructions;

The processor is configured to execute the instructions stored in the memory, so that the processor executes the point cloud segmentation method of the embodiment of the present invention.

A fifth aspect of the embodiments of the present invention provides a computer storage medium on which a computer program is stored, and when the program is executed by a processor, the point cloud segmentation method of the first aspect of the embodiment of the present invention is implemented.

The point cloud segmentation method, system and computer storage medium of the embodiments of the present invention propose a set of simple, effective, easy-to-implement, universal and adaptable point cloud segmentation solutions, which can achieve high efficiency and robustness for different types of point cloud data Segmentation.

Description of the drawings

In order to explain the technical solutions of the embodiments of the present invention more clearly, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some of the present invention. Embodiments, for those of ordinary skill in the art, without creative labor, other drawings can be obtained based on these drawings.

Fig. 1 is a schematic flowchart of a point cloud segmentation method according to an embodiment of the present invention;

Fig. 2 is a schematic flowchart of point cloud data preprocessing according to an embodiment of the present invention;

Fig. 3 is a schematic diagram of point cloud downsampling according to an embodiment of the present invention;

4 is a schematic flowchart of determining an above-ground point from a point cloud according to an embodiment of the present invention;

Fig. 5 is a schematic diagram of gridding the horizontal plane of a point cloud according to an embodiment of the present invention;

Fig. 6 is a schematic flowchart of determining a target point from above-ground points according to an embodiment of the present invention;

7A-7D are schematic diagrams of point clouds for segmentation according to an embodiment of the present invention;

Figure 7E is an enlarged view of Figure 7D;

FIG. 8 is a schematic flowchart of a point cloud segmentation method according to another embodiment of the present invention;

FIG. 9 is a schematic flowchart of a point cloud segmentation method according to another embodiment of the present invention;

Fig. 10 is a structural block diagram of a point cloud segmentation system according to an embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions, and advantages of the present invention more obvious, the exemplary embodiments according to the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited by the exemplary embodiments described herein. Based on the embodiments of the present invention described in the present invention, all other embodiments obtained by those skilled in the art without creative work should fall within the protection scope of the present invention.

In the following description, a lot of specific details are given in order to provide a more thorough understanding of the present invention. However, it is obvious to those skilled in the art that the present invention can be implemented without one or more of these details. In other examples, in order to avoid confusion with the present invention, some technical features known in the art are not described.

It should be understood that the present invention can be implemented in different forms and should not be construed as being limited to the embodiments presented here. On the contrary, the provision of these embodiments will make the disclosure thorough and complete, and will fully convey the scope of the present invention to those skilled in the art.

The purpose of the terms used here is only to describe specific embodiments and not as a limitation of the present invention. When used herein, the singular forms "a", "an" and "the/the" are also intended to include plural forms, unless the context clearly indicates otherwise. It should also be understood that the terms "composition" and/or "including", when used in this specification, determine the existence of the described features, integers, steps, operations, elements and/or components, but do not exclude one or more other The existence or addition of features, integers, steps, operations, elements, components, and/or groups. As used herein, the term "and/or" includes any and all combinations of related listed items.

In order to thoroughly understand the present invention, a detailed structure will be proposed in the following description to explain the technical solution proposed by the present invention. The optional embodiments of the present invention are described in detail as follows. However, in addition to these detailed descriptions, the present invention may also have other embodiments.

Fig. 1 shows a schematic flowchart of a point cloud segmentation method 100 according to an embodiment of the present invention. As shown in FIG. 1, the method 100 includes the following steps:

In step S110, point cloud data is acquired;

In step S120, a ground point is determined from the point cloud according to the height of the point cloud in the point cloud data;

In step S130, an above-ground point is determined from the point cloud according to the ground point;

In step S140, a target point is determined from the above-ground points according to the characteristics of the above-ground points, and the target point is a front scenic spot or a background point.

The point cloud segmentation method 100 of the embodiment of the present invention can quickly and effectively segment a point cloud into ground points, above ground points, front scenic spots, and background points, and can achieve efficient and robust segmentation for different types of point cloud data.

Exemplarily, in step S110, the point cloud data may be point cloud data collected by a distance measuring device, the distance measuring device may be a lidar, and the lidar may be a regular and repeated scanning lidar , It can also be a lidar with complex scanning trajectory with non-repetitive scanning characteristics. In one embodiment, the distance measuring device is used to sense external environmental information, for example, distance information, orientation information, reflection intensity information, speed information, etc. of environmental targets. The point cloud points in the point cloud data may include at least one of the external environment information measured by the distance measuring device.

In one implementation, the distance measuring device can detect the distance from the probe to the distance measuring device by measuring the time of light propagation between the distance measuring device and the probe, that is, the time-of-flight (TOF). . Alternatively, the ranging device may also detect the distance from the probe to the ranging device through other technologies, such as a ranging method based on phase shift measurement, or a ranging method based on frequency shift measurement.

As an example, the distance measuring device may include a transmitting circuit, a receiving circuit, a sampling circuit, and an arithmetic circuit.

Among them, the transmitting circuit can emit a light pulse sequence (for example, a laser pulse sequence). The receiving circuit can receive the light pulse sequence reflected by the object to be detected, and perform photoelectric conversion on the light pulse sequence to obtain an electrical signal, which can be output to the sampling circuit after processing the electrical signal. The sampling circuit can sample the electrical signal to obtain the sampling result. The arithmetic circuit can determine the distance between the distance measuring device and the detected object based on the sampling result of the sampling circuit.

In some implementation manners, in addition to the foregoing circuit, the distance measuring device may further include a scanning module for changing the propagation direction of at least one laser pulse sequence emitted by the transmitting circuit.

In an embodiment, the scanning module may include a plurality of optical elements for changing the propagation path of the light beam, wherein the optical element may change the propagation path of the light beam by reflecting, refracting, or diffracting the light beam. For example, the scanning module includes a lens, a mirror, a prism, a galvanometer, a grating, a liquid crystal, an optical phased array (Optical Phased Array), or any combination of the foregoing optical elements. In an example, at least part of the optical element is moving, for example, the at least part of the optical element is driven to move by a driving module, and the moving optical element can reflect, refract, or diffract the light beam to different directions at different times.

In some embodiments, multiple optical elements of the scanning module can rotate or vibrate around a common axis or different axes, and each rotating or vibrating optical element is used to continuously change the propagation direction of the incident light beam. In one embodiment, the multiple optical elements of the scanning module may rotate at different speeds or vibrate at different speeds. The rotation speed of the optical element directly determines the uniformity of the scanning point cloud of the scanning module.

In one embodiment, each optical element in the scanning module can project light to different directions by rotating, so that the space around the distance measuring device is scanned. It can be driven by the same or different drivers, so that the rotation speed and/or rotation of multiple optical elements are different, so that the collimated light beam can be projected to different directions in the outer space, and a larger space range can be scanned. Optionally, the plurality of optical elements may include two or three wedge-angle prisms rotating at different rotation speeds. It is understandable that when the speed of the optical element in the scanning module changes, the scanning pattern will also change accordingly.

In one embodiment, the distance measuring device can be applied to a mobile platform, and the distance measuring device can be installed on the platform body of the mobile platform. A mobile platform with a distance measuring device can measure the external environment, for example, measuring the distance between the mobile platform and obstacles for obstacle avoidance and other purposes, or performing two-dimensional or three-dimensional surveying and mapping of the external environment, and the point cloud collected by the distance measuring device The data may be point cloud data collected for road scenes. In some embodiments, the mobile platform includes at least one of an unmanned aerial vehicle, a car, a remote control car, a robot, and a camera.

In one embodiment, the point cloud data obtained in step S110 may be preprocessed point cloud data, and step S110 includes: obtaining original point cloud data; preprocessing the original point cloud data to obtain preprocessing After the point cloud data.

Specifically, the original point cloud data usually contains the following problems: 1. There is noise in the point cloud data, which affects the robustness of the segmentation result; 2. The ranging device that collects the point cloud data is tilted relative to the ground, making the ground in the point cloud It is not parallel to the coordinate plane, which increases the processing complexity; 3. There are too many points in the cloud, and the amount of data is too large, resulting in slow processing speed. The preprocessing step can at least partially solve the above-mentioned problems to obtain a better segmentation effect.

In an embodiment, referring to FIG. 2, the preprocessing may include step S210, removing noise points in the original point cloud data.

Specifically, there are usually noise points in the original point cloud data. These noise points may be derived from the light pulse signal reflected by the non-measurement target object, or the normal pulse echo reflected by the measurement target object is distorted in the analog circuit and caused by the depth calculation. The solution deviation of the model. The noise points in the point cloud data are usually discrete points far from the subject point cloud. Therefore, in the embodiment of the present invention, statistical filtering can be applied to denoise the original point cloud data according to the discreteness of the noise points.

Among them, the denoising process can be carried out in a three-dimensional space. Specifically, the number of adjacent point cloud points in the neighborhood of the original point cloud point in the original point cloud data in the point cloud space can be counted, and the original point cloud point whose number of adjacent point cloud points is less than the threshold can be judged as Noise, and remove the noise.

Wherein, the neighborhood of the original point cloud point in the three-dimensional space may be a spherical three-dimensional space area with a predetermined radius centered on the current point cloud point, and the size of the radius may be set according to actual needs. After counting the number of adjacent point cloud points in the neighborhood of the current point cloud point, it can be compared with a preset threshold. If the number of adjacent point cloud points is less than the threshold, the current point cloud point is judged as a noise point. And remove it.

In one embodiment, all point cloud points may adopt the same neighborhood radius and threshold value of neighboring points. In another embodiment, since the point cloud collected for objects that are closer is denser, and the point cloud collected for objects that are farther away is sparse, it is easier to be judged as noise points, so the distance is more dense. Far point clouds can be suitable for larger radii or smaller thresholds.

Alternatively, the denoising process can also be performed on the projection surface. At this time, removing the noise in the original point cloud data may include: projecting the original point cloud data on the projection surface; counting the number of adjacent projections of the original point cloud points on the projection surface in the neighborhood; The original point cloud points with the number of adjacent projections less than the threshold are judged as noise points, and the noise points are removed. Wherein, the projection surface may be a front view, that is, a vertical plane perpendicular to the axis of the lidar. In a road scene, it is simpler and more effective to denoise by projecting onto the front view.

When projecting onto the projection surface, the neighborhood of the original point cloud point may be a circular two-dimensional plane area centered on the current point cloud point. After counting the number of adjacent point cloud points in the two-dimensional neighborhood of the current point cloud point, it can be compared with the preset threshold. If the number of adjacent point cloud points is less than the threshold, the current point cloud point is judged as Noise and remove it.

In addition, the above two methods can also be combined for denoising. Combining multiple filtering algorithms for denoising can obtain a more comprehensive and effective filtering effect.

In an embodiment, the point cloud preprocessing further includes step S220 of performing coordinate transformation on the original point cloud data so that the coordinate system of the point cloud space is parallel to the ground. Step S220 may be performed after step S210, and the original point cloud data at this time is the original point cloud data after removing noise.

In step S220, for the problem of the inclination of the ranging device relative to the ground, the attitude of the ranging device that collects the point cloud data relative to the ground may be estimated first, and then the coordinate transformation of the entire point cloud is performed based on the estimation result, so that The coordinate system in the point cloud space is parallel to the ground. The method of estimating the attitude of the distance measuring device can include offline calibration, online method or any other suitable method. Firstly, the coordinate transformation matrix is calculated, and the original point cloud data and the transformation matrix are dot-producted to obtain the calibrated Point cloud data.

Further, although the point cloud filtered in step S210 has removed noise, its data is still too dense, resulting in too slow calculation speed. Therefore, in one embodiment, the point cloud preprocessing further includes step S230 of down-sampling (or referred to as down-sampling) the original point cloud data. The down-sampling adopts a voxel matrix method to appropriately simplify redundant data on the premise of maintaining the point cloud characteristics. Step S230 may be performed after S220, and the original point cloud data at this time is the original point cloud data after coordinate transformation.

In view of the problem of too much data volume, in the embodiment of the present invention, a method of down-sampling the region of interest is adopted to reduce the points to be processed. Specifically, the method includes: selecting a region of interest in the point cloud space where the original point cloud is located in the original point cloud data; dividing the region of interest into a voxel matrix composed of a plurality of voxels; If each voxel in the matrix contains multiple original point cloud points, one of the original point cloud points will be retained. Figure 3 is a schematic diagram of dividing the point cloud space into a voxel matrix.

Specifically, the division of voxels is to divide the region of interest in the point cloud space into multiple volume spaces, each volume space is a voxel, and the voxel is the abbreviation of Volume Pixel.

The region of interest may be selected according to application requirements. For example, for a driving vehicle, the region of interest may be an area in a certain range in front of the road surface. At the current point in time, objects beyond this range do not need attention temporarily, so the point cloud points in the region of interest are reserved, and the point cloud points outside the region of interest are discarded. Then, the above region of interest can be divided into a voxel matrix of size Nx*Ny*Nz, where Nx is the number of voxels divided along the x axis, Ny is the number of voxels divided along the y axis, and Nz is the number of voxels divided along the z axis. The number of voxels divided by the axis. Exemplarily, the region of interest may be a cube region with x ranging from 0m to 100m, y ranging from -20m to 20m, and z ranging from 0 to 6m, as shown in FIG. 3.

In this embodiment, only the voxel division and down-sampling are performed on the region of interest, and when the subsequent segmentation is performed, only the region of interest is segmented, and there is no need to segment the entire point cloud, which reduces the amount of calculation.

In some embodiments, down-sampling can be performed based on different standards or rules. The purpose of downsampling is to retain one or more representative point cloud points in each voxel, and remove the remaining redundant point cloud points of each voxel. As an example, for each voxel, if there are multiple point cloud points falling into it, one of the point cloud points is arbitrarily retained. As another example, it is also possible to selectively retain specific point cloud points in each voxel. For example, by calculating the normal vector and distance of the point cloud point, the center of gravity of each voxel is determined, and the position of the center of gravity of the voxel is kept or separated from the point cloud. The point cloud with the closest center of gravity.

After the above-mentioned down-sampling processing, a significant reduction in point cloud points can be achieved while still retaining basic object information in the scene.

In step S120, a ground point is determined from the point cloud according to the height of the point cloud in the point cloud data.

In one embodiment, the step of determining ground points from the point cloud includes: first, performing preliminary filtering on the point cloud data according to the height of the point cloud points in the point cloud data to determine at least part of the ground points; Then, perform plane fitting on the remaining point cloud points, and filter out the point cloud points whose distance from the fitted plane is greater than the threshold. After that, the remaining point cloud points are the ground points.

Further, referring to FIG. 4, the preliminary filtering may include the following two steps: In step S410, the horizontal plane of the point cloud space where the point cloud is located is divided into a plurality of grids, and according to the points in the grid For the lowest height of the cloud point, perform relative height filtering in each of the grids; and step S420, perform absolute height filtering according to the absolute height of the point cloud point.

It should be noted that if the above preprocessing is performed, the point cloud space involved in step S120 and step S130 is the region of interest in the point cloud space, and the point cloud points to be filtered are the point cloud after downsampling. point.

As shown in Figure 5, when performing relative height filtering, the horizontal plane of the point cloud space is first rasterized, that is, the horizontal plane is divided into meshed grids without dividing the vertical height. In each grid, there may be multiple point cloud points distributed at different heights (of course, some grids may not have point cloud points). For each grid, take the height of the point cloud point with the lowest height among all the point cloud points as the reference height of the grid, thereby obtaining the minimum height distribution map of the point cloud space; then based on the minimum height distribution map, Set a certain height threshold to filter out the point cloud points in each grid whose height difference between the reference height of the corresponding grid is greater than the threshold.

It can be understood that after the relative height filtering, the retained point cloud points are the point cloud points within a certain range above the minimum height distribution map.

After that, step S420 is executed to perform absolute height filtering based on the absolute height of all the point cloud points in the point cloud based on the point cloud points after the relative height filtering.

When performing absolute height filtering, filtering is no longer performed for each grid separately, but for the absolute height of all point cloud points in the point cloud space, so as to remove the point cloud points whose absolute height is greater than a certain threshold.

For example, suppose there is a certain platform-type obstacle in a road scene. When performing relative height filtering, the grid corresponding to the platform will use the height of the platform as the reference height, and the point cloud points collected by the platform and located above the platform will be used as the reference height. Point cloud points within a certain threshold range will not be filtered out. In the absolute height filtering process, since the judgment criterion is the absolute height of the point cloud point, this part of the point cloud point that does not belong to the ground point can be filtered out.

In one embodiment, since the ground may have a slope, the farther the distance from the distance measuring device is, the greater the possible slope of the ground. Therefore, when the absolute height filtering is performed, the distance between the point cloud point and the point cloud spatial coordinate origin (ie, and The distance between the distance measuring devices) determines the absolute height threshold, and filters out the point cloud points whose height is greater than the absolute height threshold corresponding to the distance according to the distance of the point cloud points. For example, the absolute height threshold of point cloud points within 10m from the distance measuring device is set to 0.4m, the absolute height threshold of point cloud points in the range of 10m to 40m is set to 1m, and the absolute height threshold of point cloud points outside 40m is set to 2m.

Afterwards, step S430 is performed to perform plane fitting on the remaining point cloud points after the relative height filtering and the absolute height filtering, and to filter out the point cloud points whose distance from the fitted plane is greater than the threshold.

Among them, the plane fitting can be performed by any suitable optimization algorithm such as least square method, and the plane obtained by the fitting is a plane close to the real terrain. If the remaining point cloud points after relative height filtering and absolute height filtering have a point cloud point whose distance from the plane exceeds the given point cloud point, the object corresponding to the point cloud point may deviate from the real ground, so it is filtered Divide, and accept the final remaining point cloud point as the ground point.

So far, the ground point in the point cloud has been determined through the above steps. In step S130, an above-ground point is determined from the point cloud according to the ground point. In one embodiment, after all the ground points are determined in step S120, the ground points are removed, and the remaining point cloud points are the ground points.

As an example, the front scenic spot may be a moving object. And, for different scenes, moving objects can also be different. As an example, if the scene is a road, the moving objects may be vehicles, or the moving objects may be vehicles and pedestrians. As another example, if the scene is a parking lot, the moving object may be a vehicle, for example, a car, a bicycle, and so on. Those skilled in the art can understand that for other scenes, the corresponding moving objects may be other moving objects in the scene.

As mentioned above, the target point can be a front scenic spot or a background point. When the target point is the former scenic spot, the former scenic spot is determined from the above-ground points according to the features of the former scenic spot, and the above-ground points remaining after filtering out the former scenic spot are the background points; when the target point is the background point, the ground points are taken from the ground according to the features of the background point. The background point is determined from the points, and the remaining above ground points after filtering out the background points are the former scenic spots. In the following, the description is mainly given by taking the target point as the background point as an example.

In an embodiment, as shown in FIG. 6, the determining the target point from the above-ground point according to the characteristics of the above-ground point includes:

In step S610, extract seed points in the above-ground points according to the characteristics of the above-ground points; perform regional growth based on the seed points to determine a preliminary target point;

In step S620, a target area is defined according to the preliminary target point, and an above-ground point in the target area is determined as a target point;

In step S630, cluster the ground points outside the target area to obtain multiple point cloud clusters;

In step S640, a target point cloud cluster in the point cloud cluster is determined according to the characteristics of the point cloud cluster, and an above-ground point in the target point cloud cluster is determined as the target point.

In step S610, in view of the problem of improper selection of seed points or inaccurate feature extraction and easy segmentation errors in the traditional region growing method, the embodiment of the present invention optimizes the selection criteria of seed points to make the segmentation more accurate. Wherein, the characteristics of the above-ground point may include at least one of the height, horizontal position, and reflectivity of the above-ground point.

In an embodiment, an above-ground point with a height higher than the first threshold may be used as the seed point. Wherein, the first threshold can be set according to actual needs. When the target point is a background point, the height of the first threshold can be set to be greater than the general height of the foreground object of interest. For example, in a road scene, the height of the first threshold can be higher than the general height of the vehicle. A threshold can be set to 5 meters. Further, the aboveground points with a height higher than the first threshold and lower than the second threshold may also be used as seed points. For example, aboveground points between 5 meters and 10 meters may be used as seed points for regional growth. When the target point is the previous scenic spot, the height range of the seed point can also be set according to the height of the previous scenic spot.

Since the height of the target object usually has certain rules, extracting seed points based on the height is more in line with the actual situation. Taking the target point as the background point as an example, taking the ground point whose height exceeds the general height of the foreground object as the seed point can ensure that the seed point is the background point, thereby avoiding the deviation of the extraction of the seed point.

In other embodiments, the seed point can also be selected according to the horizontal position or reflectivity of the ground point.

For example, in a road scene, the background points mainly include point cloud points collected for buildings on both sides of the road. When selecting the seed point, the ground point whose horizontal distance from the distance measuring device is greater than the width of the road can be used as the seed point. ; Alternatively, it is also possible to pre-record the reflectivity of background objects such as exterior walls of buildings and trees, and use point cloud points whose reflectivity is consistent with or close to the reflectivity of the background object as seed points.

As an example, after extracting the seed point, according to the preset growth rules, the above-ground points in the neighborhood of the seed point that have the same or similar properties as the seed point are merged into the area of the seed point, and the newly added above-ground point is used as the seed point Continue to grow, and so on, until there are no more ground points that meet the conditions around the seed point, at which point the growth stops. The growth rule of the area growth includes but is not limited to the curvature of the point cloud. After that, all the above-ground points in the area obtained by the area growth can be used as preliminary target points.

The above-mentioned method of region growth based on seed points can extract preliminary target points, but usually cannot extract all target points. Therefore, in one embodiment, after step S610, step S620-step S640 are continued, the target area is defined according to the preliminary target point, and the target point is further extracted.

As an example, when the target point is a background point, in step S620, after the preliminary background point is extracted by the above method, the area behind the preliminary background point may be divided into the background area.

Specifically, the step of defining the target area according to the preliminary target point includes: firstly, obtaining the change curve of the horizontal distance relative to the azimuth angle according to the azimuth angle and the horizontal distance of the preliminary target point relative to the coordinate origin; Smooth filtering is performed on the change curve, and an area on one side of the change curve is defined as the target area. Afterwards, all above-ground points in the target area can be determined as target points.

Wherein, when the target point is a background point, after the change curve is obtained, the area outside the change curve (that is, far from the coordinate origin) is defined as the background area, and all above-ground points in the background area can be determined as background points. When the target point is the front scenic spot, after obtaining the change curve, the area inside the change curve (that is, closer to the coordinate origin) is defined as the foreground area, and all target points in the foreground area can be determined as the front scenic spot.

Further, when obtaining the change curve of the horizontal distance relative to the azimuth angle according to the azimuth angle and the horizontal distance of the preliminary target point relative to the origin of the coordinates, when there are multiple preliminary target points on the same azimuth angle, change The horizontal distance corresponding to the azimuth angle is set as the horizontal distance of the preliminary target point closest to the coordinate origin; when the preliminary target point does not exist on a certain azimuth angle, the horizontal distance corresponding to the azimuth angle is set to infinity.

As an example, the above-mentioned smoothing and filtering of the change curve may include: for each azimuth angle, setting its corresponding horizontal distance to the minimum value of all horizontal distances in a certain neighborhood. The final change curve is a smoother change curve.

Simply put, the process of defining the target area can be regarded as the process of drawing the boundary between the background area and the foreground area. Taking the target point as the background point as an example, first extract the background point closest to the origin of the coordinate in each angle interval, and draw a change curve along all the extracted background points, then the area outside the change curve is the background area, and the upper The preliminary background points extracted in the article and some unextracted background points (for example, when extracting above-ground points higher than the first threshold value as seed points for regional growth, buildings cannot be grown due to their low height and relatively independent locations Point cloud) are located outside the curve.

Through the above steps, the target area and most of the target points are determined. After that, step S630 is performed to cluster the above-ground points outside the target area to obtain multiple point cloud clusters.

Here, the purpose of clustering the ground points to obtain point cloud clusters is to distinguish the point cloud points of different objects in the messy point cloud data, such as distinguishing point clouds belonging to the same vehicle, pedestrian, road marker, etc. point. As an example, the distance clustering method can be used to divide the above-ground points outside the target area into different point cloud clusters. That is to say, for any two ground points, if their distance is less than a given threshold, it is judged that they belong to the same point cloud cluster.

It should be noted that the clustering of above-ground points here is not limited to narrow clustering methods (for example, the above-mentioned distance-based distance, or other clustering methods such as k-means clustering, density-based clustering, etc.), Instead, it includes any suitable method of dividing point cloud clusters, such as model matching-based methods and neural network-based methods.

Further, step S630 may also include identifying the type of the object corresponding to each point cloud cluster. For example, the corresponding object type can be determined according to the shape of the outer contour of the point cloud cluster.

Among the point cloud clusters obtained by segmentation by the above method, there are mainly non-target point cloud clusters, but there may also be some point cloud clusters of target points. Therefore, step S640 may also be performed to determine the target point cloud cluster in the point cloud cluster according to the characteristics of the point cloud cluster, and determine the above-ground point in the target point cloud cluster as the target point.

Exemplarily, for each point cloud cluster, a series of features can be calculated, such as the number of points on the ground, the shape and size of the point cloud cluster, etc. Based on these features, the target point cloud cluster in the point cloud cluster can be further extracted. For example, taking the target point as the background point, in a road scene, if the shape of a point cloud cluster is slender, for example, its height is greater than 3 meters, and the width is only tens of centimeters, the point cloud cluster is likely to be The point cloud clusters of road poles can therefore be judged as background point cloud clusters; for another example, if the size of a point cloud cluster is greater than a certain threshold, for example, both length and width are greater than 10 meters, the point cloud cluster may be a building Point cloud cluster, so it can also be judged as background point cloud cluster, and so on.

If the background point cloud cluster is extracted from the point cloud cluster, on this basis, the target area can be redefined according to the ground points in the target point cloud cluster. Specifically, the point cloud points in the target point cloud cluster are used as the newly-added preliminary background points, step S620 is re-executed, the target area is redefined according to the updated preliminary target points, and the ground points in the redefined target area are determined As the target point.

So far, the ground points are divided into background points and different cloud clusters of front scenic spots, and each cloud cluster of front scenic spots corresponds to a foreground object, such as vehicles, pedestrians, etc.

It should be noted that the above description mainly uses the target point as the background point as an example to describe the above-ground point segmentation method according to the embodiment of the present invention, but the above-ground point segmentation in the embodiment of the present invention can also be performed with the previous scenic spot as the target point. Segmentation. At this time, only some of the specific segmentation criteria need to be adjusted adaptively. For example, when determining the target point cloud cluster in the point cloud cluster according to the characteristics of the point cloud cluster, the characteristics of the point cloud cluster should be consistent with the point cloud of the previous scenic spot The characteristics of the cluster, such as the characteristics of the vehicle point cloud cluster, etc.

In one embodiment, on the premise that the point cloud points are down-sampled, the above-mentioned segmentation method realizes the segmentation of the point cloud after down-sampling. Therefore, after that, the segmented point cloud data can also be up-sampled (or called up-sampling) processing to determine the segmentation category of each original point cloud point. That is to say, in the segmentation process, the down-sampling point cloud data is used as the object to segment, thereby greatly improving the segmentation efficiency; after the segmentation is completed, the up-sampling process is performed, and the original point cloud data can be segmented.

Specifically, the upsampling process mainly includes: firstly, traverse each point cloud point after segmentation, the category of the point cloud point is one of the ground point, the front scenic spot, and the background point; the category of the point cloud point is assigned to the The voxel corresponding to the point cloud point (refer to step S230). Then, each original point cloud point is traversed, and if it is outside the region of interest for the aforementioned downsampling, it is determined as a background point; otherwise, the point cloud point category is determined as the corresponding voxel category, that is In other words, the category of the original point cloud point in each voxel is consistent with the segmentation category of the point cloud point retained after downsampling in the same voxel. Thus, the segmentation category of each original point cloud point can be obtained.

In one embodiment, when performing upsampling processing, the original point cloud data is the original point cloud data after removing noise, that is, the noise removed in step S210 is not upsampled.

7A-7D are segmentation examples of a point cloud segmentation method according to an embodiment of the present invention. 7A is a schematic diagram of the original point cloud, FIG. 7B is a schematic diagram of the ground points segmented from FIG. 7A, FIG. 7C is a schematic diagram of the background points segmented from FIG. 7A, and FIG. 7D is a schematic diagram of the background points segmented from FIG. 7A Schematic diagram of the former attractions. Fig. 7E is an enlarged view of Fig. 7D, which shows the segmented point cloud clusters of foreground objects such as vehicles and pedestrians.

After that, the point cloud data can be subsequently processed based on the segmentation result. For example, in a road scene, the point cloud clusters of the segmented vehicles can be extracted, and the running status of surrounding vehicles can be tracked and monitored accordingly, so as to provide a key foundation for the realization of intelligent driving, intelligent transportation and other applications. In addition, all previous scenic spots can also be eliminated, and only the background points and ground points in each frame of point cloud data can be retained for simultaneous localization and map construction (simultaneous localization and mapping, SLAM). Eliminating the front sights can prevent moving foreground objects from interfering with SLAM.

In summary, the point cloud segmentation method of the embodiment of the present invention can quickly and effectively segment point cloud data into ground points, front scenic spots, and background points. The algorithm is simple and easy to implement; and the method of the embodiment of the present invention can be applied to Different types of lidars, including lidars with regular repetitive scanning and lidars with complex scanning trajectories with non-repetitive scanning characteristics, do not rely on the scanning trajectory information of the lidar, and have good versatility.

FIG. 8 shows a schematic flowchart of a point cloud segmentation method 800 according to an embodiment of the present invention. As shown in FIG. 8, the method 800 includes the following steps:

Step S810: Obtain point cloud data;

Step S820: Divide the horizontal plane of the point cloud space in the point cloud data into a plurality of grids, and perform processing in each grid according to the lowest height of the point cloud point in the grid. Relatively high filtering;

Step S830, performing absolute height filtering according to the absolute height of the point cloud point;

Step S840: Perform plane fitting on the point cloud points after the relative height filtering and the absolute height filtering, and determine the ground point and the above ground point according to whether the distance between the point cloud point and the fitted plane is greater than a threshold;

Step S850: Extract seed points from the above-ground points according to the characteristics of the above-ground points, and perform regional growth based on the seed points to determine a preliminary target point;

Step S860: Define a target area according to the preliminary target point, and determine an above-ground point in the target area as a target point;

Step S870, clustering the ground points outside the target area to obtain multiple point cloud clusters;

Step S880: Determine a target point cloud cluster in the point cloud cluster according to the characteristics of the point cloud cluster, and determine an above-ground point in the target point cloud cluster as the target point.

For the specific details of step S810 to step S880, reference may be made to the related description of the point cloud segmentation method 100 described above, which will not be repeated here.

FIG. 9 shows a schematic flowchart of a point cloud segmentation method 900 according to another embodiment of the present invention. As shown in FIG. 9, the method 900 includes the following steps:

In step S910, point cloud data is acquired;

In step S920, extract seed points in the point cloud data according to the characteristics of the point cloud points in the point cloud data;

In step S930, region growth is performed based on the seed point to determine a preliminary target point in the point cloud data.

Exemplarily, in step S910, the point cloud data may be point cloud data collected by a distance measuring device, the distance measuring device may be a lidar, and the lidar may be a regular and repeated scanning lidar , It can also be a lidar with complex scanning trajectory with non-repetitive scanning characteristics. The point cloud data may be ground points extracted from the original point cloud data, and the point cloud data may also be original point cloud data or a point cloud after noise filtering, coordinate transformation, and downsampling are performed on the original point cloud data. For data, refer to the above for details.

In step S920, in view of the problem of improper selection of seed points or inaccurate feature extraction and segmentation errors in the traditional region growing method, the embodiment of the present invention optimizes the selection criteria of seed points to make the segmentation more accurate. The feature of the point cloud point may include at least one of the height, horizontal position, and reflectivity of the point cloud point.

In an embodiment, a point cloud point with a height higher than a first threshold may be used as the seed point. Wherein, the first threshold can be set according to actual needs. When the target point is a background point, the height of the first threshold can be set to be greater than the general height of the foreground object of interest. For example, in a road scene, the height of the first threshold can be higher than the general height of the vehicle. A threshold can be set to 5 meters. Further, point cloud points with a height higher than the first threshold and lower than the second threshold may also be used as seed points. For example, point cloud points between 5 meters and 10 meters may be used as seed points for regional growth. When the target point is the previous scenic spot, the height range of the seed point can also be set according to the height of the previous scenic spot.

Since the height of the target object usually has certain rules, extracting seed points based on the height is more in line with the actual situation. Taking the target point as the background point as an example, taking the point cloud point whose height exceeds the general height of the foreground object as the seed point can ensure that the seed point is the background point, thereby avoiding the deviation of the extraction of the seed point.

In other embodiments, the seed point can also be selected according to the horizontal position or reflectivity of the point cloud point.

For example, in a road scene, background points mainly include point cloud points collected for buildings on both sides of the road. When selecting seed points, point cloud points whose horizontal distance from the distance measuring device is greater than the width of the road can be used as seeds Point; or, you can also pre-record the reflectivity of background objects such as building walls, trees, etc., and use point cloud points whose reflectivity is the same or close to that of the background object as the seed point.

After extracting the seed point, in step S930, according to the preset growth rules, the point cloud points in the neighborhood of the seed point that have the same or similar properties as the seed point are merged into the area of the seed point, and the newly added point cloud The point as the seed point continues to grow, and so on, until there are no more point cloud points that meet the conditions around the seed point, and the growth stops at this time. The growth rule of the area growth includes but is not limited to the curvature of the point cloud. After that, all the point cloud points in the area obtained by the area growth can be used as preliminary target points.

The above-mentioned method of region growth based on seed points can extract preliminary target points, but usually cannot extract all target points. Therefore, in one embodiment, after step S930, the method 900 may further include: defining a target area according to the preliminary target point, and further extracting the target point.

As an example, when the target point is a background point, after the preliminary background point is extracted by the above method, the area behind the preliminary background point can be divided into the background area.

Specifically, the step of defining the target area according to the preliminary target point includes: firstly, obtaining the change curve of the horizontal distance relative to the azimuth angle according to the azimuth angle and the horizontal distance of the preliminary target point relative to the coordinate origin; Smooth filtering is performed on the change curve, and an area on one side of the change curve is defined as the target area. After that, all the point cloud points in the target area can be determined as target points.

Wherein, when the target point is a background point, after obtaining the change curve, the area outside the change curve (that is, far from the coordinate origin) is defined as the background area, and the point cloud points in the background area can be determined as the background point. When the target point is the front scenic spot, after obtaining the change curve, the area inside the change curve (that is, closer to the coordinate origin) is defined as the foreground area, and all target points in the foreground area can be determined as the front scenic spot.

Simply put, the process of defining the target area can be regarded as the process of drawing the boundary between the background area and the foreground area. Taking the target point as the background point as an example, first extract the background point closest to the origin of the coordinate in each angle interval, and draw a change curve along all the extracted background points, then the area outside the change curve is the background area, and the upper The preliminary background points extracted in the article and some unextracted background points (for example, when extracting point cloud points higher than the first threshold value as seed points for regional growth, the building fails to grow due to its low height and relatively independent location The point cloud of the object is located outside the curve.

Through the above steps, the target area and most of the target points are determined. Afterwards, the point cloud points outside the target area may also be clustered to obtain multiple point cloud clusters.

Here, the purpose of clustering point cloud points to obtain point cloud clusters is to distinguish the point cloud points of different objects in the cluttered point cloud data, for example, to distinguish points belonging to the same vehicle, pedestrian, road marker, etc. Cloud point. As an example, the distance clustering method can be used to divide the point cloud points outside the target area into different point cloud clusters. That is to say, for any two point cloud points, if their distance is less than a given threshold, it is judged that they belong to the same point cloud cluster.

It should be noted that the clustering of point cloud points here is not limited to narrow clustering methods (for example, the above-mentioned distance-based distance, or other clustering methods such as k-means clustering, density-based clustering, etc.) , But includes any suitable method of dividing point cloud clusters, such as method based on model matching and method based on neural network.

Further, the method 900 may further include identifying the type of the object corresponding to each point cloud cluster. For example, the corresponding object type can be determined according to the shape of the outer contour of the point cloud cluster.

Among the point cloud clusters obtained by segmentation by the above method, there are mainly non-target point cloud clusters, but there may also be some point cloud clusters of target points. Therefore, it is also possible to determine the target point cloud cluster in the point cloud cluster according to the characteristics of the point cloud cluster, and determine the point cloud point in the target point cloud cluster as the target point. Exemplarily, for each point cloud cluster, a series of features can be calculated, such as the number of point cloud points, the shape and size of the point cloud cluster, etc. Based on these features, the target point cloud cluster in the point cloud cluster can be further extracted .

If the background point cloud cluster is extracted from the point cloud cluster, on this basis, the target area can also be redefined according to the point cloud points in the target point cloud cluster. Specifically, the point cloud points in the target point cloud cluster are used as the newly-added preliminary background points, the target area is redefined according to the updated preliminary target points, and the point cloud points in the redefined target area are determined as the target points .

So far, the point cloud points are divided into background points and different front scenic spot cloud clusters, and each front scenic spot cloud cluster corresponds to a foreground object, such as vehicles, pedestrians, etc.

It should be noted that the above description mainly uses the target point as the background point as an example to describe the point cloud point segmentation method according to the embodiment of the present invention, but the point cloud point segmentation in the embodiment of the present invention can also use the previous scenic spot as the target point. To perform segmentation, only some of the specific segmentation criteria need to be adjusted adaptively. For example, when determining the target point cloud cluster in the point cloud cluster according to the characteristics of the point cloud cluster, the characteristics of the point cloud cluster should be consistent with the previous scenic spot The characteristics of point cloud clusters, such as the characteristics of vehicle point cloud clusters, and so on.

The point cloud segmentation method 900 of the embodiment of the present invention proposes a target point extraction scheme, which is simple and effective, has a small amount of calculation, and is easy to implement.

FIG. 10 shows a schematic block diagram of a point cloud segmentation system 1000 in an embodiment of the present invention.

As shown in FIG. 10, the point cloud segmentation system 1000 includes one or more processors 1010 and one or more memories 1020. Optionally, the point cloud segmentation system 1000 may further include at least one of an input device (not shown), an output device (not shown), and an image sensor (not shown). These components are connected through a bus system and/or other forms. The connecting mechanism (not shown) is interconnected. It should be noted that the components and structure of the point cloud segmentation system 1000 shown in FIG. 10 are only exemplary and not restrictive. According to needs, the point cloud segmentation system 1000 may also have other components and structures, for example, Transceiver that sends and receives signals.

The memory 1020 is also a memory for storing processor-executable instructions, for example, for storing corresponding steps and program instructions in the point cloud segmentation method according to an embodiment of the present invention. It may include one or more computer program products, and the computer program products may include various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory. The volatile memory may include random access memory (RAM) and/or cache memory (cache), for example. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, and the like.

The input device may be a device used by a user to input instructions, and may include one or more of a keyboard, a mouse, a microphone, and a touch screen.

The output device may output various information (for example, images or sounds) to the outside (for example, a user), and may include one or more of a display, a speaker, and the like.

The communication interface (not shown) is used for communication between the point cloud segmentation system 1000 and other devices, including wired or wireless communication. The point cloud segmentation system 1000 can access a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G, 5G, or a combination thereof. In an exemplary embodiment, the communication interface receives a broadcast signal or broadcast related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication interface further includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

The processor 1010 may be a central processing unit (CPU), an image processing unit (GPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other forms with data processing capabilities and/or instruction execution capabilities It can control other components in the point cloud segmentation system 1000 to perform desired functions. The processor can execute the instructions stored in the memory 1020 to execute the point cloud segmentation method described herein. For example, the processor 1010 can include one or more embedded processors, processor cores, microprocessors, logic circuits, hardware finite state machines (FSM), digital signal processors (DSP), or combinations thereof.

One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 1010 may run the program instructions stored in the memory 1020 to implement the embodiments of the present invention described herein (implemented by the processor) And/or other desired functions, for example, to perform the corresponding steps of the point cloud segmentation method according to the embodiment of the present invention. Various application programs and various data, such as various data used and/or generated by the application program, can also be stored in the computer-readable storage medium.

In addition, the embodiment of the present invention also provides a computer storage medium on which a computer program is stored. When the computer program is executed by the processor, each step of the point cloud segmentation method of the embodiment of the present invention can be implemented. For example, the computer storage medium may include, for example, a memory card of a smart phone, a storage component of a tablet computer, a hard disk of a personal computer, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disk Read only memory (CD-ROM), USB memory, or any combination of the above storage media. The computer-readable storage medium may be any combination of one or more computer-readable storage media.

In the above-mentioned embodiments, it may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, it can be implemented in the form of a computer program product in whole or in part. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on the computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website, computer, server, or data center. Transmission to another website, computer, server, or data center via wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center integrated with one or more available media. The usable medium may be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a digital video disc (DVD)), or a semiconductor medium (for example, a solid state disk (SSD)), etc. .

Although the exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above-described exemplary embodiments are merely exemplary, and are not intended to limit the scope of the present invention thereto. Those of ordinary skill in the art can make various changes and modifications therein without departing from the scope and spirit of the present invention. All these changes and modifications are intended to be included within the scope of the present invention as claimed in the appended claims.

A person of ordinary skill in the art may realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software depends on the specific application and design constraint conditions of the technical solution. Professionals and technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered as going beyond the scope of the present invention.

In the several embodiments provided in this application, it should be understood that the disclosed device and method may be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components may be combined or It can be integrated into another device, or some features can be ignored or not implemented.

In the instructions provided here, a lot of specific details are explained. However, it can be understood that the embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures, and technologies are not shown in detail, so as not to obscure the understanding of this specification.

Similarly, it should be understood that in order to simplify the present invention and help understand one or more of the various aspects of the invention, in the description of the exemplary embodiments of the present invention, the various features of the present invention are sometimes grouped together into a single embodiment. , Or in its description. However, the method of the present invention should not be construed as reflecting the intention that the claimed invention requires more features than those explicitly stated in each claim. To be more precise, as reflected in the corresponding claims, the point of the invention is that the corresponding technical problems can be solved with features that are less than all the features of a single disclosed embodiment. Therefore, the claims following the specific embodiment are thus explicitly incorporated into the specific embodiment, wherein each claim itself serves as a separate embodiment of the present invention.

Those skilled in the art can understand that in addition to mutual exclusion between the features, any combination of all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and any method or device disclosed in this manner can be used. Processes or units are combined. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

In addition, those skilled in the art can understand that although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments means that they are within the scope of the present invention. Within and form different embodiments. For example, in the claims, any one of the claimed embodiments can be used in any combination.

The various component embodiments of the present invention may be implemented by hardware, or by software modules running on one or more processors, or by a combination of them. Those skilled in the art should understand that a microprocessor or a digital signal processor (DSP) may be used in practice to implement some or all of the functions of some modules according to the embodiments of the present invention. The present invention can also be implemented as a device program (for example, a computer program and a computer program product) for executing part or all of the methods described herein. Such a program for realizing the present invention may be stored on a computer-readable medium, or may have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

It should be noted that the above-mentioned embodiments illustrate rather than limit the present invention, and those skilled in the art can design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be constructed as a limitation to the claims. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims that list several devices, several of these devices may be embodied in the same hardware item. The use of the words first, second, and third, etc. do not indicate any order. These words can be interpreted as names.

Claims

A point cloud segmentation method, characterized in that the method includes:

Obtain point cloud data;

Divide the horizontal plane of the point cloud space in the point cloud data into multiple grids, and perform relative height filtering in each grid according to the lowest height of the point cloud points in the grid ；

Performing absolute height filtering according to the absolute height of the point cloud point;

Perform plane fitting on the point cloud points after the relative height filtering and the absolute height filtering, and determine the ground point and the above ground point according to whether the distance between the point cloud point and the fitted plane is greater than a threshold;

Extract seed points from the above-ground points according to the characteristics of the above-ground points, and perform regional growth based on the seed points to determine a preliminary target point;

Define a target area according to the preliminary target point, and determine an above-ground point in the target area as a target point;

Clustering the ground points outside the target area to obtain multiple point cloud clusters;

According to the characteristics of the point cloud cluster, a target point cloud cluster in the point cloud cluster is determined, and an above-ground point in the target point cloud cluster is determined as the target point.
A point cloud segmentation method, characterized in that the method includes:

Obtain point cloud data;

Determining a ground point from the point cloud according to the height of the point cloud in the point cloud data;

Determining an above-ground point from the point cloud according to the ground point;

A target point is determined from the above-ground points according to the characteristics of the above-ground point, and the target point is a front scenic spot or a background point.
The method according to claim 2, wherein the determining the target point from the above-ground point according to the characteristics of the above-ground point comprises:

Extracting seed points from the above-ground points according to the characteristics of the above-ground points;

Region growth is performed based on the seed point to determine a preliminary target point.
The method according to claim 3, wherein the characteristics of the above-ground point include at least one of the following: height, horizontal position, and reflectivity of the above-ground point.
The method according to claim 4, wherein extracting the seed point according to the height of the above-ground point comprises:

An aboveground point whose height is higher than the first threshold is used as the seed point.
The method according to claim 5, wherein extracting the seed point according to the height of the above-ground point comprises:

An aboveground point whose height is higher than the first threshold and lower than the second threshold is used as the seed point.
The method according to any one of claims 3-6, wherein the determining the target point from the above-ground points according to the characteristics of the above-ground points, further comprises:

Define a target area according to the preliminary target point;

The above-ground point in the target area is determined as the target point.
The method according to claim 7, wherein the defining a target area according to the preliminary target point comprises:

Obtaining a change curve of the horizontal distance relative to the azimuth angle according to the azimuth angle and the horizontal distance of the preliminary target point relative to the coordinate origin;

Smooth filtering is performed on the change curve, and an area outside the change curve is defined as the target area.
The method according to claim 8, wherein the obtaining the change curve of the horizontal distance relative to the azimuth angle according to the azimuth angle and the horizontal distance of the preliminary target point relative to the origin of the coordinate comprises:

When there are multiple preliminary target points on the same azimuth angle, the horizontal distance corresponding to the azimuth angle is set as the horizontal distance of the preliminary target point closest to the coordinate origin;

When the preliminary target point does not exist on a certain azimuth angle, the horizontal distance corresponding to the azimuth angle is set to infinity.
The method according to claim 8 or 9, wherein the smoothing and filtering the change curve comprises:

For each azimuth angle, the corresponding horizontal distance is set as the minimum value among all horizontal distances in a certain neighborhood.
The method according to any one of claims 7-10, wherein the determining a target point from the above-ground point according to the characteristics of the above-ground point further comprises:

Clustering the ground points outside the target area to obtain multiple point cloud clusters;

According to the characteristics of the point cloud cluster, a target point cloud cluster in the point cloud cluster is determined, and an above-ground point in the target point cloud cluster is determined as a target point.
The method according to claim 11, wherein the determining the target point from the above-ground point according to the characteristics of the above-ground point further comprises:

Redefining the target area according to the above-ground points in the target point cloud cluster.
The method according to claim 11, wherein the characteristics of the point cloud cluster include the number of above-ground points in the point cloud cluster, the size of the point cloud cluster, and/or the shape of the point cloud cluster .
The method according to any one of claims 3-13, wherein the target point is the background point, the target area is a background area, and the target point cloud cluster is a point cloud cluster of background points .
The method according to any one of claims 3-13, wherein the target point is the front scenic spot, the target area is a foreground area, and the target point cloud cluster is a point cloud cluster of the previous scenic spot .
The method according to claim 1, wherein the determining a ground point from the point cloud according to the height of the point cloud in the point cloud data comprises:

Performing preliminary filtering on the point cloud according to the height of the point cloud points in the point cloud to determine at least part of the ground points;

Perform plane fitting on the at least part of the ground, and filter out point cloud points whose distance from the fitted plane is greater than a threshold.
The method according to claim 16, wherein the preliminary filtering comprises:

Dividing the horizontal plane of the point cloud space where the point cloud is located into multiple grids;

Performing relative height filtering in each grid according to the lowest height of the point cloud point in the grid;

Absolute height filtering is performed according to the absolute height of the point cloud point.
The method according to claim 17, wherein the relative height filtering comprises:

Determining the lowest height of the point cloud point in the grid;

Filter out the point cloud points whose height difference between the grid and the lowest height is greater than a threshold.
The method according to claim 17, wherein the absolute height filtering comprises:

Determine the absolute height threshold according to the distance between the point cloud point and the origin of the point cloud space coordinates;

The point cloud points whose height is greater than the corresponding absolute height threshold are filtered out.
The method according to claim 2, wherein said acquiring point cloud data comprises:

Obtain the original point cloud data;

The original point cloud data is preprocessed to obtain preprocessed point cloud data.
The method according to claim 20, wherein the preprocessing comprises:

Remove the noise in the original point cloud data.
The method according to claim 21, wherein the removing noise in the original point cloud data comprises:

Count the number of adjacent point cloud points in the neighborhood of the original point cloud point in the original point cloud data in the point cloud space;

The original point cloud point whose number of adjacent point cloud points is less than the threshold is judged as a noise point, and the noise point is removed.
The method according to claim 21, wherein the removing noise in the original point cloud data comprises:

Projecting the original point cloud data onto the projection surface;

Counting the number of adjacent projections of the original point cloud points on the projection surface in the neighborhood;

The original point cloud points whose number of adjacent projections are less than the threshold are judged as noise points, and the noise points are removed.
The method according to claim 20, wherein the preprocessing comprises:

Perform coordinate transformation on the original point cloud data so that the coordinate system of the point cloud space is parallel to the ground.
The method according to claim 20, wherein the preprocessing comprises:

Down-sampling the original point cloud data.
The method according to claim 25, wherein the downsampling comprises:

Selecting a region of interest in the point cloud space where the point cloud is located in the original point cloud data;

Dividing the region of interest into a voxel matrix composed of multiple voxels;

For each voxel in the voxel matrix, if it contains multiple original point cloud points, one of the original point cloud points is retained.
The method according to claim 26, characterized in that, after dividing the above-ground point into a front scenic spot and a background point, the method further comprises:

Upsampling processing is performed according to the segmentation category of the point cloud point obtained by the segmentation to determine the segmentation category of the original point cloud point in the original point cloud data.
The method according to claim 27, wherein the upsampling processing comprises:

Defining the segmentation category of the original point cloud point that belongs to the same voxel as the point cloud point in the region of interest as the segmentation category of the point cloud point;

The segmentation category of the original point cloud points outside the region of interest is defined as background points.
The method according to claim 27 or 28, wherein the original point cloud data is original point cloud data after noise points have been removed.
The method according to claim 2, further comprising:

Remove the former scenic spot in the point cloud;

Synchronous positioning and map construction are performed according to the point cloud data after excluding the front scenic spots.
A point cloud segmentation method, characterized in that the method includes:

Obtain point cloud data;

Extracting seed points in the point cloud data according to the characteristics of the point cloud points in the point cloud data;

Perform region growth based on the seed point to determine a preliminary target point in the point cloud data.
The method according to claim 31, wherein the feature of the point cloud point comprises the height, horizontal position and/or reflectivity of the point cloud point.
The method according to claim 32, wherein extracting the seed point in the point cloud data according to the height of the point cloud point comprises:

A point cloud point with a height higher than the first threshold is used as the seed point.
The method according to claim 33, wherein extracting the seed point in the point cloud data according to the height of the point cloud point comprises:

A point cloud point whose height is higher than the first threshold and lower than the second threshold is used as the seed point.
The method according to any one of claims 31-34, further comprising:

Define a target area according to the preliminary target point;

The above-ground point in the target area is determined as the target point.
The method according to claim 35, wherein the defining a target area according to the preliminary target point comprises:

Obtaining a change curve of the horizontal distance relative to the azimuth angle according to the azimuth angle and the horizontal distance of the preliminary target point relative to the coordinate origin;

Smooth filtering is performed on the change curve, and an area outside the change curve is defined as the target area.
The method according to claim 36, wherein the obtaining the change curve of the horizontal distance relative to the azimuth angle according to the azimuth angle and the horizontal distance of the preliminary target point relative to the origin of the coordinate comprises:

When there are multiple preliminary target points on the same azimuth angle, the horizontal distance corresponding to the azimuth angle is set as the horizontal distance of the preliminary target point closest to the coordinate origin;

When the preliminary target point does not exist on a certain azimuth angle, the horizontal distance corresponding to the azimuth angle is set to infinity.
The method according to claim 36 or 37, wherein the smoothing and filtering the change curve comprises:

For each azimuth angle, the corresponding horizontal distance is set as the minimum value among all horizontal distances in a certain neighborhood.
The method according to any one of claims 35-38, further comprising:

Clustering the point cloud points outside the target area to obtain multiple point cloud clusters;

According to the characteristics of the point cloud cluster, a target point cloud cluster in the point cloud cluster is determined, and a point cloud point in the target point cloud cluster is determined as a target point.
The method according to claim 39, further comprising:

Redefine the target area according to the point cloud points in the target point cloud cluster.
The method according to claim 39, wherein the characteristics of the point cloud cluster include the number of point cloud points in the point cloud cluster, the size of the point cloud cluster, and/or the size of the point cloud cluster. shape.
The method according to any one of claims 31-41, wherein the target point is the background point, the target area is a background area, and the target point cloud cluster is a point cloud cluster of background points .
The method according to any one of claims 31-41, wherein the target point is the front scenic spot, the target area is a foreground area, and the target point cloud cluster is a point cloud cluster of the previous scenic spot .
A point cloud segmentation system, characterized in that the system includes:

Memory, used to store executable instructions;

The processor is configured to execute the instructions stored in the memory, so that the processor executes the point cloud segmentation method according to any one of claims 1 to 43.
A computer storage medium having a computer program stored thereon, wherein the program is executed by a processor to implement the point cloud segmentation method according to any one of claims 1 to 43.