CN115439526A - Tree diameter measurement method based on halcon binocular stereo vision - Google Patents

Abstract

A method for measuring tree diameter at breast height based on halcon binocular stereo vision disclosed by the invention belongs to the technical field of artificial intelligence and includes the following steps: step S1, calibrate the binocular camera; step S2, perform binocular correction according to the calibration parameters; step S3, perform stereo matching and disparity map calculation on the image; step S4, preprocess the collected image; step S5, perform tree target detection training based on deep learning; step S6, measure the distance between the intersection points on both sides according to the edge of the tree , to obtain the tree diameter at breast height. The method for measuring tree diameter at breast height of the present invention collects images of artificial forest trees through binocular cameras, and obtains the size of tree diameter at breast height through deep learning and binocular vision technology, which reduces manual workload, improves measurement accuracy at the same time, and avoids losses due to errors.

Description

Tree diameter measurement method based on halcon binocular stereo vision

technical field

The invention relates to the field of artificial intelligence automatic size measurement, in particular to a tree diameter measurement method based on halcon binocular stereo vision.

Background technique

In the process of planting plantations, it is very necessary to timely and fully grasp the information of tree diameter at breast height. The data information can directly reflect the growth rate of trees and the volume of trees. important indicators.

The collection of DBH data of traditional forest plantations mostly relies on manual operations, mainly using calipers, calipers, diameter tapes and other instruments for manual measurement and recording, which is extremely inefficient, labor-intensive, and poor in accuracy. The use of deep learning and computer vision technology is of positive significance in the field of tree diameter information collection. It can not only meet the needs of fast, efficient, accurate and automatic collection at the technical level, improve the efficiency of tree diameter measurement, but also promote the development of plantation forest industry. Efficient development and high-quality production have significant practical significance, further liberating the labor force and reducing labor costs.

Contents of the invention

The technical problem to be solved by the present invention is: to overcome the deficiencies of the prior art, to provide an image collected by a binocular camera, and to automatically obtain the diameter of the forest tree after processing the image and deep learning, reducing manpower At the same time, it improves the measurement efficiency and accuracy, and avoids the loss caused by measurement errors. The tree diameter measurement method based on halcon binocular stereo vision.

In order to achieve the stated purpose, a tree diameter measurement method based on halcon binocular stereo vision is provided, including:

S1, binocular camera calibration

Calibrate the internal and external parameters of the binocular camera through the dot-type calibration plate, obtain the calibration parameters based on the halcon nonlinear model handle calibration method, correct the image according to the calibration parameters, and use the Block Matching algorithm for stereo matching and disparity map calculation;

S2. Deep learning and model training

(1) Preprocessing the image collected by the binocular camera, including using the gen_rectangle operator to segment the left and right images, using smooth operators or mean operators to smooth the image, removing noise, and using operators such as invert_image and scale_image to enhance the image, And carry out image annotation to complete the construction of deep learning image data set;

(2) Carry out tree target detection training based on the halcon deep learning network framework, and obtain the optimal model by adjusting parameters;

S3, tree diameter statistics

(1) Use the read_cam_par and read_pose operators to read the camera internal parameters of the left and right cameras obtained by binocular stereo vision calibration and the pose of the right camera relative to the left camera;

(2) Use the gen_binocular_rectification_map operator to obtain the transformation matrix between the non-standard outer limit image and the standard outer limit image, that is, the mapped image. Perform image correction on the left and right camera images according to the transformation matrix;

(3) Use the read_dl_model operator to introduce the trained tree DBH detection model, and use the parameter setting set_dl_model_param operator to set the corresponding parameters;

(4) Use list_files and tuple_regexp_select operators to open the camera or read the image, and the image reading type can be set to various types such as bmp, gif, jpg, etc.;

(5) Use gen_dl_samples_from_images and preprocess_dl_samples operators to detect the left and right images respectively, and obtain the boundary area below the DBH of the forest tree after threshold segmentation through the morphological processing method, select the coordinate positions of the upper left and upper right target points on the boundary, and define the difference in abscissa is the breast diameter size;

The tree image target detected is specifically:

The gray_features(RectangleR,V,“mean”,ValueR) operator is used to calculate the gray feature value of the specified area. The input is a set of areas, and the features of each area are stored in a set of value arrays, so as to obtain the read According to the threshold(B,RegionB,L1,255) and threshold(H,RegionH,L2,255) operators, the gray value information of the image area is obtained, and then the two boundary thresholds of the trunk diameter and the following area are determined L1:=ValueR/1.3, L2:=ValueR/1.7, finally divide the two channels, and then take the union to obtain the target detection area;

(6) Using the open_framegrabber operator, in the loop body, read in the forest images sequentially from the image data, select the first two frames of images at the beginning; keep an image later each time, read in a new image, overwrite An image earlier in time;

(7) Calculation of mean diameter at breast height

Use tuple_mean and disp_message operators, and define All:=[All,Mean] to count the number of recognition times. If the number of recognition times meets the specified setting requirements, calculate the mean value of the tree diameter at breast height, otherwise the recognition will be considered invalid, and the diameter at breast height will not be counted;

(8) Detect effectively identified trees, when the target trees disappear at the edge of the field of view and obtain an effective DBH value, the corresponding tree quantity statistics counter adds 1;

(9) Use the open_file and fwrite_string operators to export statistical values from the Excel table, mainly including the forest tree number counter value and the corresponding forest tree DBH average value.

Preferably, said step S1 includes the following steps:

The binocular camera calibration uses the halcon calibration data model. During the collection process, after the previous calibration image is collected, the direction of the collected image is rotated, and then the next calibration image is collected until all calibration image sets are obtained. During the collection of a single calibration image Controls stay still to ensure there is no movement, avoiding blur and sync issues.

Preferably, said step S2 includes the following steps:

Step (1), the image enhancement is mainly the adjustment of contrast and brightness, and the image annotation utilizes the MVTecDeep Learning Tool annotation tool, and the annotation mode is mainly for the chest diameter and the following area rectangular frame annotation;

In step (2), the model network is SqueezeNet, and its corresponding operator is pretrained_dl_classifier_compact.hdl, and the main parameters to be adjusted are batch-size (batch size), iterations (number of iterations) and learning_rate (learning rate) three parameters .

Preferably, said step S3 includes the following steps:

In steps (1), (2), and (3), the parameters, transformation matrix and tree diameter at breast height detection model need to be placed in a fixed position of the computer for reading;

In step (4), the read image is a color image;

In step (5), the rectangular frame of tree diameter at breast height is obtained according to the labeling method, and the detection area is obtained according to the gray value information and the two boundary threshold settings.

In step (7) of step S3 of the present invention, the counting of forest tree identification times is to prevent misidentification. The same forest tree will be photographed ten times. If the number of recognition times is less than six times, the image is generally not a forest tree as the main body. .

Compared with prior art, the beneficial effect that the present invention has is:

1. Collect images of artificial forest trees through binocular cameras, and obtain the size of tree diameter at breast height through deep learning and binocular vision technology, which reduces manual workload, improves measurement accuracy, and avoids losses caused by measurement errors ;

2. Using the deep learning convolutional neural network model, while ensuring no loss of accuracy, it is more efficient and fast to achieve the target detection task and obtain the target rectangular frame area;

3. Determine the two boundary thresholds of the diameter of the trunk and the following area through the acquisition of the gray value of the area. Finally, segment the two channels and then take the union, which can make the edge contour to be extracted more obvious and enhance the discrimination of the image. It is possible to reduce the influence of other background environmental factors on the measured object;

4. Using binocular stereo vision technology for distance measurement, there is no limit to the recognition rate compared with monocular vision, because in principle, it is not necessary to first identify and then measure, but directly measure the target recognition object. Binocular stereo vision is more accurate than monocular vision, and directly uses parallax to calculate distance;

5. When measuring the diameter at breast height of forest images, it can process a large amount of data and automatically obtain statistical results. For some non-forest targets such as shrubs and obstacles, the number of recognition times is insufficient, and actual statistics and diameter measurement are not performed to avoid misidentification and mismeasurement.

The description is only an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention, it can be implemented according to the contents of the description, and in order to make the above and other purposes, features and advantages of the present invention more obvious It is easy to understand that the specific embodiments of the present invention are given below.

Description of drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiment. The drawings are only for the purpose of illustrating a preferred embodiment and are not to be considered as limiting the invention. Also throughout the drawings, the same stage is denoted by the same reference numeral.

Fig. 1 is a hardware schematic diagram of a tree diameter measurement method based on halcon binocular stereo vision according to the present invention.

Fig. 2 is a flowchart of a tree diameter measurement method based on halcon binocular stereo vision according to the present invention.

detailed description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided for more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

Referring to FIG. 1 , the present embodiment provides a device schematic diagram of a tree diameter measurement method based on halcon binocular stereo vision, including a tree image acquisition platform, a forest image processing platform, and a tree diameter measurement display platform. The forest tree image acquisition platform is responsible for image acquisition and is connected to the industrial computer; the forest tree image processing platform is embedded with halcon processing software to process the collected image data to obtain the DBH data of the collected images; the forest tree DBH measurement and display platform is mainly responsible for image collection Real-time monitoring with image processing for human-computer interaction and display results, which mainly includes a display screen connected to an industrial computer.

The forest image acquisition platform given in this embodiment can realize real-time image acquisition through the combination of infrared sensor and binocular camera through the setting of the development board program, and set the acquisition time for the interval of forest trees to reach the specified number of images collected by each tree, and then pass halcon automatically creates a folder on the industrial computer, imports the collected forest images, and makes a naming distinction. Among them, the infrared sensor 1 is triggered by the tree perception and transmits instructions to the industrial computer. At this time, the binocular camera takes image shooting. When the infrared sensor 2 is triggered by the perception, the binocular camera ends the shooting. , use the transmission line to transmit the number to the industrial computer, and start the binocular camera after processing by the industrial computer. When the binocular camera starts to take pictures, collect photos and start number storage, the acquisition frequency and time interval are controlled by the touch screen and the industrial computer. Infrared After the sensor 2 is sensed and triggered, it transmits a signal to the industrial computer, and the industrial computer controls the binocular camera to finish collecting photos after processing.

The core algorithm of the industrial computer in the tree image processing platform given in this embodiment is the SqueezeNet convolutional neural network model based on the halcon target detection framework, which is used to obtain the target detection and recognition rectangular frame, so as to obtain the forest tree diameter at breast height and the trunk position area below, combined with gray_features and the threshold operator to obtain the standard tree diameter at breast height and the boundary contour of the following areas.

The display screen in the tree diameter measurement and display platform given in this embodiment is mainly used for result display and image processing for real-time monitoring.

See Figure 2. The realization of tree diameter measurement and statistics mainly consists of four parts: binocular camera calibration, deep learning and model training, tree image recognition and collection, forest tree diameter measurement and statistics.

The specific implementation steps are:

1. Binocular camera calibration;

The internal and external parameters of the binocular camera are calibrated through the dot-type calibration plate, and the calibration parameters are obtained based on the halcon nonlinear model handle calibration method. The image is corrected according to the calibration parameters to obtain the corrected image, and the BlockMatching algorithm is used to perform stereo matching and disparity map calculation on the corrected image.

In the complete camera imaging model, it is composed of a camera and a lens. In the present invention, a combination of an area array camera and a common lens is used. To complete the measurement of tree image diameter at breast height, it is necessary to determine the three-dimensional geometric position of forest tree diameter at breast height and its corresponding phase relationship in the captured image, mainly the world coordinate system and camera coordinate system, camera coordinate system and imaging plane coordinate system, imaging plane coordinate system and image The mutual conversion between the coordinate systems undergoes three transformations between the four coordinate systems, so as to realize the conversion of the three-dimensional space points from the world coordinate system to the image coordinate system. Due to the lens processing technology and other reasons, the collected images will have different degrees of distortion. In practical applications, the distortion of industrial lenses is mainly radial distortion, which will change the point (u, v) ^T of the imaging plane coordinate system:

为由于镜头畸变在成像平面上偏移的投影点坐标。where k is the lens radial distortion coefficient,

is the coordinate of the projected point offset on the imaging plane due to lens distortion.

Therefore, the radial distortion will cause the actual coordinates of the target point in the image to shift, and the accuracy of the spatial image positioning algorithm is closely related to the coordinate information. It is necessary to calibrate the camera to obtain the width and height of a single pixel of the internal camera. The focal length of the lens, the vertical projection of the optical center of the lens on the imaging plane, and the lens distortion coefficient are 6 parameters, and the image is corrected for distortion.

(1) Select a 160×160mm dot-shaped calibration plate for camera calibration. The diameter of the circle on it is 10mm, the distance between the centers of the circles is 20mm, and the black triangle in the upper left corner is its direction mark. Through the halcon calibration assistant window, according to the camera parameters, set the camera model to the area scanning type, and determine the focal length and the width and height of a single pixel;

(2) Constantly change the pose of the calibration board in the camera field of view and use the left and right cameras to collect 16 calibration images, calibrate the camera through the calibration board images, and obtain the internal parameters of the camera. And the method of back projection error is used to calculate the calibration accuracy of the camera, and the average error of the left and right camera calibration is obtained;

(3) Use the gen_radial_distortion_map operator to establish the camera internal reference mapping relationship with and without distortion, and use the Block Matching algorithm to perform stereo matching and disparity map calculations to correct image distortion and improve the accuracy of spatial positioning of target points.

2. Tree image recognition and collection;

The infrared sensor 1 signal line is connected to the corresponding pin of the development board, and an infrared judgment signal is given; the development board uses a USB data cable to connect to the industrial computer, and sends a logic judgment signal to the industrial computer to judge whether to save the image; if the above judgment is yes, then The binocular camera is connected to the industrial computer with a USB cable, and the collected images are transmitted to the industrial computer. When the infrared sensor 2 gives an infrared judgment signal, the image acquisition of a single tree is completed, and the program loops until the program is closed;

3. Deep learning and model training;

(1) Data set construction. Firstly, the collected image is preprocessed. Because the collected image is a binocular image, the left and right images are equally divided along the vertical central axis by using the gen_rectangle operator; after the segmentation, the smooth operator or the mean operator is used to smooth the image and remove Noise, to ensure that the image removes background interference before training; use invert_image, scale_image and other operators for image enhancement, mainly image flipping, contrast and brightness changes, etc., so as to increase the image data of the dataset and improve the robustness of the model ;Finally, image annotation is carried out, and MVTecDeep Learning Tool is used to mark the rectangular frame of the breast diameter part, and the construction of the deep learning image data set is completed;

(2) Model training and optimization. Based on the halcon deep learning network framework for tree target detection training, the SqueezeNet network model is introduced through the operator pretrained_dl_classifier_compact.hdl, and the three parameters of batch-size (batch processing), iterations (iterations) and learning_rate (learning rate) are continuously optimized. Model, determine the optimal training model by determining the average precision mean, precision and recall;

(3) Optimal model export. During training, specify the optimal model best_dl_model_detection.hdl save path to ensure the application of the best model during training.

4. Forest tree diameter measurement and statistics;

(1) Use the read_cam_par and read_pose operators to read the internal camera parameters of the left and right cameras obtained by binocular stereo vision calibration and the pose of the right camera relative to the left camera to ensure that the next step is to correct the binocular image in real time; use gen_binocular_rectification_map Operator, to obtain the transformation matrix between the non-standard outer limit image and the standard outer limit image, that is, the mapping image, and perform image correction on the left and right camera images according to the transformation matrix, so as to obtain the corrected left and right camera image interfaces; use read_dl_model to calculate Introduce the trained tree DBH detection model, and use the parameter setting set_dl_model_param operator to set the corresponding parameters to meet the working requirements of real-time DBH measurement;

(2) Real-time image interface reading. When the binocular camera recognizes and does not save the image, the program performs the tree diameter measurement process. The list_files and tuple_regexp_select operators are used to open the binocular camera to read image data in real time, and the image reading type can be set Various types such as bmp, gif, jpg;

(3) Set the corresponding recognition time, and when the forest image is recognized within the recognition time, the image detection link is performed. First, invert_image, scale_image and other operators are used to enhance the image to ensure the robustness of the detection process, and then the images of the left and right cameras are respectively corrected according to the transformation matrix; then the gray_features operator is used to calculate the gray feature value of the specified area , so as to obtain the gray value of the read image area.

(4) Use the gen_dl_samples_from_images and preprocess_dl_samples operators to detect the left and right images respectively, and obtain the gray value information according to the threshold(B, RegionB, L1, 255) and threshold(H, RegionH, L2, 255) operators, and then perform the trunk diameter Determination of the two boundary thresholds of some and the following areas: L1:=ValueR/1.3, L2:=ValueR/1.7, and finally divide the two channels, and then take the union to finally obtain the target detection area.

(5) Through the morphological processing method, use get_dict_tuple and gen_circle to obtain the boundary area below the tree diameter after threshold segmentation and the coordinate positions of the upper left and upper right target points of the boundary.

(6) Using the open_framegrabber operator, in the loop body, read in the forest images sequentially from the image data, select the first two frames of images at the beginning; keep an image later each time, read in a new image, overwrite An image earlier in time;

(7) Calculation of mean DBH. Use tuple_mean and disp_message operators, and define All:=[All,Mean] to count the number of recognition times. If the number of recognition times meets the specified setting requirements, calculate the mean value of the tree diameter at breast height, otherwise the recognition will be considered invalid, and the diameter at breast height will not be counted;

(8) Detect effectively identified trees, when the target trees disappear at the edge of the field of view and obtain an effective DBH value, the corresponding tree quantity statistics counter adds 1;

(9) Use the open_file and fwrite_string operators to export statistical values from the Excel table, mainly including the forest tree number counter value and the corresponding forest tree DBH average value.

Compared with prior art, the beneficial effect that the present invention has is:

1. Collect images of artificial forest trees through binocular cameras, and obtain the size of tree diameter at breast height through deep learning and binocular vision technology, which reduces manual workload, improves measurement accuracy, and avoids losses caused by measurement errors ;

2. Using the deep learning convolutional neural network model SqueezeNet, the size of the original AlexNet model is compressed to 1/50 of the original while ensuring no loss of accuracy, and its parameters are 50 times less than AlexNet, which can achieve target detection tasks more efficiently and quickly , get the area of the target rectangle;

3. Determine the two boundary thresholds of the diameter of the trunk and the following area through the acquisition of the gray value of the area. Finally, segment the two channels and then take the union, which can make the edge contour to be extracted more obvious and enhance the discrimination of the image. It is possible to reduce the influence of other background environmental factors on the measured object;

4. Using binocular stereo vision technology for distance measurement, there is no limit to the recognition rate compared with monocular vision, because in principle, it is not necessary to first identify and then measure, but directly measure the target recognition object. Binocular stereo vision is more accurate than monocular vision, and directly uses parallax to calculate distance;

5. When measuring the diameter at breast height of forest images, it can process a large amount of data and automatically obtain statistical results. For some non-forest targets such as shrubs and obstacles, the number of recognition times is insufficient, and actual statistics and diameter measurement are not performed to avoid misidentification and mismeasurement.

In this embodiment, a computer-readable storage medium is also provided, wherein the computer-readable storage medium stores one or more programs, and when the one or more programs are executed by a processor, the method is implemented.

In this embodiment, an electronic device is also provided, wherein the electronic device includes:

Processor; and,

A memory arranged to store computer-executable instructions which, when executed, cause the processor to perform the described method.

It should be noted:

The method used in this embodiment can be transformed into program steps and devices that can be stored in a computer storage medium, and implemented by being called and executed by a controller, wherein the devices should be understood as functional modules implemented by computer programs.

The algorithms and displays presented herein are not inherently related to any particular computer, virtual appliance, or other device. Various general purpose devices can also be used with the teachings based on this. The structure required to construct such an apparatus will be apparent from the foregoing description. Furthermore, the present invention is not specific to any particular programming language. It should be understood that various programming languages can be used to implement the content of the present invention described herein, and the above description of specific languages is for disclosing the best mode of the present invention.

In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure the understanding of this description.

Similarly, it should be appreciated that in the foregoing description of exemplary embodiments of the invention, in order to streamline this disclosure and to facilitate an understanding of one or more of the various inventive aspects, various features of the invention are sometimes grouped together in a single embodiment, figure, or its description. This method of disclosure, however, is not to be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the Detailed Description are hereby expressly incorporated into this Detailed Description, with each claim standing on its own as a separate embodiment of this invention.

Those skilled in the art can understand that the modules in the device in the embodiment can be adaptively changed and arranged in one or more devices different from the embodiment. Modules or units or components in the embodiments may be combined into one module or unit or component, and furthermore may be divided into a plurality of sub-modules or sub-units or sub-assemblies. All features disclosed in this specification (including accompanying claims, abstract and drawings) and any method or method so disclosed may be used in any combination, except that at least some of such features and/or processes or units are mutually exclusive. All processes or units of equipment are combined. Each feature disclosed in this specification (including accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.

Furthermore, those skilled in the art will understand that although some embodiments described herein include some features included in other embodiments but not others, combinations of features from different embodiments are meant to be within the scope of the invention. and form different embodiments. For example, in the following claims, any of the claimed embodiments may be used in any combination.

The various component embodiments of the present invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. The present invention can also be implemented as an apparatus or an apparatus program (for example, a computer program and a computer program product) for performing a 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 be in the form of one or more signals. Such a signal may be downloaded from an Internet site, or provided on a carrier signal, or provided in any other form.

It should be noted that the embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means can be embodied by one and the same item of hardware. The use of the words first, second, and third, etc. does not indicate any order. These words can be interpreted as names.

Claims (10)

1. A tree breast height diameter measuring method based on halcon binocular stereo vision is characterized by comprising the following steps:

s1, calibrating a binocular camera;

s2, acquiring images by using a binocular camera to construct a deep learning image data set, carrying out tree target detection training based on a halcon deep learning network frame, and obtaining an optimal model by adjusting parameters;

s3, reading camera internal parameters of the left camera and the right camera and the pose of the right camera relative to the left camera, which are obtained by binocular stereo vision calibration, by adopting a read _ cam _ par and a read _ position operator; (2) Acquiring a transformation matrix between the non-standard outer limit image and the standard outer limit image by adopting a gen _ binocular _ recommendation _ map operator, and carrying out image correction on left and right camera imaging according to the transformation matrix; (3) Introducing a trained tree breast diameter detection model by adopting a read _ dl _ model operator, and setting corresponding parameters by using a parameter setting set _ dl _ model _ param operator; (4) Opening the camera or reading an image by using a list _ files and a tuple _ regexp _ select operator; (5) Adopting gen _ dl _ samples _ from _ images and preprocess _ dl _ samples operators to respectively detect left and right images, obtaining a boundary region below the breast diameter of the forest after threshold segmentation, selecting coordinate positions of two target points on the upper left and the upper right of the boundary, defining the transverse coordinate difference value as the breast diameter, adopting a gray _ features operator to calculate the gray characteristic value of a specified region, thereby obtaining the gray value of the read image region, according to threshold (B, regionB, L1, 255), threshold (H, regionH, L2, 255) operators to obtain gray value information, then determining threshold values of the breast diameter part of the trunk and two boundaries of the following regions, wherein L1 is = ValueR/1.3, L2 is = ValueR/1.7, finally segmenting two channels, merging the two channels, and obtaining a target detection region; (6) Sequentially reading in forest images from image data in a cycle body by adopting an open _ frame classifier operator, selecting the first two frames of images at the beginning, reserving one image with later time each time, reading in a new image, and covering one image with earlier time; (7) Adopting a tuple _ Mean operator and a disp _ message operator, defining All: (= [ All, mean ]) to count the identification times, and if the identification times meet the specified setting requirements, calculating the Mean value of the breast diameter of the forest, otherwise, determining that the identification is invalid, and not counting the breast diameter value; (8) Detecting the effectively identified forest, and adding 1 to a corresponding forest number statistical counter when the target forest disappears at the edge of the visual field and an effective chest diameter value is obtained; (9) And (3) deriving statistical values by using open _ file and fwrite _ string operators, wherein the statistical values comprise the value of a forest number counter and the mean value of the breast diameters of the corresponding forests.

2. The tree breast diameter measuring method of claim 1, wherein the binocular camera calibration method of the step S1 further comprises:

calibrating internal and external parameters of the binocular camera through a dot type calibration plate, obtaining calibration parameters based on a halcon nonlinear model handle calibration method, correcting an image according to the calibration parameters, and performing stereo Matching and disparity map calculation by adopting a Block Matching algorithm.

3. The tree breast diameter measuring method of claim 2, wherein: in the step S1, the binocular camera calibration adopts a halcon calibration data model to acquire a plurality of calibration images, during the acquisition, after the acquisition of the previous calibration image is completed, the rotation of the acquired image direction is performed, then the acquisition of the next calibration image is performed until all calibration image sets are obtained, and the single calibration image is controlled to be kept in a static state during the acquisition.

4. The tree breast diameter measuring method of claim 2, wherein the calibration of the inside and outside parameters of the binocular camera is performed by a dot type calibration plate, calibration parameters are obtained based on a halcon nonlinear model handle calibration method, the image is corrected according to the calibration parameters, and stereo Matching and disparity map calculation are performed by a Block Matching algorithm, further comprising:

selecting a dot type calibration plate to calibrate the camera, calibrating an assistant window through a halcon, setting a camera model as a surface scanning type according to camera parameters, and determining a focal length and 3 parameters of the width and the height of a single pixel;

continuously changing the position and pose of a calibration plate in the visual field of a camera, acquiring a plurality of calibration images by using a left camera and a right camera respectively, calibrating the cameras by using the calibration plate images to acquire internal parameters of the cameras, and calculating the calibration precision of the cameras by adopting a back projection error method to obtain the calibration average error of the left camera and the right camera;

and establishing a coherent mapping relation with and without distortion through a gen _ radial _ distortion _ map operator, and performing stereo Matching and disparity map calculation through a Block Matching algorithm.

5. The tree breast diameter measuring method according to claim 1, wherein said step S2 further comprises, before constructing the deep learning image data set:

adopting a gen _ rectangle operator to carry out average segmentation on the left image and the right image along the central axis in the vertical direction;

smoothing the image by adopting smooth operators or mean operators after segmentation;

and (5) carrying out image enhancement by using an invert _ image operator and a scale _ image operator, and carrying out image annotation.

6. The tree breast height diameter measuring method of claim 5, wherein:

performing the image enhancement by adjusting contrast and brightness; and/or

The image labeling further comprises labeling the chest diameter and the rectangular frame below the chest diameter by using an MVTec Deep Learning Tool.

7. The method for measuring the breast height diameter of the tree according to claim 1, wherein the step S2 of performing tree target detection training based on a halcon deep learning network framework to obtain an optimal model by adjusting parameters further comprises:

the model network is configured as a SqueezeNet, with its corresponding operator configured as predicted _ dl _ classifier _ compact.hdl; and/or

The adjusted parameters are configured to include at least a batch-size, iterations, and spare _ rate.

8. The tree breast-height diameter measuring method of claim 1, wherein in step S3, the step of obtaining a boundary region below the breast-height diameter of the tree after threshold segmentation, selecting coordinate positions of two target points at the upper left and upper right of the boundary, and defining a difference value of horizontal coordinates of the target points as the breast-height diameter further comprises:

and acquiring a boundary area below the breast diameter of the forest and coordinate positions of two target points, namely the upper left point and the upper right point, of the boundary after threshold segmentation by adopting a morphological processing method through get _ direct _ tuple and gen _ circle.

9. The tree breast diameter measuring method of claim 1, wherein before calculating the gray feature value of the designated area by using the gray _ features operator in step S3, the method further comprises:

carrying out image enhancement by adopting operators such as invert _ image, scale _ image and the like;

and respectively carrying out image correction on the left camera and the right camera according to the transformation matrix.

10. The tree breast diameter measuring method according to claim 1, wherein the step S3 further comprises:

in the steps (1), (2) and (3), the internal parameters of the camera, the transformation matrix and the forest breast-height diameter detection model are read by placing a computer at a fixed position. And/or

In the step (4), the read image is configured into a color image; and/or

In the step (5), a tree breast diameter rectangular frame is obtained according to the labeling mode, and a detection area is obtained according to the gray value information and the threshold value setting of two boundaries.

Images