CN113362468A - Dimension measuring method for hub of train wheel - Google Patents

Abstract

The invention relates to a dimension measurement method of a train wheel hub. The method includes: acquiring a three-dimensional point cloud of a train wheel hub, and performing down-sampling processing on the three-dimensional point cloud of the train wheel hub to determine the down-sampled three-dimensional point cloud; filtering out outliers in the down-sampled three-dimensional point cloud, and determining to filter out The three-dimensional point cloud of the train hub after filtering; the outliers include noise point clouds and artifacts; rotate the filtered three-dimensional point cloud of the train hub at different angles to perform multiple measurements to determine the three-dimensional points of the train hub at different positions. The coordinates of the cloud; the multi-scale fusion Gaussian weight distribution algorithm is used to process the coordinates of the three-dimensional point cloud of the train hub at different positions to determine the size of the position to be measured. The invention can accurately and reliably measure the dimensions of the train wheel hub at different positions and angles.

Description

A kind of dimension measurement method of train wheel hub

technical field

The invention relates to the field of three-dimensional point cloud processing technology and computer science technology, in particular to a dimension measurement method of a train wheel hub.

Background technique

The measurement of products and parts in industry is often performed manually, which is time-consuming, inefficient and labor-intensive. Existing common measuring instruments that can be manually operated can often only measure the size of smaller objects, but cannot measure objects with larger dimensions. In addition, during manual measurement, the measurement accuracy sometimes depends on the operator's proficiency. For example, when using an inner diameter dial indicator to measure the inner diameter, if the inner diameter dial indicator is not placed accurately, the accurate value cannot be measured. In addition to the above shortcomings, the efficiency of manual measurement is low, and the measurement scale values obtained from some measuring tools (such as vernier calipers, radius micrometers) require further calculation to obtain accurate measurement values. For large-sized objects such as train wheels, the manual method is more time-consuming, inefficient and labor-intensive.

In order to solve the above problems, a series of automatic dimension measuring instruments have been developed. However, when these instruments measure the size, the object still needs to be placed on the platform, which results in that the size range of the measurement is still not large, and manual operation is still required, resulting in low measurement efficiency. In addition to the above-mentioned disadvantages, these measuring instruments are expensive and cost-effective. And due to the particularity of the train wheel hub, the measurement efficiency of manual-assisted operation is lower, and the dimensions of the train wheel hub at different positions and angles cannot be accurately and reliably measured.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a method for measuring the dimensions of a train wheel hub, so as to solve the problems of low measurement efficiency and inability to accurately and reliably measure the dimensions of the train wheel hub at different positions and angles.

For achieving the above object, the present invention provides the following scheme:

A method for measuring the dimensions of a train wheel hub, comprising:

Obtaining a three-dimensional point cloud of the train hub, and performing down-sampling processing on the three-dimensional point cloud of the train hub to determine the down-sampling three-dimensional point cloud;

Filter out outliers in the down-sampled three-dimensional point cloud, and determine the filtered train hub three-dimensional point cloud; the outliers include noise point clouds and artifacts;

Rotating the filtered three-dimensional point cloud of the train hub at different angles to perform multiple measurements to determine the coordinates of the three-dimensional point cloud of the train hub at different positions;

Using a multi-scale fusion Gaussian weight distribution algorithm, the coordinates of the three-dimensional point cloud of the train hub at the different positions are processed to determine the size of the position to be measured.

Optionally, the obtaining of the three-dimensional point cloud of the train hub, and performing downsampling processing on the three-dimensional point cloud of the train hub to determine the down-sampling three-dimensional point cloud, specifically including:

The space of the three-dimensional point cloud of the train hub is divided into a plurality of cubic spaces; each of the cubic spaces includes a plurality of three-dimensional points of the train hub;

Calculate the average value of all the three-dimensional points of the train hub in the cubic space, and use the average value of the three-dimensional points of the train hub as a down-sampled three-dimensional point cloud.

Optionally, the filtering out outliers in the down-sampled three-dimensional point cloud, and determining the three-dimensional point cloud of the train hub after filtering, specifically includes:

Acquiring multiple domain points of each down-sampled three-dimensional point in the down-sampled three-dimensional point cloud, and calculating an average distance value between the multiple domain points and the down-sampled three-dimensional point;

Determine the mean value and the standard deviation according to the average distance value of all the down-sampled three-dimensional points;

The outliers in the down-sampled 3D point cloud are filtered out according to the mean value and the standard deviation, and the filtered train hub 3D point cloud is determined.

Optionally, rotating the filtered three-dimensional point cloud of the train hub at different angles to perform multiple measurements to determine the coordinates of the three-dimensional point cloud of the train hub at different positions, specifically including:

Adjusting the three-dimensional coordinate system in which the filtered train hub three-dimensional point cloud is located so that the X axis coincides with the axis of the train hub;

Taking 1° as the rotation angle, the filtered three-dimensional point cloud of the train hub is rotated and measured several times around its own X-axis using the two-dimensional rotation formula, and the coordinates of the three-dimensional point cloud of the train hub at different positions are determined.

Optionally, the multi-scale fusion Gaussian weight distribution algorithm is used to process the coordinates of the three-dimensional point cloud of the train hub at the different positions to determine the size of the position to be measured, specifically including:

Divide the position to be measured into 5 large areas;

Rotating the three-dimensional point cloud of the train hub to different large regions, and dividing the three-dimensional point cloud of the train wheel in the large region into 5 small regions;

According to calculating the coordinate average value of the coordinates of the three-dimensional point cloud of the train hub in each small area, and storing the coordinate average value of 5 small areas in a one-dimensional list with a length of 5;

According to the fixed Gaussian weight, the average value of the coordinates of the five small areas is weighted to determine the size of the different positions of the large area;

The size at the position to be measured is determined according to the size at the different positions of the large area.

According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: the present invention provides a method for measuring the size of a train wheel hub, which directly performs high-efficiency multi-angle and multi-position accurate measurement on the three-dimensional point cloud data of the train wheel hub. And robust measurement does not require heavy and low-efficiency manual operation, the invention is applied to the size measurement and wear determination of the train wheel hub, and can accurately and reliably measure the exact size of the train wheel hub at different positions and angles to check the wear condition. The determination can greatly improve the efficiency and accuracy of the dimension measurement and wear determination of the train wheel, greatly reduce the labor cost and greatly improve the accuracy, and has a strong practical value in the industry.

Description of drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings required in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some of the present invention. In the embodiments, for those of ordinary skill in the art, other drawings can also be obtained according to these drawings without any creative effort.

Fig. 1 is the flow chart of the dimension measuring method of the train wheel hub provided by the present invention;

Figure 2 is a schematic diagram of the multiplication calculation of the positioning value and the one-dimensional Gaussian weight kernel of two scales;

3 is a schematic diagram of a measurement position and its adjacent small area;

Figure 4 is a schematic diagram of size value calculation;

5 is a schematic diagram of adjusting the point cloud so that its axis coincides with the x-axis;

FIG. 6 is a schematic diagram of the dimension measurement method of a train wheel hub provided by the present invention in an actual operation process.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

The purpose of the present invention is to provide a method for measuring the dimensions of a train wheel hub, which can accurately and reliably measure the dimensions of the train wheel hub at different positions and angles.

In order to make the above objects, features and advantages of the present invention more clearly understood, the present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

Fig. 1 is the flow chart of the size measurement method of the train wheel hub provided by the present invention, as shown in Fig. 1, a kind of size measurement method of the train wheel hub, including:

Step 101: Acquire a three-dimensional point cloud of the train hub, and perform down-sampling processing on the three-dimensional point cloud of the train hub to determine the down-sampled three-dimensional point cloud.

The step 101 specifically includes: dividing the space of the three-dimensional point cloud of the train hub into a plurality of cubic spaces; each of the cubic spaces includes a plurality of three-dimensional points of the train hub; calculating all the train hubs in the cubic space The average value of the three-dimensional points, and the average value of the three-dimensional points of the train hub is used as the down-sampled three-dimensional point cloud.

In practical applications, the space of the 3D point cloud of the train hub obtained by the matrix camera is first divided into small cubic spaces (voxels) with a fixed size of v*v*v. At this time, the 3D point of the train hub is divided into each voxel. , calculate the average of the coordinates of all the three-dimensional points of the train hub in each voxel, and use this average as the coordinates of the voxel, that is, downsample the three-dimensional point cloud to achieve the purpose of downsampling.

Step 102: Filter out outliers in the down-sampled 3D point cloud, and determine a filtered train hub 3D point cloud; the outliers include noise point clouds and artifacts.

The step 102 specifically includes: acquiring multiple domain points of each down-sampled three-dimensional point in the down-sampled three-dimensional point cloud, and calculating the average distance value between the multiple domain points and the down-sampled three-dimensional point; The average distance value of the down-sampled three-dimensional points determines the mean value and the standard deviation; according to the mean value and the standard deviation, the outliers in the down-sampled three-dimensional point cloud are filtered out, and the filtered train wheel three-dimensional point cloud is determined. .

In practical applications, the downsampled 3D point cloud still has many "noisy" point clouds far away from the train hub point cloud, as well as some artifacts. Using the statistical properties of mean and variance, the "noise" and artifacts far away from the main point cloud are removed, so that only the main point cloud (train hub point cloud) is left in the entire three-dimensional space to filter out outliers. The formula is as follows:

Assuming that there are n three-dimensional points in the down-sampled three-dimensional point cloud, _i is any three-dimensional point, and it is specified that each point Pi in the down-sampled three-dimensional point cloud has N neighbor points, and the distance between the N points and _Pi is calculated. The average distance value S _i of :

Among them, _Wi is the neighborhood of point _Pi , (x, y, z) is the point coordinate of the neighborhood point, and D(x, y, z) is the distance between each neighborhood point and _Pi .

Each point P _i has its own S _i , (the average distance of N neighboring points of this point from this point ₎ , and these average distances Si obey a Gaussian distribution. At this point the Gaussian distribution has a mean and a standard deviation. The mean and standard deviation are both calculated from the global mean:

Find the mean:

Find the standard deviation σ:

Set the threshold T to determine whether the point Q is filtered out:

Among them: S _Q is the average distance between the neighboring points of point Q and point Q. If μ-σT<S _Q <μ+σT is satisfied, then keep this point, otherwise, it is regarded as an outlier and filtered out.

Step 103: Rotate the filtered three-dimensional point cloud of the train wheel at different angles to perform multiple measurements, and determine the coordinates of the three-dimensional point cloud of the train wheel at different positions.

The step 103 specifically includes: adjusting the three-dimensional coordinate system in which the filtered train hub three-dimensional point cloud is located so that the X axis coincides with the axis of the train hub; taking 1° as the rotation angle, using a two-dimensional rotation formula to The filtered three-dimensional point cloud of the train wheel is rotated around its own X axis for many times and measured, and the coordinates of the three-dimensional point cloud of the train wheel at different positions are determined.

In order to accurately measure the train wheel hub and improve the robustness of the measurement results, the point cloud rotation algorithm is used to adjust and rotate the point cloud to perform multiple measurements (measure results every 1°). First adjust the point cloud to the position to be rotated (measure the size of the train hub in the axial direction, first adjust the three-dimensional coordinate system so that its x-axis coincides with the axis of the train hub) (Figure 5), and then use the two-dimensional rotation formula to convert The point cloud is rotated around its own x-axis (Figure 5) with the following formula:

Among them, x is unchanged, (x, y', z') is the new coordinates of each point in the point cloud after rotation, that is, the coordinates of the three-dimensional point cloud of the train hub at different positions.

Step 104: Using a multi-scale fusion Gaussian weight distribution algorithm, the coordinates of the three-dimensional point cloud of the train hub at different positions are processed to determine the size of the position to be measured.

The step 104 specifically includes: dividing the position to be measured into five large areas; rotating the three-dimensional point cloud of the train hub to different large areas, and dividing the three-dimensional point cloud of the train hub in the large area into 5 small areas; according to calculating the coordinate average value of the coordinates of the three-dimensional point cloud of the train hub in each small area, and storing the coordinate average value of the 5 small areas in a one-dimensional list with a length of 5; according to The fixed Gaussian weight performs weighted calculation on the average value of the coordinates of the five small areas to determine the size of the different positions of the large area; the size of the position to be measured is determined according to the size of the different positions of the large area.

存入一个长度为5的一维列表中，按照固定高斯权重对这五个平均值进行加权计算(中间权重高，两侧权重低)，计算结果即为此小区域的精确定位值，如图2所示的尺度一。In practical applications, the point cloud is rotated according to step 103, and the position to be measured on the point cloud is transferred to the measurement position. At this time, the point cloud obtains new coordinates (x, y', z'). For example, to measure the precise positioning value of the small middle area in Figure 2, it is necessary to turn the small area of the point cloud to the measurement position, as shown in Figure 3. At this time, the three-dimensional point obtains new coordinates (x, y', z'), divides the three-dimensional point in the middle small area into 5 small areas, and obtains the average value of each small area

Stored in a one-dimensional list with a length of 5, the five averages are weighted according to the fixed Gaussian weight (the middle weight is high, and the weight on both sides is low), and the calculation result is the precise positioning value of this small area, as shown in the figure 2 shows scale one.

This method is similar to the mean value method. The mean value method uses the average coordinates of all three-dimensional points in this small area as the positioning value of this small area. The present invention uses the idea of weighting to assign different weights to different parts of the points, and finally obtain this Small area targeting value.

The formula is as follows:

Assuming that there are m points in the small area, divide it into a one-dimensional list of length 5:

I={m ₁ , m ₂ , m ₃ , m ₄ , m ₅ } (6)

Among them: m ₁ , m ₂ , m ₃ , m ₄ , and m ₅ are points divided into these five intervals, respectively.

Average the points in the above 5 intervals to obtain the list I ₂ :

I ₂ ={A ₁ , A ₂ , A ₃ , A ₄ , A ₅ } (7)

Among them: A ₁ , A ₂ , A ₃ , A ₄ , A ₅ are the average coordinates of points in 5 intervals respectively

Define the Gaussian kernel G:

G={g ₁ , g ₂ , g ₃ , g ₄ , g ₅ } (8)

Among them: g ₁ , g ₂ , g ₃ , g ₄ , g ₅ are approximate to Gaussian distribution, the value of g ₃ is the largest, and gradually decreases on both sides, and g ₁ and g ₂ are the smallest.

Calculate the precise positioning value L of the small area, as shown in the scale 1 in Figure 2:

表示列表对应元素相乘并相加，即：in:

Indicates that the corresponding elements of the list are multiplied and added, that is:

There are overlapping parts between the middle small area and the 4 adjacent small areas on the left and right. Rotate the point cloud and move the adjacent small area to the measurement position as shown in Figure 3. According to the above method, measure the 4 adjacent small areas respectively. Precise positioning value for a small area. Then put the precise positioning values of these 5 small areas into a one-dimensional list with a length of 5, as shown in the scale 2 in Figure 2, and perform the weighted calculation according to the fixed Gaussian weight. The calculation result is the small area 1 in Figure 3. Accurate and robust positioning value, as shown in Figure 3, in practical applications, the Gaussian weight of itself is high, and the Gaussian weight of adjacent small areas on both sides is low. The formula is as follows:

The precise positioning values of the above 5 small areas are put into the list I ₃ :

I ₃ ={L ₁ , L ₂ , L ₃ , L ₄ , L ₅ } (11)

Define the Gaussian kernel G:

G={g ₁ , g ₂ , g ₃ , g ₄ , g ₅ } (12)

Among them, g ₁ , g ₂ , g ₃ , g ₄ , and g ₅ are approximate to Gaussian distribution, the value of g ₃ is the largest, and gradually decreases on both sides, and g ₁ and g ₂ are the smallest.

Calculate the accurate and robust localization value L' in a small area, that is, the final localization value, which is the scale 2 in Figure 2:

表示列表对应元素相乘并相加，即：in,

Indicates that the corresponding elements of the list are multiplied and added, that is:

The exact and robust positioning value L' at the specified position of the point cloud is not only related to the coordinates of the three-dimensional point in the small area at this position, that is, the scale 1 in Figure 2, to obtain the precise positioning value L, which is also related to the adjacent position. The positioning value of the small area at the location is related, that is, the scale 2 in Figure 2, and the final accurate and robust positioning value L' is obtained based on L.

The calculation method of the accurate and robust positioning value of the small area 2 in FIG. 3 is the same as above; after obtaining the accurate and robust positioning values of the two small areas, the accurate and robust positioning values can be obtained by calculating (subtracting) the positioning values of the two small areas. Robust dimension value. For example, as shown in Figure 4, when the positioning values of small area 1 and small area 2 are obtained, accurate and robust size values in the x-direction and z-direction can be obtained by subtracting the operation. The formula is as follows:

Wherein, L′ _x1 and L′ _y1 are the precise and robust positioning values of the small area 1 , and L′ _x2 and L′ _y2 are the precise and robust positioning values of the small area 2 . L' _x and L' _y are the precise and robust positioning difference values in the x-direction and the y-direction, that is, the size value, respectively. After obtaining the accurate and robust positioning difference (dimension value), the wear judgment can be carried out according to the standard dimension value.

FIG. 6 is a schematic diagram of the method for measuring the size of a train wheel hub provided by the present invention in an actual operation process. As shown in FIG. 6 , a matrix camera is used to obtain point cloud data (x, y, z) of the train wheel hub; using voxel downsampling The algorithm converts point coordinates in space into a downsampling model of voxel coordinates; constructs an outlier filtering model that uses a statistical outlier removal algorithm to remove "noise" points in space; uses the proposed point cloud rotation algorithm Multi-angle rotation is performed on the point cloud to measure accurate and robust size values at different positions; for the above processed point cloud, the proposed multi-scale Gaussian weight fusion algorithm is used to calculate the accurate and robust point cloud at the specified position. The positioning value of , and the two positioning values are subtracted to obtain an accurate and robust size value.

The measurement of the train hub no longer requires heavy and inefficient manual operations, but can directly perform high-efficiency, multi-angle, multi-position accurate and robust measurements on the 3D point cloud data of the train hub. The method is applied to the size measurement and wear determination of the train wheel hub, and can accurately and reliably measure the size of the train wheel hub at different positions and angles to determine the wear condition, which greatly improves the efficiency of the size measurement and wear determination of the train wheel. Accuracy, while greatly reducing labor costs, greatly improves accuracy, and has strong practical value in industry.

The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same and similar parts between the various embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant part can be referred to the description of the method.

In this paper, specific examples are used to illustrate the principles and implementations of the present invention. The descriptions of the above embodiments are only used to help understand the methods and core ideas of the present invention; meanwhile, for those skilled in the art, according to the present invention There will be changes in the specific implementation and application scope. In conclusion, the contents of this specification should not be construed as limiting the present invention.

Claims (5)

1. A method of measuring dimensions of a hub of a railway wheel, comprising:

acquiring a three-dimensional point cloud of a train hub, and performing down-sampling processing on the three-dimensional point cloud of the train hub to determine the down-sampling three-dimensional point cloud;

filtering outliers in the down-sampling three-dimensional point cloud, and determining the filtered three-dimensional point cloud of the hub of the train; the outliers comprise noise point clouds and artifacts;

rotating the filtered three-dimensional point cloud of the train wheel hub at different angles to measure for multiple times, and determining coordinates of the three-dimensional point cloud of the train wheel hub at different positions;

and processing the coordinates of the three-dimensional point cloud of the hub of the train wheel at different positions by utilizing a multi-scale fusion Gaussian weight distribution algorithm to determine the size of the position to be measured.

2. The method for measuring the size of the hub of the train as claimed in claim 1, wherein the step of obtaining the three-dimensional point cloud of the hub of the train and performing down-sampling processing on the three-dimensional point cloud of the hub of the train to determine the down-sampled three-dimensional point cloud specifically comprises the following steps:

dividing the space of the three-dimensional point cloud of the hub of the train into a plurality of cubic spaces; each cubic space comprises a plurality of three-dimensional points of the train hubs;

and calculating the average value of all the three-dimensional points of the hub of the train in the cubic space, and taking the average value of the three-dimensional points of the hub of the train as down-sampling three-dimensional point cloud.

3. The method for measuring the dimensions of a train hub according to claim 1, wherein the filtering out outliers in the down-sampled three-dimensional point cloud and determining the filtered out train hub three-dimensional point cloud specifically comprises:

acquiring a plurality of field points of each down-sampling three-dimensional point in the down-sampling three-dimensional point cloud, and calculating the average distance value between the plurality of field points and the down-sampling three-dimensional point;

determining a mean value and a standard deviation according to the average distance values of all the down-sampling three-dimensional points;

and filtering outliers in the down-sampling three-dimensional point cloud according to the mean value and the standard deviation, and determining the filtered train wheel hub three-dimensional point cloud.

4. The method for measuring the size of the hub of the train as claimed in claim 1, wherein the rotating the filtered three-dimensional point cloud of the hub of the train at different angles for a plurality of times to determine the coordinates of the three-dimensional point cloud of the hub of the train at different positions specifically comprises:

adjusting a three-dimensional coordinate system where the filtered three-dimensional point cloud of the hub of the train is located until an X axis is coincident with the axis of the hub of the train;

and taking 1 degree as a rotation angle, and performing multiple rotations and measurement on the filtered three-dimensional point cloud of the hub of the train around the X axis of the three-dimensional point cloud of the hub of the train by using a two-dimensional rotation formula to determine coordinates of the three-dimensional point cloud of the hub of the train at different positions.

5. The method for measuring the dimensions of the hub of the train as claimed in claim 1, wherein the determining the dimensions of the positions to be measured by processing the coordinates of the three-dimensional point cloud of the hub of the train at the different positions by using a multi-scale fusion gaussian weight distribution algorithm specifically comprises:

dividing the position to be measured into 5 large areas;

rotating the train wheel hub three-dimensional point cloud to different large areas, and dividing the train wheel hub three-dimensional point cloud of the large area into 5 small areas;

calculating the coordinate average value of the coordinates of the three-dimensional point cloud of the hub of the train in each small area, and storing the coordinate average value of 5 small areas into a one-dimensional list with the length of 5;

carrying out weighted calculation on the coordinate average value of the 5 small areas according to the fixed Gaussian weight, and determining the sizes of different large area positions;

and determining the size of the position to be measured according to the sizes of the different large-area positions.

Images