A method and system for constructing a point cloud model of a green new energy lightweight vehicle

Abstract

The present invention relates to the field of image processing technology, and proposes a method and system for constructing a point cloud model of a green new energy lightweight automobile, including: collecting point cloud data of a new energy lightweight automobile; marking target point cloud data, target points, comparison point cloud data and comparison points, determining the derivative sequence and color change trend of the target point, and determining the texture change degree of the target point; determining the matching degree of the target point and the comparison point, and dividing the matching point pairs in the target point cloud data and the comparison point cloud data according to the matching degree; combining the Euclidean distance between all matching point pairs in the target point cloud data and the comparison point cloud data, representing the target point cloud data and the comparison point cloud data in the same coordinate system, and completing the construction of the point cloud model of the green new energy lightweight automobile according to the result of representing the point cloud data of all new energy lightweight automobiles in the same coordinate system. The present invention can balance the efficiency and accuracy of point cloud matching and improve the accuracy of the point cloud model of the lightweight automobile.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a method and system for constructing a green new energy lightweight automobile point cloud model. Background Art

[0002] Lightweight technology is one of the key means to achieve lightweight design and manufacturing of automobiles. In the design and manufacturing process of lightweight automobiles, reverse engineering is widely used to optimize automobile parts. The core of reverse engineering lies in the construction of point cloud models of lightweight automobiles. When constructing a point cloud model of a lightweight automobile, due to the large size of the automobile, it is often necessary to obtain point cloud data from different angles, and then align the point cloud data according to the overlapping parts of the point cloud data from different angles, so as to transform the point clouds from different angles to the same coordinates and obtain the complete point cloud data of the lightweight automobile. Therefore, aligning different point cloud data is the key to ensuring the accuracy of point cloud data.

[0003] Generally, the ICP point cloud registration algorithm is used to register point clouds. However, the ICP point cloud registration algorithm often uses a fixed convergence threshold and cannot achieve a balance between efficiency and accuracy based on the point cloud data itself. Specifically, in areas with less detailed information in the point cloud data, a convergence threshold that is too small will result in too high computing resources being required, affecting the efficiency of the algorithm. In areas with more detailed information in the point cloud data, a convergence threshold that is too large will result in the details being ignored and insufficient calculation accuracy. Therefore, it is necessary to balance efficiency and accuracy based on the specific needs of the point cloud data to improve the efficiency and accuracy of point cloud registration. Summary of the invention

[0004] The present invention provides a method and system for constructing a green new energy lightweight automobile point cloud model to solve the problem that the value of the convergence threshold cannot balance the efficiency and accuracy of point cloud matching, making the construction of the lightweight automobile point cloud model inaccurate. The technical solution adopted is as follows:

[0005] In a first aspect, an embodiment of the present invention provides a method for constructing a point cloud model of a green new energy lightweight automobile, the method comprising the following steps:

[0006] Collect point cloud data of new energy lightweight vehicles with overlapping collection areas from different angles. Each point in the point cloud data corresponds to a color value and a three-dimensional coordinate.

[0007] Record any point cloud data as target point cloud data, record any point in the target point cloud data as the target point, determine the neighboring points of the target point according to the spatial position distribution relationship between the target point and other points in the target point cloud data, determine the derivative sequence and color change trend of the target point according to the distribution of color values ​​and spatial position distribution of the target point and all the neighboring points of the target point, and determine the texture change degree of the target point according to the color values, derivative sequences and color change trends of all the neighboring points of the target point;

[0008] According to the relative position relationship and the difference between the color values ​​of the target point and all the neighboring points, the color distribution sequence of the target point is determined, any point cloud data different from the point cloud data where the target point is located is recorded as the comparison point cloud data, and any point in the comparison point cloud data is recorded as the comparison point. According to the data distribution of the texture variation of all points contained in the two point cloud data where the target point and the comparison point are located, the correlation of the color distribution sequence of the target point and the comparison point, and the texture variation of the target point and the comparison point, the matching degree of the target point and the comparison point is determined, and the matching point pairs in the target point cloud data and the comparison point cloud data are divided according to the matching degree;

[0009] According to the Euclidean distance and matching degree between all matching point pairs in the target point cloud data and the comparison point cloud data, the target point cloud data and the comparison point cloud data are represented in the same coordinate system. According to the result that the point cloud data of all new energy lightweight vehicles are represented in the same coordinate system, the construction of the green new energy lightweight vehicle point cloud model is completed.

[0010] Furthermore, the specific method for determining the neighboring points of the target point is:

[0011] The first preset number of points in the target point cloud data that are closest to the target point in space are recorded as neighboring points of the target point.

[0012] Furthermore, the method for determining the derivative sequence and color change trend of the target point is:

[0013] Perform plane fitting on the target point and all its neighboring points to obtain the normal vector of the fitting plane, arrange the color values ​​of the target point and all its neighboring points in the order of the points in the direction of the normal vector, and obtain the color value sequence of the target point; perform polynomial fitting on the color value sequence of the target point to obtain the fitting curve of the target point, calculate the derivative corresponding to each color value in the color value sequence of the target point on the fitting curve of the target point, arrange the derivative corresponding to each color value in the color value sequence of the target point in the order of the color value in the color value sequence, and obtain the derivative sequence of the target point;

[0014] The second preset value is used as the window length, and the moving variance method is used to process the derivative sequence of the target point to obtain the moving variance sequence of the target point. A straight line fitting is performed on the moving variance sequence to obtain the moving variance straight line equation of the target point, and the slope of the moving variance straight line equation of the target point is recorded as the color change trend of the target point.

[0015] Further, the texture change degree of the target point is determined according to the color values, derivative sequences and color change trends of all neighboring points of the target point, including the specific method of:

[0016] The ratio of the number of 0s contained in the derivative sequence of the target point to the number of all neighboring points of the target point is recorded as the first ratio of the target point;

[0017] The absolute value of the color change trend of the target point is recorded as the first absolute value of the target point, the sum of the first absolute value of the target point and the number 1 is recorded as the first sum value of the target point, the first sum value of the target point is used as the true number of the logarithmic function with base 10, and the value of the logarithmic function is recorded as the first function value of the target point;

[0018] The product of the first function value of the target point, the first ratio and the standard deviation of the color values ​​of all neighboring points of the target point is recorded as the texture variation degree of the target point.

[0019] Furthermore, the method for determining the color distribution sequence of the target point is:

[0020] Calculate the Chebyshev distance between the target point and each neighboring point;

[0021] The absolute value of the difference between the color values ​​of the target point and the neighboring points is recorded as the color difference of the neighboring points;

[0022] The color differences of all neighboring points of the target point are arranged in ascending order according to the Chebyshev distance between the neighboring point and the target point to obtain the color distribution sequence of the target point.

[0023] Further, the matching degree between the target point and the comparison point is determined according to the data distribution of the texture variation degree of all points contained in the two point cloud data where the target point and the comparison point are located, the correlation of the color distribution sequence of the target point and the comparison point, and the texture variation degree of the target point and the comparison point, including the specific method of:

[0024] Clustering the texture variation of all points contained in the two point cloud data where the target point and the comparison point are located to obtain cluster clusters;

[0025] When the texture variation of the target point and the comparison point is in the same cluster, the direct matching value between the target point and the comparison point is recorded as 1; when the texture variation of the target point and the comparison point is not in the same cluster, the direct matching value between the target point and the comparison point is recorded as 0;

[0026] The matching degree between the target point and the comparison point is determined according to the correlation of the color distribution sequences of the target point and the comparison point, the direct matching value, and the texture variation between the target point and the comparison point.

[0027] Further, the matching degree between the target point and the comparison point is determined according to the correlation of the color distribution sequences of the target point and the comparison point, the direct matching value, and the texture variation degree between the target point and the comparison point, including the specific method of:

[0028] The absolute value of the difference between the texture variation of the target point and the comparison point is recorded as the texture variation difference between the target point and the comparison point, and the ratio of the correlation between the color distribution sequence of the target point and the comparison point to the texture variation difference is recorded as the second ratio between the target point and the comparison point;

[0029] The product of the second ratio between the target point and the comparison point and the direct matching value is recorded as the matching degree between the target point and the comparison point.

[0030] Furthermore, the method of dividing the matching point pairs in the target point cloud data and the comparison point cloud data according to the matching degree includes the following specific methods:

[0031] The maximum value of the matching degree between the target point and all the points in the comparison point cloud data is taken as the matching point of the target point in the comparison point cloud data;

[0032] Obtain the matching point of any point in the target point cloud data in the comparison point cloud data;

[0033] The matching degrees of all points in the target point cloud data and the matching points in the comparison point cloud data are divided using the maximum inter-class variance method to obtain a division threshold, and the points in the target point cloud data with a matching degree greater than the division threshold and the matching points in the comparison point cloud data are taken as a set of matching point pairs.

[0034] Further, the target point cloud data and the comparison point cloud data are represented in the same coordinate system according to the Euclidean distance and matching degree between all matching point pairs in the target point cloud data and the comparison point cloud data, and the construction of the green new energy lightweight vehicle point cloud model is completed according to the result of representing the point cloud data of all new energy lightweight vehicles in the same coordinate system, including the specific method of:

[0035] The mean value of the Euclidean distances between all matching point pairs in the target point cloud data and the comparison point cloud data is recorded as the matching distance between the target point cloud data and the comparison point cloud data, and the mean value of the matching degrees between all matching point pairs in the target point cloud data and the comparison point cloud data is recorded as the average matching degree between the target point cloud data and the comparison point cloud data;

[0036] According to the matching distance and average matching degree between the target point cloud data and the comparison point cloud data, the optimal convergence threshold between the target point cloud data and the comparison point cloud data is determined. The specific calculation formula is:

[0037]

[0038] in, Indicates the optimal convergence threshold of the target point cloud data and the comparison point cloud data; Indicates the initial value of the convergence threshold; represents a natural constant; Indicates the average matching degree between the target point cloud data and the comparison point cloud data; Indicates the matching distance between the target point cloud data and the comparison point cloud data; represents the inverse tangent function; represents pi;

[0039] The optimal convergence threshold of the target point cloud data and the comparison point cloud data is taken as the value of the convergence threshold, and the target point cloud data and the comparison point cloud data are registered using the ICP point cloud registration algorithm, and the target point cloud data and the comparison point cloud data are represented in the same coordinate system;

[0040] Obtain the result of expressing the collected point cloud data of any two different new energy lightweight vehicles in the same coordinate system; obtain the result of expressing the collected point cloud data of all new energy lightweight vehicles in the same coordinate system;

[0041] The Poisson surface reconstruction algorithm is used to obtain the point cloud model of the green new energy lightweight vehicle based on the result that the point cloud data of all new energy lightweight vehicles are represented in the same coordinate system.

[0042] In the second aspect, an embodiment of the present invention also provides a system for constructing a green new energy lightweight automobile point cloud model, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of any one of the above-mentioned methods when executing the computer program.

[0043] The beneficial effects of the present invention are:

[0044] Based on the characteristics that the texture changes of each point in the point cloud data corresponding to the car body area are relatively small, the texture changes of each point corresponding to the wheel are relatively large, and the texture of each point at the junction between different parts and modules of the car will show sudden changes, and based on the characteristics that the texture features presented by points at different or the same positions are different, the texture features in the local range of the target point are analyzed to obtain the texture change degree of the target point; further, the relative position of the target point and the neighboring points and the characteristics of the color value changes of the neighboring points are continuously combined to evaluate the matching degree between the target point in the target point cloud data and the comparison point in the comparison point cloud data, and the matching degree is evaluated according to the matching characteristics. The matching point pairs in the target point cloud data and the comparison point cloud data are divided according to the degree of matching; finally, the target point cloud data and the comparison point cloud data are represented in the same coordinate system according to the Euclidean distance and the matching degree between all the matching point pairs in the target point cloud data and the comparison point cloud data, so as to balance the efficiency and accuracy of point cloud matching. According to the result that the point cloud data of all new energy lightweight vehicles are represented in the same coordinate system, the construction of the green new energy lightweight vehicle point cloud model is completed, the accuracy of the lightweight vehicle point cloud model is improved, and the problem that the value of the convergence threshold cannot balance the efficiency and accuracy of point cloud matching, resulting in inaccurate construction of the lightweight vehicle point cloud model, is solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0046] Figure 1 A schematic flow chart of a method for constructing a point cloud model of a green new energy lightweight automobile provided by an embodiment of the present invention;

[0047] Figure 2 A flowchart for obtaining texture variation provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] See also Figure 1 , which shows a flow chart of a method for constructing a green new energy lightweight automobile point cloud model provided by an embodiment of the present invention, the method comprising the following steps:

[0050] Step S001, collecting point cloud data of a new energy lightweight vehicle with overlapping collection areas from different angles, wherein each point in the point cloud data corresponds to a color value and a three-dimensional coordinate.

[0051] A binocular stereo camera is used to collect point cloud data of new energy lightweight vehicles from different angles. Each point in the point cloud data corresponds to a color value and a three-dimensional coordinate. The color value and three-dimensional coordinate corresponding to the point are both data information of the point.

[0052] It is understandable that the reason for collecting point cloud data of new energy lightweight vehicles from different angles is that the size of the car is large, and different shooting angles need to be changed to ensure the integrity of the point cloud data of the new energy lightweight vehicle. The acquisition areas of the binocular cameras at two adjacent shooting angles need to ensure that there is at least 20% overlap until the complete point cloud data corresponding to the new energy lightweight vehicle is obtained.

[0053] Each acquired point cloud data is pre-processed using voxel grid filtering and statistical filtering to remove redundant information in the point cloud data. Voxel grid filtering and statistical filtering are both well-known technologies and will not be described in detail.

[0054] At this point, point cloud data of new energy lightweight vehicles at different angles are obtained.

[0055] Step S002, record any point cloud data as target point cloud data, record any point in the target point cloud data as the target point, determine the neighboring points of the target point based on the spatial position distribution relationship between the target point and other points in the target point cloud data, determine the derivative sequence and color change trend of the target point based on the distribution of color values ​​and spatial position distribution of the target point and all its neighboring points, and determine the texture change degree of the target point based on the color values, derivative sequences and color change trends of all its neighboring points.

[0056] When aligning point cloud data with overlapping acquisition areas, the coordinate systems of the three-dimensional coordinates corresponding to the two point cloud data are different due to different shooting angles, which is equivalent to performing the same spatial translation transformation on all points in the corresponding overlapping area of ​​one of the point cloud data. Therefore, the texture features presented by the points in the corresponding overlapping area of ​​the two point cloud data are still consistent with the adjacent points.

[0057] It is understandable that the car body is usually made of metal and sprayed with paint, so the car body presents the characteristics of high reflectivity and continuity, that is, the texture changes of each point corresponding to the car body area in the point cloud data are relatively small; there are patterns and geometric hollows on the surface of the car wheel, showing low reflectivity, that is, the texture changes of each point corresponding to the wheel in the point cloud data are relatively large; there are obvious gaps at the junctions between different parts and modules of the car, and the characteristics and textures of different parts and modules are different, so the texture of each point corresponding to the junction position between different parts and modules of the car in the point cloud data will show mutations. Therefore, there are differences in the texture features presented by points at different or the same positions.

[0058] Any point cloud data is recorded as target point cloud data, and any point in the target point cloud data is recorded as the target point.

[0059] Construct a KD tree for the target point, and obtain the first preset value of neighboring points in the target point cloud data that are closest to the target point in space. Perform plane fitting on the target point and all its neighboring points, obtain the normal vector of the fitting plane, arrange the color value of the target point and the color values ​​of all its neighboring points in the order of the points in the direction of the normal vector, and obtain the color value sequence of the target point. Perform polynomial fitting on the color value sequence of the target point, obtain the fitting curve of the target point, calculate the derivatives corresponding to each color value in the color value sequence of the target point on the fitting curve of the target point, arrange the derivatives corresponding to each color value in the color value sequence of the target point in the order of the color value in the color value sequence, and obtain the derivative sequence of the target point. Use the second preset value as the window length, process the derivative sequence of the target point using the moving variance method, obtain the moving variance sequence of the target point, perform linear fitting on the moving variance sequence, obtain the moving variance straight line equation of the target point, and record the slope of the moving variance straight line equation of the target point as the color change trend of the target point.

[0060] Among them, in this embodiment, the value of the first preset value is 100, and the value of the second preset value is 5; the least squares method is used to perform plane fitting on the target point and all the neighboring points of the target point, the least squares method is used to perform polynomial fitting on the color value sequence, and the least squares method is used to perform straight line fitting on the moving variance sequence, wherein the use of the least squares method for straight line fitting, plane fitting and polynomial fitting are well-known technologies and will not be repeated; using the moving variance method to obtain a moving variance sequence for a sequence, and constructing a KD tree for the target point are both well-known technologies and will not be repeated.

[0061] When the color value of the target point and the color values ​​of all the neighboring points of the target point are arranged in the order of the points in the normal vector direction, all the color values ​​of the color value of the target point and the color values ​​of all the neighboring points of the target point are arranged. At the same time, if there are multiple points with the same order in the normal vector direction, these points are arranged in the order of the size of the coordinate values ​​of the three-dimensional coordinates. Specifically, the x-axis coordinates of the points with the same order in the normal vector direction are first compared, and the points are arranged in the order of the x-axis coordinate values ​​from small to large. When the x-axis coordinate values ​​of these points are the same, the y-axis coordinates are continued to be compared, and the points with the same x-axis coordinate values ​​are arranged in the order of the y-axis coordinate values ​​from small to large. When the y-axis coordinate values ​​of the points with the same x-axis coordinate values ​​are also the same, the points with the same y-axis coordinate values ​​are arranged in the order of the z-axis coordinate values ​​from small to large. Since the three-dimensional coordinates of different points in the point cloud data are different, the color value sequence of the target point can be obtained according to the above rules.

[0062] The number of zeros contained in the derivative sequence of the target point is the number of neighborhood points where the color value along the normal vector direction of the fitting plane where the target point is located has obvious changes. The absolute value of the color change trend of the target point is used to evaluate the color value change trend along the normal vector direction of the fitting plane where the target point is located. The faster the color value changes, the greater the absolute value of the color change trend of the target point.

[0063] The texture variation degree of the target point is determined based on the color values, derivative sequences and color variation trends of all neighboring points of the target point.

[0064] Preferably, as an embodiment of the present application, the ratio of the number of 0s contained in the derivative sequence of the target point to the number of all neighboring points of the target point is recorded as the first ratio of the target point; the absolute value of the color change trend of the target point is recorded as the first absolute value of the target point, the sum of the first absolute value of the target point and the number 1 is recorded as the first sum value of the target point, the first sum value of the target point is taken as the true number of a logarithmic function with base 10, and the value of the logarithmic function is recorded as the first function value of the target point; the product of the first function value of the target point, the first ratio and the standard deviation of the color values ​​of all neighboring points of the target point is recorded as the texture variation degree of the target point.

[0065] The texture variation acquisition flow chart is as follows Figure 2 shown.

[0066] The same method can be used to obtain the texture variation of any point in any point cloud data.

[0067] Since the degree of change of texture features corresponding to the car body, the junction between different modules of the car and the car wheels increases successively, the degree of texture change of points corresponding to the car body, the junction between different modules of the car and the car wheels increases successively.

[0068] At this point, the texture variation of any point in any point cloud data can be obtained in the same way.

[0069] Step S003, determine the color distribution sequence of the target point according to the relative position relationship and the difference between the color values ​​between the target point and all neighboring points, record any point cloud data different from the point cloud data where the target point is located as comparison point cloud data, record any point in the comparison point cloud data as a comparison point, determine the matching degree of the target point and the comparison point according to the data distribution of the texture variation degree of all points contained in the two point cloud data where the target point and the comparison point are located, the correlation between the color distribution sequence of the target point and the comparison point, and the texture variation degree of the target point and the comparison point, and divide the matching point pairs in the target point cloud data and the comparison point cloud data according to the matching degree.

[0070] Furthermore, the features of the target point are extracted by combining the relative positions of the target point and the neighboring points and the features of the color value changes of the neighboring points.

[0071] Get the Chebyshev distance between the target point and each neighboring point, and at the same time, calculate the absolute value of the difference in color value between the target point and each neighboring point, and record the absolute value of the difference in color value between the target point and the neighboring point as the color difference of the neighboring point. Arrange the color differences of all the neighboring points of the target point in the order of the Chebyshev distance between the neighboring point and the target point from small to large, and obtain the color distribution sequence of the target point.

[0072] It should be noted that when the Chebyshev distances between different neighboring points and the target point are the same, the color differences of the neighboring points are arranged in order from small to large.

[0073] Any point cloud data different from the point cloud data where the target point is located is recorded as comparison point cloud data, and any point in the comparison point cloud data is recorded as a comparison point.

[0074] The texture variation of all points contained in the two point cloud data where the target point and the comparison point are located are clustered to obtain clusters. When the texture variation of the target point and the comparison point are in the same cluster, the direct matching value between the target point and the comparison point is recorded as 1; when the texture variation of the target point and the comparison point are not in the same cluster, the direct matching value between the target point and the comparison point is recorded as 0.

[0075] This embodiment uses the K-means clustering algorithm to cluster the texture variation, and the number of clusters is 3. Using the K-means clustering algorithm to cluster the texture variation is a well-known technology and will not be described in detail.

[0076] When the target point and the comparison point are two points corresponding to the same position of the car in different point cloud data, the texture variation of the target point and the comparison point is relatively close, the texture variation of the target point and the comparison point are in the same cluster, and the direct matching value of the target point and the comparison point is 1.

[0077] The matching degree between the target point and the comparison point is determined according to the correlation of the color distribution sequences of the target point and the comparison point, the direct matching value, and the texture variation between the target point and the comparison point.

[0078] Preferably, as an embodiment of the present application, the absolute value of the difference in texture variation between the target point and the comparison point is recorded as the texture variation difference between the target point and the comparison point, the ratio of the correlation of the color distribution sequence of the target point and the comparison point to the texture variation difference is recorded as the second ratio of the target point to the comparison point, and the product of the second ratio of the target point to the comparison point and the direct matching value is recorded as the matching degree between the target point and the comparison point.

[0079] Preferably, the correlation between the color distribution sequences of the target point and the comparison point can be the Pearson correlation coefficient. As other implementations, on the basis of achieving the purpose of measuring the correlation of the two sequences, the implementer can adopt other methods in the prior art such as cosine correlation to obtain the correlation of the two sequences, and this application does not impose any special restrictions.

[0080] When the target point and the comparison point are two points corresponding to the same position of the car in different point cloud data, the difference between the texture change degree of the target point and the comparison point is small, the direct matching value between the target point and the comparison point is 1, and the correlation between the color distribution sequence of the target point and the comparison point is large. At this time, the matching degree between the target point and the comparison point is large.

[0081] The same method can be used to obtain the matching degree between any point in the target point cloud data and any point in the comparison point cloud data.

[0082] The maximum value of the matching degree between the target point and all the points in the comparison point cloud data is taken as the matching point of the target point in the comparison point cloud data. The matching point of any point in the target point cloud data in the comparison point cloud data can be obtained in the same way.

[0083] The matching degrees of all points in the target point cloud data and the matching points in the comparison point cloud data are divided using the maximum inter-class variance method to obtain a division threshold, and the points in the target point cloud data with a matching degree greater than the division threshold and the matching points in the comparison point cloud data are taken as a set of matching point pairs.

[0084] Using the maximum inter-class variance method to obtain the segmentation threshold is a well-known technique and will not be described in detail.

[0085] The same method can be used to obtain the corresponding matching point pairs in any two different point cloud data.

[0086] At this point, the corresponding matching point pairs in any two different point cloud data are obtained.

[0087] Step S004, based on the Euclidean distance and matching degree between all matching point pairs in the target point cloud data and the comparison point cloud data, the target point cloud data and the comparison point cloud data are represented in the same coordinate system, and based on the result that the point cloud data of all new energy lightweight vehicles are represented in the same coordinate system, the construction of the green new energy lightweight vehicle point cloud model is completed.

[0088] When using the ICP point cloud registration algorithm to register the target point cloud data with the comparison point cloud data, the value of the convergence threshold determines the accuracy of the registration result. When the value of the convergence threshold is small, the accuracy of the registration result is high, but it will cause too many computing resources to be consumed in areas with less detail information in the point cloud data, resulting in too low operating efficiency of the point cloud matching. When the value of the convergence threshold is large, the accuracy of the registration result is insufficient. Therefore, the value of the convergence threshold is determined according to the matching point pairs to balance the registration efficiency and registration accuracy.

[0089] Specifically, when the matching degree of the matching point pair is greater, the texture consistency in the local range of the two points corresponding to the matching point pair is stronger, and the detail features in the local range of the two points corresponding to the matching point pair are fewer, a larger convergence threshold should be selected.

[0090] The average of the Euclidean distances between all matching point pairs in the target point cloud data and the comparison point cloud data is recorded as the matching distance between the target point cloud data and the comparison point cloud data, and the average of the matching degrees of all matching point pairs in the target point cloud data and the comparison point cloud data is recorded as the average matching degree between the target point cloud data and the comparison point cloud data.

[0091] According to the matching distance and average matching degree between the target point cloud data and the comparison point cloud data, the optimal convergence threshold between the target point cloud data and the comparison point cloud data is determined.

[0092]

[0093] in, Indicates the optimal convergence threshold of the target point cloud data and the comparison point cloud data; represents the initial value of the convergence threshold. In this embodiment, the initial value of the convergence threshold is 0.1 meter; represents a natural constant; Indicates the average matching degree between the target point cloud data and the comparison point cloud data; Indicates the matching distance between the target point cloud data and the comparison point cloud data; represents the inverse tangent function; In this embodiment, pi is calculated with two decimal places.

[0094] When the matching distance between the target point cloud data and the comparison point cloud data is smaller and the average matching degree is larger, the optimal convergence threshold of the target point cloud data and the comparison point cloud data is smaller, the target point cloud data and the comparison point cloud data contain more detailed information, and the accuracy of the registration result is higher.

[0095] It should be noted that when aligning the target point cloud data with the comparison point cloud data, the distance between the points contained in the point cloud includes units, and the convergence threshold of the ICP point cloud alignment algorithm also includes units. Therefore, the initial value of the convergence threshold includes units, and the optimal convergence threshold of the target point cloud data and the comparison point cloud data also includes units, and the specific unit is meter.

[0096] The optimal convergence threshold of the target point cloud data and the comparison point cloud data is used as the value of the convergence threshold, and the target point cloud data and the comparison point cloud data are registered using the ICP point cloud registration algorithm, and the target point cloud data and the comparison point cloud data are represented in the same coordinate system. The same method can be used to obtain the result of the point cloud data of any two different new energy lightweight vehicles collected in the same coordinate system. Furthermore, the result of the point cloud data of all new energy lightweight vehicles collected in the same coordinate system can be obtained.

[0097] The Poisson surface reconstruction algorithm is used to obtain the point cloud model of the green new energy lightweight vehicle based on the result that the point cloud data of all new energy lightweight vehicles are represented in the same coordinate system.

[0098] Among them, using the ICP point cloud registration algorithm to register the target point cloud data with the comparison point cloud data and using the Poisson surface reconstruction algorithm to obtain the point cloud model are both well-known technologies and will not be described in detail.

[0099] At this point, the construction of the green new energy lightweight vehicle point cloud model is completed.

[0100] Based on the same inventive concept as the above method, an embodiment of the present invention also provides a system for constructing a green new energy lightweight automobile point cloud model, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned methods for constructing a green new energy lightweight automobile point cloud model are implemented.

[0101] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for constructing a green new energy lightweight automobile point cloud model, characterized in that: The method comprises the following steps: Collect point cloud data of new energy lightweight vehicles with overlapping collection areas from different angles. Each point in the point cloud data corresponds to a color value and a three-dimensional coordinate. Record any point cloud data as target point cloud data, record any point in the target point cloud data as the target point, determine the neighboring points of the target point according to the spatial position distribution relationship between the target point and other points in the target point cloud data, determine the derivative sequence and color change trend of the target point according to the distribution of color values ​​and spatial position distribution of the target point and all the neighboring points of the target point, and determine the texture change degree of the target point according to the color values, derivative sequences and color change trends of all the neighboring points of the target point; According to the relative position relationship and the difference between the color values ​​of the target point and all the neighboring points, the color distribution sequence of the target point is determined, any point cloud data different from the point cloud data where the target point is located is recorded as the comparison point cloud data, and any point in the comparison point cloud data is recorded as the comparison point. According to the data distribution of the texture variation of all points contained in the two point cloud data where the target point and the comparison point are located, the correlation of the color distribution sequence of the target point and the comparison point, and the texture variation of the target point and the comparison point, the matching degree of the target point and the comparison point is determined, and the matching point pairs in the target point cloud data and the comparison point cloud data are divided according to the matching degree; According to the Euclidean distance and matching degree between all matching point pairs in the target point cloud data and the comparison point cloud data, the target point cloud data and the comparison point cloud data are represented in the same coordinate system. According to the result that the point cloud data of all new energy lightweight vehicles are represented in the same coordinate system, the construction of the green new energy lightweight vehicle point cloud model is completed. The method for determining the derivative sequence of the target point is: Perform plane fitting on the target point and all its neighboring points to obtain the normal vector of the fitting plane, arrange the color values ​​of the target point and all its neighboring points in the order of the points in the direction of the normal vector to obtain the color value sequence of the target point; perform polynomial fitting on the color value sequence of the target point to obtain the fitting curve of the target point, calculate the derivative corresponding to each color value in the color value sequence of the target point on the fitting curve of the target point, arrange the derivative corresponding to each color value in the color value sequence of the target point in the order of the color value in the color value sequence to obtain the derivative sequence of the target point.

2. The method for constructing a green new energy lightweight automobile point cloud model according to claim 1, characterized in that: The specific method for determining the neighboring points of the target point is: The first preset number of points in the target point cloud data that are closest to the target point in space are recorded as neighboring points of the target point.

3. The method for constructing a green new energy lightweight automobile point cloud model according to claim 1, characterized in that: The method for determining the color change trend is: The second preset value is used as the window length, and the moving variance method is used to process the derivative sequence of the target point to obtain the moving variance sequence of the target point. A straight line fitting is performed on the moving variance sequence to obtain the moving variance straight line equation of the target point, and the slope of the moving variance straight line equation of the target point is recorded as the color change trend of the target point.

4. The method for constructing a green new energy lightweight automobile point cloud model according to claim 1, characterized in that: The method of determining the texture variation degree of the target point according to the color values, derivative sequences and color variation trends of all neighboring points of the target point includes the following specific methods: The ratio of the number of 0s contained in the derivative sequence of the target point to the number of all neighboring points of the target point is recorded as the first ratio of the target point; The absolute value of the color change trend of the target point is recorded as the first absolute value of the target point, the sum of the first absolute value of the target point and the number 1 is recorded as the first sum value of the target point, the first sum value of the target point is used as the true number of the logarithmic function with base 10, and the value of the logarithmic function is recorded as the first function value of the target point; The product of the first function value of the target point, the first ratio and the standard deviation of the color values ​​of all neighboring points of the target point is recorded as the texture variation degree of the target point.

5. The method for constructing a green new energy lightweight automobile point cloud model according to claim 1, characterized in that: The method for determining the color distribution sequence of the target point is: Calculate the Chebyshev distance between the target point and each neighboring point; The absolute value of the difference between the color values ​​of the target point and the neighboring points is recorded as the color difference of the neighboring points; The color differences of all neighboring points of the target point are arranged in ascending order according to the Chebyshev distance between the neighboring point and the target point to obtain the color distribution sequence of the target point.

6. The method for constructing a green new energy lightweight automobile point cloud model according to claim 1, characterized in that: The method of determining the matching degree between the target point and the comparison point according to the data distribution of the texture variation degree of all points contained in the two point cloud data where the target point and the comparison point are located, the correlation of the color distribution sequence of the target point and the comparison point, and the texture variation degree of the target point and the comparison point includes the following specific methods: Clustering the texture variation of all points contained in the two point cloud data where the target point and the comparison point are located to obtain cluster clusters; When the texture variation of the target point and the comparison point is in the same cluster, the direct matching value between the target point and the comparison point is recorded as 1; when the texture variation of the target point and the comparison point is not in the same cluster, the direct matching value between the target point and the comparison point is recorded as 0; The matching degree between the target point and the comparison point is determined according to the correlation of the color distribution sequences of the target point and the comparison point, the direct matching value, and the texture variation between the target point and the comparison point.

7. The method for constructing a green new energy lightweight automobile point cloud model according to claim 6, characterized in that: The method of determining the matching degree between the target point and the comparison point according to the correlation of the color distribution sequences between the target point and the comparison point, the direct matching value, and the texture variation between the target point and the comparison point includes the following specific methods: The absolute value of the difference between the texture variation of the target point and the comparison point is recorded as the texture variation difference between the target point and the comparison point, and the ratio of the correlation between the color distribution sequence of the target point and the comparison point to the texture variation difference is recorded as the second ratio between the target point and the comparison point; The product of the second ratio between the target point and the comparison point and the direct matching value is recorded as the matching degree between the target point and the comparison point.

8. The method for constructing a green new energy lightweight automobile point cloud model according to claim 1, characterized in that: The specific method of dividing the matching point pairs in the target point cloud data and the comparison point cloud data according to the matching degree includes: The maximum value of the matching degree between the target point and all the points in the comparison point cloud data is taken as the matching point of the target point in the comparison point cloud data; Obtain the matching point of any point in the target point cloud data in the comparison point cloud data; The matching degrees of all points in the target point cloud data and the matching points in the comparison point cloud data are divided using the maximum inter-class variance method to obtain a division threshold, and the points in the target point cloud data with a matching degree greater than the division threshold and the matching points in the comparison point cloud data are taken as a set of matching point pairs.

9. The method for constructing a green new energy lightweight automobile point cloud model according to claim 1, characterized in that: The target point cloud data and the comparison point cloud data are represented in the same coordinate system according to the Euclidean distance and matching degree between all matching point pairs in the target point cloud data and the comparison point cloud data, and the construction of the green new energy lightweight vehicle point cloud model is completed according to the result of representing the point cloud data of all new energy lightweight vehicles in the same coordinate system, including the specific method of: The mean value of the Euclidean distances between all matching point pairs in the target point cloud data and the comparison point cloud data is recorded as the matching distance between the target point cloud data and the comparison point cloud data, and the mean value of the matching degrees between all matching point pairs in the target point cloud data and the comparison point cloud data is recorded as the average matching degree between the target point cloud data and the comparison point cloud data; According to the matching distance and average matching degree between the target point cloud data and the comparison point cloud data, the optimal convergence threshold between the target point cloud data and the comparison point cloud data is determined. The specific calculation formula is: in, Indicates the optimal convergence threshold of the target point cloud data and the comparison point cloud data; Indicates the initial value of the convergence threshold; represents a natural constant; Indicates the average matching degree between the target point cloud data and the comparison point cloud data; Indicates the matching distance between the target point cloud data and the comparison point cloud data; represents the inverse tangent function; represents pi; The optimal convergence threshold of the target point cloud data and the comparison point cloud data is taken as the value of the convergence threshold, and the target point cloud data and the comparison point cloud data are registered using the ICP point cloud registration algorithm, and the target point cloud data and the comparison point cloud data are represented in the same coordinate system; Obtain the result of expressing the collected point cloud data of any two different new energy lightweight vehicles in the same coordinate system; obtain the result of expressing the collected point cloud data of all new energy lightweight vehicles in the same coordinate system; The Poisson surface reconstruction algorithm is used to obtain the point cloud model of the green new energy lightweight vehicle based on the result that the point cloud data of all new energy lightweight vehicles are represented in the same coordinate system.

10. A system for constructing a point cloud model of a green new energy lightweight automobile, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

Images