Cross-platform laser point cloud data registration quality evaluation method

Abstract

The invention belongs to the technical field of point cloud data processing, and particularly relates to a cross-platform laser point cloud data registration quality evaluation method, which comprises the following steps: for registered cross-platform laser point clouds, extracting spatial geometric features of points, straight lines and planes from laser point clouds obtained by platforms; constructing a cross-platform laser point registration quality evaluation index based on spatial point features, a cross-platform laser point registration quality evaluation index based on spatial straight line features and a cross-platform laser point registration quality evaluation index based on spatial plane features, and combining to form a cross-platform laser point cloud registration quality evaluation index based on spatial features; and performing cross-platform laser point cloud registration quality comprehensive measurement evaluation on the registered cross-platform laser point cloud by using the cross-platform laser point cloud registration quality evaluation index based on the spatial characteristics. According to the method, the registration quality of the multi-source point cloud data can be evaluated more comprehensively and accurately.

Description

Technical Field

[0001] The present invention belongs to the technical field of point cloud data processing, and in particular relates to a cross-platform laser point cloud data registration quality evaluation method. Background Art

[0002] With the rapid development of laser scanning technology, laser point cloud data has become an important data source in many fields such as geospatial information acquisition, urban planning, environmental monitoring and autonomous driving. However, in practical applications, due to the complex and changeable data acquisition environment and the performance differences between different platforms and sensors, the acquired laser point cloud data has problems such as diversity, multi-source and heterogeneity. Therefore, the precise registration of cross-platform laser point cloud data has become the key to data fusion applications, and the registration quality directly affects the accuracy and reliability of subsequent data processing and applications.

[0003] At present, there are many studies on multi-platform laser point cloud data registration methods, and the types of registration methods are diverse. However, there are few studies on registration quality evaluation, and they lack systematicity. Most of the existing laser point cloud data registration quality evaluation methods are based on a single feature or a single indicator, which is difficult to fully reflect the differences and uncertainties in scale, rotation, translation, etc. of multi-source cross-platform point cloud data. Summary of the invention

[0004] In view of the above deficiencies in the prior art, the present invention provides a cross-platform laser point cloud data registration quality evaluation method, which can more comprehensively and accurately evaluate the quality of multi-source point cloud data registration.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is a cross-platform laser point cloud data registration quality evaluation method, comprising the following steps: S1. For the registered cross-platform laser point cloud (refers to aligning and fusing laser radar point cloud data from different sources or platforms through technical means to accurately reflect the spatial information of the real world in the same coordinate system; these platforms include drones, ground laser scanners, vehicle-mounted laser radar systems, etc.; the registration process can make up for the shortcomings of a single data source, improve the integrity and accuracy of the data, and enhance the details and continuity of three-dimensional spatial information; this process can be achieved by known technologies), extract the spatial geometric features of points, lines, and planes in the laser point clouds obtained from each platform respectively; S2, construct a cross-platform laser point registration quality evaluation index based on spatial point features; S3, construct a cross-platform laser point registration quality evaluation index based on spatial straight line features; S4, construct a cross-platform laser point registration quality evaluation index based on spatial plane features; S5, combining the indicators in S2-S4 to form a cross-platform laser point cloud registration quality evaluation indicator based on spatial features; S6. For the registered cross-platform laser point cloud, a comprehensive measurement evaluation of the cross-platform laser point cloud registration quality is performed using a cross-platform laser point cloud registration quality evaluation index based on spatial features.

[0006] As a preferred solution of the present invention, in S2, the process of constructing a cross-platform laser point registration quality evaluation index based on spatial point features is: S21. For the cross-platform laser point cloud after registration of two platforms, k pairs of feature points with the same name are extracted from the laser point clouds of the two platforms, which are expressed as , , i=1,2,…,k; where and represents the coordinates of the corresponding point in the Cartesian coordinate system; S22. The quality evaluation indicators of cross-platform laser point registration based on spatial point features include: and ,in It is an indicator for evaluating the average deviation between cross-platform laser point clouds after registration. It is an indicator to evaluate the average deviation direction between cross-platform laser point clouds after registration, which can be expressed as: (1); (2); In the formula, α and β are weight coefficients; Representation Matrix The corresponding maximum eigenvalue; Representation Matrix The eigenvector corresponding to the largest eigenvalue; denotes a normalized vector; the superscript T denotes the transpose.

[0007] As a preferred solution of the present invention, in S22, the values ​​of α and β are both 1 / 2.

[0008] As a preferred solution of the present invention, in said S3, the process of constructing a cross-platform laser point registration quality evaluation index based on spatial straight line features is: S31. Extract c pairs of feature lines with the same name from the laser point clouds of the two platforms, respectively: and , j = 1, 2, ..., c; , The corresponding direction vectors are , ; S32, cross-platform laser point registration quality evaluation indicators based on spatial straight line features include and ,in It is an indicator to evaluate the degree of rotation between cross-platform laser point clouds after registration. It is an indicator for evaluating the rotation direction between cross-platform laser point clouds after registration, which can be expressed as: (3); (4); In the formula, Representation Matrix The eigenvector corresponding to the largest eigenvalue.

[0009] As a preferred solution of the present invention, in said S4, the process of constructing a cross-platform laser point registration quality evaluation index based on spatial plane features is: S41. Extract h pairs of surface elements with the same name and complete boundaries from the laser point clouds of the two platforms, and calculate the areas and normal vectors of the surface elements with the same name respectively (using known techniques), and set their areas to be , , the normal vectors are , , f = 1, 2, ..., h; S42. The quality evaluation indicators of cross-platform laser point registration based on spatial plane features include: and ,in It is an indicator for evaluating the scale consistency between cross-platform laser point clouds after registration. It is an indicator to evaluate the consistency of the rotation direction between the laser point clouds across platforms after registration, which can be expressed as: (5); (6); In the formula, For the matrix The eigenvector corresponding to the largest eigenvalue.

[0010] As a preferred embodiment of the present invention, in S5, , , , , , Together they constitute a cross-platform laser point cloud registration quality evaluation index based on spatial features.

[0011] As a preferred solution of the present invention, in S6, the process of performing comprehensive metric evaluation of cross-platform laser point cloud registration quality is as follows: S61, dividing the registered cross-platform laser point cloud into M regions according to the density of the laser point cloud; S62, extracting spatial geometric features of points, lines, and planes in each region respectively. If a line feature or a surface element feature cannot be extracted from a region, the number of feature lines or surface elements (i.e., plane features) with the same name in the region is zero; S63. According to the spatial geometric features extracted in each area, the cross-platform laser point cloud registration quality evaluation index based on spatial features is calculated for each area, including , , , , , , m=1,2,…,M, , , , , , That is, the mth region , , , , , ; S64. Calculate the comprehensive evaluation index of each area separately , ,in It is an indicator to evaluate the degree of alignment between laser point clouds in the cross-platform area after registration. It is an indicator for evaluating the alignment direction between laser point clouds in the cross-platform area after registration, which can be expressed as: (7); (8); In the formula, , , are the weights of the registration quality evaluation index of the point, line, and plane features in the mth region respectively; S65, based on the calculated values ​​of each region and , generate heat maps and vector maps (using known techniques), visually displaying the distribution of laser point cloud registration quality in each area; S66, calculate the M areas and The average value of the registration quality of the laser point cloud in the entire area is obtained. , ,in It is an indicator for evaluating the comprehensive alignment between cross-platform laser point clouds after registration. It is an indicator for evaluating the comprehensive alignment direction between cross-platform laser point clouds after registration, which can be expressed as: (9); (10).

[0012] As a preferred embodiment of the present invention, in the S61, the area division method is to establish a plane grid according to a preset size, then project the original laser point cloud into the corresponding grid, extract density features from the point set in each grid, and perform area division based on the density features.

[0013] As a preferred embodiment of the present invention, in the S62, the spatial geometric features of points, lines and planes are extracted in each area respectively, and then the corresponding features are matched to obtain the feature points, feature lines and surface elements with the same name in the area. Specifically, for points, the feature points with the same name are obtained through key point extraction, feature descriptor generation and descriptor matching; for line features, a depth map is generated according to the laser point cloud, and the line features are extracted and matched in the depth map; for plane features, plane fitting and clustering methods are used to extract planes, and plane feature matching is performed according to the criteria of minimum plane normal vector angle and maximum plane overlap area.

[0014] As a preferred embodiment of the present invention, in the S64, , , The calculation method is: (11); (12); (13); In the formula, , , are the numbers of feature points, feature lines and surface elements with the same name extracted in the mth region respectively.

[0015] The algorithm involved in the present invention can be executed by an electronic device. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The algorithm is implemented by executing the program by the processor.

[0016] The beneficial effects of the present invention are: The present invention comprehensively considers multiple spatial features such as points, lines, and planes, and constructs a multi-feature joint registration quality evaluation system, which effectively solves the problem of single quality evaluation indicators and single features in existing laser point cloud data registration. Compared with existing laser point cloud data registration quality evaluation methods, it can more comprehensively reflect the differences and uncertainties of multi-source point cloud data in terms of scale, rotation, translation, etc., more comprehensively and accurately evaluate the quality of multi-source point cloud data registration, and can perform visual expression to intuitively display the distribution of point cloud registration quality in each area. The application of the present invention can provide strong support for the precise registration and high-quality fusion of cross-platform laser point cloud data. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow diagram of the present invention. DETAILED DESCRIPTION

[0018] The embodiments of the present invention are further described below in conjunction with the accompanying drawings: like Figure 1 As shown, a cross-platform laser point cloud data registration quality evaluation method includes the following steps: S1. For the laser point cloud obtained by each platform after cross-platform registration, the spatial geometric features of points, lines and planes are extracted; S2, construct a cross-platform laser point registration quality evaluation index based on spatial point features; S3, construct a cross-platform laser point registration quality evaluation index based on spatial straight line features; S4, construct a cross-platform laser point registration quality evaluation index based on spatial plane features; S5, combining the indicators in S2-S4 to form a cross-platform laser point cloud registration quality evaluation indicator based on spatial features; S6. For the registered cross-platform laser point cloud, a comprehensive measurement evaluation of the cross-platform laser point cloud registration quality is performed using a cross-platform laser point cloud registration quality evaluation index based on spatial features.

[0019] In S2, the process of constructing the cross-platform laser point registration quality evaluation index based on spatial point features is as follows: S21. For the cross-platform laser point cloud after registration of two platforms, point features are the most basic elements in laser point cloud data. They directly reflect the spatial position and shape of objects. k pairs of feature points with the same name are extracted from the laser point clouds of the two platforms, which are expressed as , , i=1,2,…,k; where and represents the coordinates of the corresponding point in the Cartesian coordinate system; S22. The quality evaluation indicators of cross-platform laser point registration based on spatial point features include: and ,in It is an indicator for evaluating the average deviation between cross-platform laser point clouds after registration. It is an indicator to evaluate the average deviation direction between cross-platform laser point clouds after registration, which can be expressed as: (1); (2); In the formula, α and β are weight coefficients, both of which are 1 / 2; Representation Matrix The corresponding maximum eigenvalue; Representation Matrix The eigenvector corresponding to the largest eigenvalue; denotes a normalized vector; the superscript T denotes the transpose.

[0020] If there are multiple platforms, that is, the number of platforms is greater than two, any two platforms can be selected to perform calculations according to the method of this embodiment, and the obtained and Find the average value and use it as the new and , used for subsequent indicator calculations, the same below.

[0021] In S3, the process of constructing the cross-platform laser point registration quality evaluation index based on spatial straight line features is as follows: S31. Compared with point features, spatial straight line features can better reflect the contour and shape information of objects, thus having a unique advantage in evaluating the rotation error after registration. c pairs of feature straight lines with the same name are extracted from the laser point clouds of the two platforms, which are and , j = 1, 2, ..., c; , The corresponding direction vectors are , ; S32, cross-platform laser point registration quality evaluation indicators based on spatial straight line features include and ,in It is an indicator to evaluate the degree of rotation between cross-platform laser point clouds after registration. It is an indicator for evaluating the rotation direction between cross-platform laser point clouds after registration, which can be expressed as: (3); (4); In the formula, Representation Matrix The eigenvector corresponding to the largest eigenvalue.

[0022] In S4, the process of constructing the cross-platform laser point registration quality evaluation index based on spatial plane features is as follows: S41. Extract h pairs of surface elements with the same name and complete boundaries from the laser point clouds of the two platforms, and calculate the areas and normal vectors of the surface elements with the same name respectively. Let their areas be , , the normal vectors are , , f = 1, 2, ..., h; S42. The quality evaluation indicators of cross-platform laser point registration based on spatial plane features include: and ,in It is an indicator for evaluating the scale consistency between cross-platform laser point clouds after registration. It is an indicator to evaluate the consistency of the rotation direction between the laser point clouds across platforms after registration, which can be expressed as: (5); (6); In the formula, For the matrix The eigenvector corresponding to the largest eigenvalue.

[0023] In S5, , , , , , Together they constitute a cross-platform laser point cloud registration quality evaluation index based on spatial features.

[0024] In S6, the process of comprehensive metric evaluation of cross-platform laser point cloud registration quality is as follows: S61, dividing the registered cross-platform laser point cloud into M regions according to the density of the laser point cloud; S62, extracting spatial geometric features of points, lines, and planes in each region respectively. If a line feature or a surface element feature cannot be extracted from a certain region, the number of feature lines or surface elements with the same name in the region is zero; S63. According to the spatial geometric features extracted in each area, the cross-platform laser point cloud registration quality evaluation index based on spatial features is calculated for each area, including , , , , , , m=1,2,…,M, , , , , , That is, the mth region , , , , , ; Used to describe the differences and uncertainties in deviation, rotation, scale, etc. of the laser point cloud in the area after registration; S64. Calculate the comprehensive evaluation index of each area separately , ,in It is an indicator to evaluate the degree of alignment between laser point clouds in the cross-platform area after registration. It is an indicator for evaluating the alignment direction between laser point clouds in the cross-platform region after registration. It is used to describe the overall difference and uncertainty of the laser point clouds in the region after registration, and is expressed as: (7); (8); In the formula, , , are the weights of the registration quality evaluation index of the point, line, and plane features in the mth region respectively; S65, based on the calculated values ​​of each region and , generate heat maps and vector maps to visually display the distribution of laser point cloud registration quality in each area; S66, calculate the M areas and The average value of the registration quality of the laser point cloud in the entire area is obtained. , ,in It is an indicator for evaluating the comprehensive alignment between cross-platform laser point clouds after registration. It is an indicator for evaluating the comprehensive alignment direction between cross-platform laser point clouds after registration, which can be expressed as: (9); (10).

[0025] In the heat map of S65, the color depth can be used to directly reflect the distribution of the registration quality index. The darker the color, the lower the registration quality of the area, and the lighter the color, the higher the registration quality. The vector map shows the distribution of the rotation direction index between the cross-platform point clouds.

[0026] In S61, the region division method is to use the plane grid method to establish a plane grid according to a preset size, then project the original laser point cloud into the corresponding grid, extract density features from the point set in each grid, and perform region division based on the density features. This region division method can reflect the difference in local region registration quality.

[0027] In S62, the spatial geometric features of points, lines and planes are extracted in each area respectively, and then the corresponding features are matched to obtain the feature points, feature lines and surface elements with the same name in the area. Specifically, for points, the feature points with the same name are obtained through key point extraction, feature descriptor generation and descriptor matching; for line features, a depth map is generated according to the laser point cloud, and the line features are extracted and matched in the depth map; for plane features, plane fitting and clustering methods are used to extract planes, and plane feature matching is performed according to the criteria of minimum angle between plane normal vectors and maximum plane overlap area.

[0028] In S64, , , The calculation method is: (11); (12); (13); In the formula, , , are the numbers of feature points, feature lines and surface elements with the same name extracted in the mth region respectively.

Claims

1. A cross-platform laser point cloud data registration quality evaluation method, characterized in that The following steps are involved: S1. For the laser point cloud obtained by each platform after cross-platform registration, the spatial geometric features of points, lines and planes are extracted; S2, construct a cross-platform laser point registration quality evaluation index based on spatial point features; S3, construct a cross-platform laser point registration quality evaluation index based on spatial straight line features; S4, construct a cross-platform laser point registration quality evaluation index based on spatial plane features; S5, combining the indicators in S2-S4 to form a cross-platform laser point cloud registration quality evaluation indicator based on spatial features; S6. For the registered cross-platform laser point cloud, a comprehensive measurement evaluation of the cross-platform laser point cloud registration quality is performed using a cross-platform laser point cloud registration quality evaluation index based on spatial features.

2. A cross-platform laser point cloud data registration quality evaluation method according to claim 1, characterized in that: In S2, the process of constructing a cross-platform laser point registration quality evaluation index based on spatial point features is as follows: S21. For the cross-platform laser point cloud after registration of two platforms, k pairs of feature points with the same name are extracted from the laser point clouds of the two platforms, which are expressed as , , i=1,2,…,k; where and represents the coordinates of the corresponding point in the Cartesian coordinate system; S22. The quality evaluation indicators of cross-platform laser point registration based on spatial point features include: and ,in It is an indicator for evaluating the average deviation between cross-platform laser point clouds after registration. It is an indicator to evaluate the average deviation direction between cross-platform laser point clouds after registration, which can be expressed as: （1）； （2）； In the formula, α and β are weight coefficients; Representation Matrix The corresponding maximum eigenvalue; Representation Matrix The eigenvector corresponding to the largest eigenvalue; denotes a normalized vector; the superscript T denotes the transpose.

3. A cross-platform laser point cloud data registration quality evaluation method according to claim 2, characterized in that: In the above S22, the values ​​of α and β are both 1 / 2.

4. A cross-platform laser point cloud data registration quality evaluation method according to claim 2, characterized in that: In the above S3, the process of constructing a cross-platform laser point registration quality evaluation index based on spatial straight line features is as follows: S31. Extract c pairs of feature lines with the same name from the laser point clouds of the two platforms, respectively: and , j = 1, 2, ..., c; , The corresponding direction vectors are , ; S32, cross-platform laser point registration quality evaluation indicators based on spatial straight line features include and ,in It is an indicator to evaluate the degree of rotation between cross-platform laser point clouds after registration. It is an indicator for evaluating the rotation direction between cross-platform laser point clouds after registration, which can be expressed as: （3）； （4）； In the formula, Representation Matrix The eigenvector corresponding to the largest eigenvalue.

5. A cross-platform laser point cloud data registration quality evaluation method according to claim 4, characterized in that: In the above S4, the process of constructing a cross-platform laser point registration quality evaluation index based on spatial plane features is as follows: S41. Extract h pairs of surface elements with the same name and complete boundaries from the laser point clouds of the two platforms, and calculate the areas and normal vectors of the surface elements with the same name respectively. Suppose their areas are , , the normal vectors are , , f = 1, 2, ..., h; S42. The quality evaluation indicators of cross-platform laser point registration based on spatial plane features include: and ,in It is an indicator for evaluating the scale consistency between cross-platform laser point clouds after registration. It is an indicator to evaluate the consistency of the rotation direction between the laser point clouds across platforms after registration, which can be expressed as: （5）； （6）； In the formula, For the matrix The eigenvector corresponding to the largest eigenvalue.

6. A cross-platform laser point cloud data registration quality evaluation method according to claim 5, characterized in that: In the S5, , , , , , Together they constitute a cross-platform laser point cloud registration quality evaluation index based on spatial features.

7. A cross-platform laser point cloud data registration quality evaluation method according to claim 6, characterized in that: In S6, the process of performing comprehensive metric evaluation of cross-platform laser point cloud registration quality is as follows: S61, dividing the registered cross-platform laser point cloud into M regions according to the density of the laser point cloud; S62, extracting spatial geometric features of points, lines, and planes in each region respectively. If a line feature or a surface element feature cannot be extracted from a certain region, the number of feature lines or surface elements with the same name in the region is zero; S63. According to the spatial geometric features extracted in each area, the cross-platform laser point cloud registration quality evaluation index based on spatial features is calculated for each area, including , , , , , , m=1,2,…,M, , , , , , That is, the mth region , , , , , ; S64. Calculate the comprehensive evaluation index of each area separately , ,in It is an indicator to evaluate the degree of alignment between laser point clouds in the cross-platform area after registration. It is an indicator for evaluating the alignment direction between laser point clouds in the cross-platform area after registration, which can be expressed as: （7）； （8）； In the formula, , , are the weights of the registration quality evaluation index of the point, line, and plane features in the mth region respectively; S65, based on the calculated values ​​of each region and , generate heat maps and vector maps to visually display the distribution of laser point cloud registration quality in each area; S66. Calculate the M regions and The average value of the registration quality of the laser point cloud in the entire area is obtained. , ,in It is an indicator for evaluating the comprehensive alignment between cross-platform laser point clouds after registration. It is an indicator for evaluating the comprehensive alignment direction between cross-platform laser point clouds after registration, which can be expressed as: （9）； （10）。 8. The cross-platform laser point cloud data registration quality evaluation method according to claim 7, characterized in that: In the above-mentioned S61, the region division method is to establish a plane grid according to a preset size, then project the original laser point cloud into the corresponding grid, extract density features from the point set in each grid, and perform region division based on the density features.

9. The cross-platform laser point cloud data registration quality evaluation method according to claim 7, characterized in that: In the above-mentioned S62, the spatial geometric features of points, lines and planes are extracted in each area respectively, and then the corresponding features are matched to obtain the feature points, feature lines and surface elements with the same name in the area. Specifically, for points, the feature points with the same name are obtained through key point extraction, feature descriptor generation and descriptor matching; for line features, a depth map is generated according to the laser point cloud, and the line features are extracted and matched in the depth map; for plane features, plane fitting and clustering methods are used to extract planes, and plane feature matching is performed according to the criteria of minimum plane normal vector angle and maximum plane overlap area.

10. The cross-platform laser point cloud data registration quality evaluation method according to claim 7, characterized in that: In the S64, , , The calculation method is: （11）； （12）； （13）； In the formula, , , are the numbers of feature points, feature lines and surface elements with the same name extracted in the mth region respectively.

Images