Method and device for determining posture, electronic device, and computer-readable storage medium

Abstract

The present disclosure provides a posture determination method and device, an electronic device, and a computer-readable storage medium, wherein the present disclosure first obtains at least two frames of point cloud data collected by a radar; then determines the attribute characteristics of each point in each frame of point cloud data; and for each frame of point cloud data except the first frame of point cloud data in the at least two frames of point cloud data, matches the points in the frame of point cloud data with the points in the previous frame of point cloud data that have the same attribute characteristics to obtain multiple registration point pairs; finally, based on the coordinate information of each point in each registration point pair and the attribute characteristics of each registration point pair, determines the posture feature information of the radar when collecting each frame of point cloud data.

Description

Technical Field

[0001] The present disclosure relates to computer vision technology and the field of autonomous driving, and more specifically, to a method and device for determining a posture, an electronic device, and a computer-readable storage medium. Background Art

[0002] With the development of computer vision and autonomous driving, the requirements for the accuracy of maps used in them are getting higher and higher. In order to generate high-precision maps, a dedicated map collection vehicle is generally used. The map collection vehicle is equipped with a top-mounted laser radar and a high-precision strapdown inertial navigation system to build a high-precision map based on the position of the laser radar. The above solution can meet the requirements of building high-precision maps under the condition of good Global Navigation Satellite System (GNSS) signals (such as highways).

[0003] However, the map constructed by the above scheme has reduced accuracy in the case of weak GNSS, and is also susceptible to inaccurate parameter calibration used in data fusion between high-precision strapdown inertial navigation system and lidar. Summary of the invention

[0004] In view of this, the present disclosure at least provides a method and device for determining posture.

[0005] In a first aspect, the present disclosure provides a method for determining a posture, comprising:

[0006] Obtain at least two frames of point cloud data collected by the radar;

[0007] Determine the attribute characteristics of each point in each frame of point cloud data;

[0008] For each frame of point cloud data except the first frame of point cloud data in the at least two frames of point cloud data, matching points in the frame of point cloud data with points in a previous frame of point cloud data having the same attribute features to obtain a plurality of registration point pairs;

[0009] Based on the coordinate information of each point in each registration point pair and the attribute characteristics of each registration point pair, the position and posture characteristic information of the radar when collecting each frame of point cloud data is determined.

[0010] In this aspect, by pairing points with the same attribute characteristics in the point cloud data, the success rate and accuracy of the point cloud data registration can be improved. Then, by using more accurate registration point pairs, the accuracy of the determined posture feature information can be improved. Subsequently, using accurate posture feature information, a more accurate map can be generated.

[0011] In a possible implementation manner, before determining the attribute features of each point in each frame of point cloud data, the method further includes:

[0012] Eliminate points in each frame of point cloud data whose distance from the radar is greater than a first set threshold;

[0013] The points in each frame of point cloud data whose reflectivity is less than the second set threshold are downsampled.

[0014] In this embodiment, points that are far away from the radar and have low reflectivity are eliminated, which is beneficial to improving the accuracy of the determined attribute features and posture feature information.

[0015] In a possible implementation manner, the attribute feature includes whether it is a ground point;

[0016] Determining the attribute features of each point in each frame of point cloud data includes:

[0017] Determine the preset segmentation area to which each point in the frame belongs according to the coordinate information of each point;

[0018] For each preset segmentation area, if the distances between points in other preset segmentation areas adjacent to the preset segmentation area and the ground are all smaller than the shortest distance between the point in the preset segmentation area and the ground, it is determined that the point in the preset segmentation area is not a ground point.

[0019] This embodiment utilizes the distance between the points in each preset segmentation area and the ground to more accurately determine whether the corresponding point is a ground point, that is, it can more accurately determine the attribute characteristics of the point, which is beneficial to improve the accuracy of the determined posture feature information based on more accurate attribute characteristics.

[0020] In a possible implementation manner, when the attribute feature of any point includes a feature of a non-ground point, the feature of the non-ground point includes a shape feature, and the shape feature is any one of a linear feature, a planar feature, and a spherical feature;

[0021] The determining of the attribute features of each point in each frame of point cloud data also includes:

[0022] For any point that is not a ground point, a plurality of points whose distances from the any point are within a preset distance range are obtained; and based on the obtained plurality of points, a three-dimensional feature value of the any point is determined;

[0023] Based on the three-dimensional eigenvalues, determining the linear eigenvalue, the planar eigenvalue and the spherical eigenvalue corresponding to the any point;

[0024] Based on the linearity eigenvalues, the planarity eigenvalues ​​and the sphericality eigenvalues, a shape feature corresponding to any point cloud is determined.

[0025] In this implementation manner, a plurality of points close to a certain point are used to determine the attribute features of the point cloud data, thereby improving the accuracy of the attribute features of the point being determined.

[0026] In a possible implementation manner, determining the shape feature corresponding to any point based on the linear feature value, the planar feature value, and the spherical feature value includes:

[0027] If the linear feature value of a point is greater than the fourth set threshold, it is determined that the point has a linear feature.

[0028] In this embodiment, by using the linear feature value of a certain point and the fourth set threshold, it is possible to more accurately determine whether the point has a linear feature.

[0029] In a possible implementation manner, determining the shape feature corresponding to any point cloud based on the linearity eigenvalue, the planarity eigenvalue, and the sphericality eigenvalue further includes:

[0030] If the linear feature value of a point is less than or equal to the fourth set threshold, and the surface feature value of the point is greater than the fifth set threshold, it is determined that the point has the surface feature.

[0031] In this embodiment, by using the linear feature value, planar feature value, fourth set threshold value and fifth set threshold value of a certain point, it is possible to more accurately determine whether the point has the planar feature.

[0032] In a possible implementation manner, determining the shape feature corresponding to any point cloud based on the linearity eigenvalue, the planarity eigenvalue, and the sphericality eigenvalue further includes:

[0033] If the linear feature value of a point is less than or equal to the fourth set threshold, the surface feature value of the point is less than or equal to the fifth set threshold, and the spherical feature value of the point is greater than the sixth set threshold, it is determined that the point cloud has a spherical feature.

[0034] In this embodiment, the linear feature value, planar feature value, spherical feature value and several preset thresholds of a certain point are used to more accurately determine whether the point has a spherical feature.

[0035] In a possible implementation, determining the position and posture feature information of the radar when collecting each frame of point cloud data based on the coordinate information of each point in each registration point pair and the attribute characteristics of each registration point pair includes:

[0036] Based on the coordinate information of the points in each registration point pair and the attribute characteristics of each registration point pair, the error metric information of each registration point pair is determined respectively;

[0037] Based on the error measurement information of each registration point pair, determine the pose transformation matrix between each two adjacent frames of point cloud data;

[0038] Based on the pose conversion matrix between each two adjacent frames of point cloud data, the pose feature information of the radar when collecting each frame of point cloud data is determined.

[0039] In this embodiment, by using the attribute features corresponding to each pair of registration points, more accurate error measurement information can be determined, and then, based on the more accurate error measurement information, the accuracy of the determined posture feature information can be improved.

[0040] In a possible implementation manner, the attribute feature includes features of whether it is a ground point or a non-ground point, the feature of the non-ground point includes a shape feature, and the shape feature includes a spherical feature; the error metric information includes Euclidean distance error information;

[0041] The step of determining the error metric information of each registration point pair based on the coordinate information of the points in each registration point pair and the attribute characteristics of each registration point pair includes:

[0042] For a registration point pair having a non-ground spherical feature, the Euclidean distance error information of the registration point pair is determined based on the coordinate information of each point in the registration point pair.

[0043] In a possible implementation manner, the attribute feature includes whether it is a ground point or a non-ground point feature, the non-ground point feature includes a shape feature, and the shape feature includes a linear feature; the error metric information includes point-line distance error information;

[0044] The method of determining the error metric information of each registration point pair based on the coordinate information of the points in each registration point pair and the attribute characteristics of each registration point pair also includes:

[0045] For a registration point pair with non-ground linear features, point-line distance error information of the registration point pair is determined based on the coordinate information of each point in the registration point pair and the vector of the straight line where one of the points is located.

[0046] In a possible implementation manner, the attribute feature includes whether it is a ground point or a non-ground point feature, the non-ground point feature includes a shape feature, and the shape feature includes a surface feature; the error metric information includes point-to-surface distance error information;

[0047] The method of determining the error metric information of each registration point pair based on the coordinate information of the points in each registration point pair and the attribute characteristics of each registration point pair also includes:

[0048] For a pair of registration points with non-ground planar features or a pair of registration points with ground features, the point-to-surface distance error information of the pair of registration points is determined based on the coordinate information of each point in the pair of registration points and the normal vector of the surface where the point in the pair of registration points belongs to the next frame of point cloud data.

[0049] The above implementation method determines different error measurement information for registration point pairs with different attribute characteristics, thereby improving the accuracy of the determined error measurement information, which is beneficial to improving the accuracy of the determined posture feature information.

[0050] In a possible implementation, based on the pose conversion matrix between each two adjacent frames of point cloud data, determining the pose feature information of the radar when collecting each frame of point cloud data includes:

[0051] Based on the pose conversion matrix between each two adjacent frames of point cloud data, the pose transformation constraint information of the radar when collecting each two adjacent frames of point cloud data is determined;

[0052] Pre-integrate multiple measurement pose information of the radar when the radar collects each two adjacent frames of point cloud data, and determine pre-integration constraint information of the radar when collecting each two adjacent frames of point cloud data;

[0053] Based on the pose transformation constraint information and pre-integration constraint information, the pose feature information of the radar when collecting each frame of point cloud data is determined.

[0054] This embodiment utilizes the posture transformation constraint information determined based on point cloud registration and the pre-integration constraint information determined based on the measured posture, so as to improve the accuracy of the determined posture feature information.

[0055] In a possible implementation, before determining the pose feature information of the radar when collecting each frame of point cloud data based on the pose transformation constraint information and the pre-integration constraint information, the method further includes:

[0056] Determining a first weight corresponding to the posture transformation constraint information based on a covariance corresponding to the posture transformation matrix;

[0057] Pre-integrate multiple measurement pose information of the radar when the radar collects each two adjacent frames of point cloud data to obtain a second weight corresponding to the pre-integration constraint information;

[0058] The determining of the posture feature information of the radar when collecting each frame of point cloud data based on the posture transformation constraint information and the pre-integration constraint information includes:

[0059] Based on the posture transformation constraint information, the pre-integration constraint information, the first weight and the second weight, the posture feature information of the radar when collecting each frame of point cloud data is determined.

[0060] This implementation, by combining the weights determined by the covariance corresponding to the posture conversion matrix and the weights determined based on the measured posture, can improve the accuracy of the determined posture feature information.

[0061] In a possible implementation manner, before determining the position and posture feature information of the radar when collecting each frame of point cloud data, the method further includes:

[0062] Based on the global positioning information of the radar when the radar collects each frame of point cloud data, the estimated posture information of the radar when collecting each frame of point cloud data is determined;

[0063] Determine the global positioning transformation constraint information based on the estimated pose information of the radar when collecting each two adjacent frames of point cloud data;

[0064] The determining of the posture feature information of the radar when collecting each frame of point cloud data based on the posture transformation constraint information and the pre-integration constraint information includes:

[0065] The method determines the posture feature information of the radar when collecting each frame of point cloud data based on the posture transformation constraint information, the pre-integration constraint information and the global positioning transformation constraint information.

[0066] This implementation, combined with the global positioning transformation constraint information determined based on the global positioning information, can improve the accuracy of the determined posture feature information.

[0067] In a possible implementation, the above-mentioned posture determination method further includes:

[0068] Based on the position and posture feature information of the radar when collecting continuous multi-frame point cloud data, the movement mileage information of the radar is determined.

[0069] This embodiment utilizes multiple accurate position and posture feature information of the radar to accurately determine the movement mileage information of the radar.

[0070] In a second aspect, the present disclosure provides a posture determination device, comprising:

[0071] A data acquisition module, used to acquire at least two frames of point cloud data collected by the radar;

[0072] An attribute determination module is used to determine the attribute characteristics of each point in each frame of point cloud data;

[0073] a point pair registration module, for matching points in each frame of point cloud data except the first frame of point cloud data in the at least two frames of point cloud data with points in a previous frame of point cloud data that have the same attribute features, to obtain a plurality of registration point pairs;

[0074] The posture determination module is used to determine the posture feature information of the radar when collecting each frame of point cloud data based on the coordinate information of each point in each registration point pair and the attribute characteristics of each registration point pair.

[0075] In a third aspect, the present disclosure provides an electronic device comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the above-mentioned posture determination method are performed.

[0076] In a fourth aspect, the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned posture determination method are executed.

[0077] The above-mentioned devices, electronic devices, and computer-readable storage media disclosed in the present invention at least contain technical features that are substantially the same as or similar to the technical features of any aspect of the above-mentioned method disclosed in the present invention or any implementation method of any aspect. Therefore, for the description of the effects of the above-mentioned devices, electronic devices, and computer-readable storage media, please refer to the description of the effects of the above-mentioned method content, which will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present disclosure and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0079] Figure 1 A flow chart of a method for determining a posture provided by an embodiment of the present disclosure is shown;

[0080] Figure 2 A flowchart of determining whether each point is a ground point in another method for determining a position and posture provided by an embodiment of the present disclosure is shown;

[0081] Figure 3 A flowchart of determining attribute features of a point in another method for determining a position and posture provided by an embodiment of the present disclosure is shown;

[0082] Figure 4 A flowchart of determining the posture feature information of a radar when collecting each frame of point cloud data in another posture determination method provided by an embodiment of the present disclosure is shown;

[0083] Figure 5 A schematic diagram of the structure of a posture determination device provided by an embodiment of the present disclosure is shown;

[0084] Figure 6 A schematic structural diagram of an electronic device provided by an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0085] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. It should be understood that the drawings in the present disclosure are only for the purpose of illustration and description and are not used to limit the scope of protection of the present disclosure. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowcharts used in the present disclosure illustrate the operations implemented according to some embodiments of the present disclosure. It should be understood that the operations of the flowcharts can be implemented out of sequence, and steps without logical contextual relationships can be reversed in order or implemented simultaneously. In addition, those skilled in the art, under the guidance of the contents of the present disclosure, can add one or more other operations to the flowchart, or remove one or more operations from the flowchart.

[0086] In addition, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The components of the embodiments of the present disclosure generally described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the drawings is not intended to limit the scope of the present disclosure claimed for protection, but merely represents the selected embodiments of the present disclosure. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present disclosure.

[0087] It should be noted that the term “comprising” will be used in the embodiments of the present disclosure to indicate the existence of the features declared thereafter, but does not exclude the addition of other features.

[0088] The present disclosure provides a posture determination method and device, an electronic device, and a computer-readable storage medium, wherein the present disclosure uses points with the same attribute characteristics for pairing, which can improve the registration success rate and accuracy of point cloud data, and then uses more accurate registration point pairs to improve the accuracy of the determined posture feature information. Subsequently, using accurate posture feature information, a more accurate map can be generated.

[0089] The posture determination method and device, electronic device, and computer-readable storage medium disclosed in the present invention are described below through specific embodiments.

[0090] The present disclosure provides a method for determining a posture, which is applied to a terminal device or a server for determining a posture of a radar. Specifically, Figure 1 As shown, the posture determination method provided by the embodiment of the present disclosure includes the following steps:

[0091] S110: Acquire at least two frames of point cloud data collected by the radar.

[0092] The at least two frames of point cloud data may be point cloud data of multiple adjacent frames collected by the radar.

[0093] The radar transmits scan lines in batches according to a preset frequency, with multiple scan lines in each batch.

[0094] Specifically, taking a laser radar that emits 64 scan lines in each batch as an example, 64 laser emitters are installed along the longitudinal direction. After the laser radar is started, the laser emitter starts to rotate. During the rotation process, each laser emitter emits and receives lasers at a certain preset frequency to collect point cloud data. Therefore, the number of scan points in a frame of point cloud data is 64*fires, where fires represents the number of batches of scan lines emitted by a laser emitter in one rotation.

[0095] For a rotating laser radar, it rotates one circle, during which it emits multiple batches of scan lines to obtain a frame of point cloud data. This frame of point cloud data is a circle of arcs. The farther the distance from the laser radar, the lower the density of the point cloud data. Point cloud data with low density has little effect on the subsequent determination of registration point pairs. Therefore, after collecting one or more frames of point cloud data, and before determining the attribute characteristics of the points in the point cloud data, the points that are farther away from the radar and have lower density can be eliminated. In specific implementation, points in each frame of point cloud data whose distance from the radar is greater than a first set threshold can be eliminated. The above-mentioned first set threshold is flexibly set based on the specific application scenario, for example, it can be set to 40 meters.

[0096] In addition, after collecting multiple frames of point cloud data, since the amount of point cloud data is large, in order to improve the efficiency of determining the registration point pairs, downsampling processing can be performed based on the reflectivity of the points in the point cloud data, that is, downsampling processing is performed on the points in each frame of point cloud data whose reflectivity is less than the second set threshold. Specifically, points whose reflectivity is greater than the second set threshold are retained, and points whose reflectivity is less than the second set threshold are randomly downsampled.

[0097] The above-mentioned elimination of points that are far away from the radar and have low reflectivity is conducive to improving the accuracy of the determined attribute features and posture feature information.

[0098] S120: Determine the attribute characteristics of each point in each frame of point cloud data.

[0099] Here, the attribute feature may specifically include whether it is a ground point, and the feature of a non-ground point includes a shape feature, and the shape feature is any one of a linear feature, a planar feature, and a spherical feature.

[0100] When determining whether a point is a ground point, the preset segmentation area to which each point in the frame belongs can be determined based on the coordinate information of each point; then, for each preset segmentation area, if the distance between the points in other preset segmentation areas adjacent to the preset segmentation area and the ground is smaller than the shortest distance between the points in the preset segmentation area and the ground, it is determined that the point in the preset segmentation area is not a ground point.

[0101] For a point having features that are not ground points, multiple points that are close to the point may be acquired, and then the shape features of the point may be determined based on the acquired multiple points.

[0102] S130. For each frame of point cloud data except the first frame of point cloud data in the at least two frames of point cloud data, match the points in the frame of point cloud data with the points in the previous frame of point cloud data that have the same attribute features to obtain multiple registration point pairs.

[0103] Here, points with the same attribute characteristics are paired, for example, points with the characteristics of ground data in two frames of point cloud data are paired, points with linear characteristics in two frames of point cloud data are paired, and points with surface characteristics in two frames of point cloud data are paired.

[0104] When pairing points with the same attribute characteristics, the points with a certain attribute characteristic in the two frames of point cloud data are respectively taken as a set. For each point in one of the sets, the point closest to the point is selected from the other set, and the point and the point closest to the point obtained by the selection are taken as a registration point pair.

[0105] S140. Based on the coordinate information of each point in each registration point pair and the attribute characteristics of each registration point pair, determine the posture characteristic information of the radar when collecting each frame of point cloud data.

[0106] Here, specifically, based on the coordinate information of each registration point pair and the attribute characteristics of each registration point pair, the error measurement information of each registration point pair can be determined respectively; based on the error measurement information of each registration point pair, the pose transformation matrix between each two adjacent frames of point cloud data can be determined; finally, based on the pose transformation matrix between each two adjacent frames of point cloud data, the pose feature information of the radar when collecting each frame of point cloud data can be determined.

[0107] In this step, the attribute features of each registration point pair can be used to determine more accurate error measurement information, and then, based on the more accurate error measurement information, the accuracy of the determined posture feature information can be improved.

[0108] The disclosed embodiment utilizes points with the same attribute characteristics in point cloud data for pairing, which can improve the registration success rate and accuracy of point cloud data. Subsequently, by utilizing more accurate registration point pairs, the accuracy of determined posture feature information can be improved. Subsequently, by utilizing accurate posture feature information, a more accurate map can be generated.

[0109] In some embodiments, after collecting one or more frames of point cloud data, and before determining the attribute features of each point in each frame of point cloud data, the coordinate information of the points collected by non-first batch scanning lines in each frame of point cloud data needs to be uniformly transformed into the coordinate system of the radar corresponding to the emission moment of the first batch of scanning lines in the frame of point cloud data. In this way, using points in the same coordinate system is conducive to improving the accuracy of the determined attribute features and posture feature information.

[0110] In specific implementation, the coordinate conversion can be performed using the following steps:

[0111] Step 1: obtaining multiple measurement pose information of the radar during the time period when the radar collects each frame of point cloud data;

[0112] Each frame of the point cloud data includes point cloud data collected by the radar by transmitting multiple batches of scanning lines. Since the radar takes a long time to collect each frame of point cloud data, the radar's posture may have changed many times during this period. Therefore, here, the inertial measurement unit IMU can be used to collect the radar's posture at different times at a certain frequency to obtain multiple measurement posture information.

[0113] Step 2: Based on the multiple measured posture information, determine the posture information of the radar when transmitting each batch of scan lines.

[0114] The radar transmits scan lines at its own frequency, and the IMU measures the radar's posture information at its own frequency. Therefore, the measured posture information measured by the IMU may not be the posture information when the radar transmits the scan lines. Therefore, the measured posture information measured by the IMU cannot be directly used to convert the coordinates of the point cloud data collected by non-first-batch scan lines into the coordinate system of the radar corresponding to the time when the first batch of scan lines were emitted.

[0115] In response to the above problem, this step uses multiple measurement pose information collected by the IMU to determine the pose information of the radar when it launches each batch of scan lines. Specifically, the pose information of the radar when launching a batch of scan lines can be determined using the following steps: first, the time when the radar launches the batch of scan lines and the measurement time of each measurement pose information can be obtained. Then, based on the above time, the measurement pose information before the radar launches the batch of scan lines and the measurement pose information after the radar launches the batch of scan lines are screened. Finally, the two measurement pose information are interpolated to obtain the pose information when the radar launches the batch of scan lines.

[0116] Step 3: For the point cloud data collected by non-first batch of scanning lines in each frame of point cloud data, based on the posture information of the radar when emitting the batch of scanning lines, the coordinates of the point cloud data collected by the batch of scanning lines are converted to the coordinate system of the radar corresponding to the emission moment of the first batch of scanning lines in the frame of point cloud data.

[0117] According to the above description, the attribute characteristics of a point include whether it is a ground point. In specific implementation, such as Figure 2 As shown, the following steps can be used to determine whether each point is a ground point:

[0118] S210 . Determine, according to coordinate information of each point, a preset segmentation area to which each point in the frame belongs.

[0119] Here, the frame point cloud data may be specifically two-dimensionally rasterized on the XOY plane to obtain a plurality of preset segmentation areas.

[0120] S220: For each preset segmentation area, if the distances between points of other preset segmentation areas adjacent to the preset segmentation area and the ground are all smaller than the shortest distances between each point in the preset segmentation area and the ground, it is determined that each point in the preset segmentation area is not a ground point.

[0121] Before executing this step, the shortest distance between each point in each preset segmentation area and the ground needs to be calculated, that is, the minimum Z value of each point in each preset segmentation area is calculated. Then, based on the determined shortest distance, it is determined whether the point is a ground point.

[0122] The above steps can only determine the preset segmentation area that does not include points with the characteristics of ground points, and a preset segmentation area may include points with the characteristics of ground points and points with the characteristics of non-ground points at the same time. In specific implementation, the following steps can be used to determine whether a point in a preset segmentation area is a ground point:

[0123] Determine the target point within the preset segmentation area that is closest to the ground; determine that the point within the preset segmentation area whose distance to the target point is greater than a third set threshold is not a ground point, and the point within the preset segmentation area whose distance to the target point is less than or equal to the third set threshold is a ground point.

[0124] This embodiment utilizes the distance between the points in each preset segmentation area and the ground to more accurately determine whether the corresponding point is a ground point, that is, it can more accurately determine the attribute characteristics of the corresponding point, which is beneficial to improve the accuracy of the determined posture feature information based on more accurate attribute characteristics.

[0125] In the case where the attribute feature of any point includes the feature of a non-ground point, the feature of the non-ground point includes a shape feature, and the shape feature is any one of a linear feature, a planar feature, and a spherical feature.

[0126] In specific implementation, Figure 3 As shown in the figure, for a point with non-ground point characteristics, the attribute characteristics of the point can be determined using the following steps:

[0127] S310. For any point that is not a ground point, obtain multiple points whose distances to the any point are within a preset distance range.

[0128] S320: Determine a three-dimensional feature value of any point based on the multiple points obtained.

[0129] Here, the principal component analysis method PCA may be specifically used to determine the three-dimensional eigenvalue and three-dimensional eigenvector of any point.

[0130] The three eigenvalues ​​are v1, v2, v3, and the three eigenvectors are e1, e2, e3.

[0131] S330. Based on the three-dimensional eigenvalues, determine the linear eigenvalue, planar eigenvalue, and spherical eigenvalue corresponding to any point.

[0132] In specific implementation, the linearity characteristic value linearty can be determined using the following formula:

[0133] linearty=(v1-v2) / v1 (1)

[0134] The planarity characteristic value can be determined using the following formula:

[0135] planarity=(v2-v3) / v1 (2)

[0136] The spherical characteristic value curvature can be determined using the following formula:

[0137] curvature=v3 / (v1+v2+v3) (3)

[0138] S340: Determine the shape feature corresponding to any point based on the linear feature value, the planar feature value and the spherical feature value.

[0139] Specifically, if the linear feature value is greater than the fourth set threshold, it is determined that the corresponding point has a linear feature; if the linear feature value is less than or equal to the fourth set threshold, and the surface feature value is greater than the fifth set threshold, it is determined that the corresponding point has a surface feature; if the linear feature value is less than or equal to the fourth set threshold, the surface feature value is less than or equal to the fifth set threshold, and the spherical feature value is greater than the sixth set threshold, it is determined that the corresponding point has a spherical feature.

[0140] If, based on the above judgment, it is determined that the shape feature of any point is not a linear feature, a planar feature or a spherical feature, then the shape feature of the cloud is determined to be a spherical feature.

[0141] The disclosed embodiment can improve the accuracy of the determined attribute features by utilizing the linear feature value, the planar feature value, the spherical feature value and several preset thresholds of a certain point.

[0142] In some embodiments, the error metric information includes Euclidean distance error information (or called point-to-point distance error), point-to-line distance error information, and point-to-plane distance error information. Based on the coordinate information of the points in each registration point pair and the attribute features corresponding to each registration point pair, the error metric information corresponding to each registration point pair is determined respectively, and is specifically implemented as follows:

[0143] Method 1: For a registration point pair having a non-ground spherical feature, based on the coordinate information of each point in the registration point pair, determine the Euclidean distance error information corresponding to the registration point pair.

[0144] In specific implementation, the above-mentioned Euclidean distance error information can be determined using the following formula:

[0145] {t,r} * =argmin(∑||Rp i +tq i ||2) (4)

[0146] In the formula, p i Represents the coordinate information of a point in the registration point pair, q i Represents the coordinate information of another point in the registration point pair, R represents the rotation matrix from one point to another point in the registration point pair, and t represents the translation vector from one point to another point in the registration point pair.

[0147] Method 2: For a registration point pair with non-ground linear features, based on the coordinate information of each point in the registration point pair and the vector of the straight line where one of the points is located, determine the point-line distance error information of the registration point pair.

[0148] In specific implementation, the above point-line distance error information can be determined using the following formula:

[0149] {t,r} * =argmin(∑(Rp i +tq i )×v line ) (5)

[0150] In the formula, v line The vector representing the straight line where one of the points in the registration point pair lies can be the eigenvector corresponding to the largest eigenvalue among the three-dimensional eigenvalues ​​corresponding to one of the points in the registration point pair. In specific implementation, the principal component analysis method PCA can be used to determine the three-dimensional eigenvalue of any point and the eigenvector corresponding to each three-dimensional eigenvalue.

[0151] Method three: for a pair of registration points with non-ground planar features or a pair of registration points with ground features, the point-to-surface distance error information of the pair of registration points is determined based on the coordinate information of each point in the pair of registration points and the normal vector of the surface where the point in the pair of registration points belongs to the next frame of point cloud data.

[0152] In specific implementation, the above point-to-surface distance error information can be determined using the following formula:

[0153] {t,r} * =argmin(∑(Rp i +tq i )·n q ) (6)

[0154] Where n q The normal vector representing the surface where the point in the registration point pair belonging to the next frame of point cloud data is located may be a eigenvector corresponding to the smallest eigenvalue among the three-dimensional eigenvalues ​​corresponding to the point in the registration point pair belonging to the next frame of point cloud data.

[0155] After determining the error information corresponding to each pair of registration points, the error information corresponding to all pairs of registration points is summed up to obtain the optimization objective function, and then the posture transformation matrix is ​​solved to minimize the optimization objective function.

[0156] When solving the pose transformation matrix, the method of approximate linear solution can be used. Since the nonlinearity of pose transformation comes from rotation, the rotation is linearized. Assuming that the rotation change is very small, because the radian of the angle at a small angle is approximately equal to its sine value, and the cosine value of the angle is approximately equal to 1, the nonlinear trigonometric function terms can be eliminated by substituting the optimization objective function into the rotation matrix of the Euler angle transformation. The original nonlinear optimization problem is also transformed into the least squares problem of a linear system of equations, and the pose transformation matrix can be directly obtained by pseudo-inverse solution.

[0157] The above embodiment performs point cloud registration to obtain multiple registration point pairs, and determines the optimization objective function based on the registration point pairs, solves the posture transformation matrix that minimizes the value of the optimization objective function, repeats the above steps until the iteration converges, and obtains the final posture transformation matrix.

[0158] The above-mentioned embodiment determines different error metric information for registration point pairs with different attribute characteristics, thereby improving the accuracy of the determined error metric information, thereby facilitating improving the accuracy of the determined posture feature information.

[0159] In some embodiments, Figure 4 As shown, the above-mentioned posture conversion matrix between each two adjacent frames of point cloud data is used to determine the posture feature information of the radar when collecting each frame of point cloud data. Specifically, it can be implemented by the following steps:

[0160] S410, based on the pose conversion matrix between each two adjacent frames of point cloud data, determining the pose transformation constraint information of the radar when collecting each two adjacent frames of point cloud data.

[0161] Here, the pose transformation constraint information is used to represent the constraint information of the radar pose transformation between adjacent frames.

[0162] S420, pre-integrate the multiple measurement posture information of the radar when the radar collects each two adjacent frames of point cloud data, and determine the pre-integration constraint information of the radar when the radar collects each two adjacent frames of point cloud data.

[0163] Here, the pre-integration constraint information can not only represent the constraint information of the radar posture transformation between adjacent frames, but also represent the constraint information of the radar speed change between adjacent frames.

[0164] S430, based on the posture transformation constraint information and the pre-integration constraint information, determine the posture feature information of the radar when collecting each frame of point cloud data.

[0165] When determining the pose feature information, the pose transformation constraint information and pre-integration constraint information can be used to construct a pose graph. The pose graph uses the pose feature information of each frame of point cloud data as an optimization variable, i.e., a node, and connects the pose transformation constraint information and pre-integration constraint information to the corresponding optimization variables in the form of edges. After the pose graph is constructed, the pose feature information of each frame of point cloud data can be obtained by solving the pose graph.

[0166] When constructing the pose graph, you can also use the following steps to add global positioning transformation constraint information to improve the accuracy of the determined pose feature information:

[0167] Based on the global positioning information of the radar when collecting each frame of point cloud data, the estimated posture information of the radar when collecting each frame of point cloud data is determined; based on the estimated posture information of the radar when collecting each two adjacent frames of point cloud data, the global positioning transformation constraint information is determined.

[0168] After determining the global positioning transformation constraint information, the pose feature information of the radar when collecting each frame of point cloud data is determined based on the pose transformation constraint information, the pre-integration constraint information and the global positioning transformation constraint information. Specifically, when constructing the pose graph, the global positioning transformation constraint information can be used as an edge to connect the corresponding optimization variables. After that, solving the pose graph can obtain the pose feature information of each frame of point cloud data.

[0169] In specific implementation, since the speed of determining the estimated pose information is slow, the estimated pose information of the radar when collecting each frame of point cloud data can be determined by interpolation.

[0170] After constructing the pose graph, the edges in the pose graph can also be weighted, where the weights of the edges corresponding to the pose transformation constraint information can be determined based on the covariance corresponding to the pose transformation matrix. The weights of the edges corresponding to the pre-integration constraint information can be obtained by pre-integrating the multiple measured pose information of the radar when the radar collects each two adjacent frames of point cloud data. The weights of the edges corresponding to the global positioning transformation constraint information can be pre-set.

[0171] After weighting the edges in the pose graph, the pose graph is solved to obtain the pose feature information of each frame of point cloud data.

[0172] The disclosed embodiments, by combining the weights determined by the covariance corresponding to the posture conversion matrix and the weights determined based on the measured posture, can improve the accuracy of the determined posture feature information.

[0173] After obtaining the position and posture feature information of each frame of point cloud data, the movement mileage information of the radar can be determined based on the position and posture feature information of the radar when collecting multiple frames of continuous point cloud data.

[0174] By utilizing multiple accurate position and posture feature information of the radar, the movement mileage information of the radar can be accurately determined.

[0175] Corresponding to the above-mentioned posture determination method, the present disclosure also provides a posture determination device, which is applied to a terminal device or server for determining the posture feature information of the radar, and each module can implement the same method steps as in the above-mentioned method and achieve the same beneficial effects. Therefore, for the same parts, the present disclosure will not repeat them.

[0176] like Figure 5As shown, the present disclosure provides a posture determination device, comprising:

[0177] The data acquisition module 510 is used to acquire at least two frames of point cloud data collected by the radar;

[0178] The attribute determination module 520 is used to determine the attribute characteristics of each point cloud data in each frame of point cloud data;

[0179] A point pair registration module 530 is used to match points in each frame of point cloud data except the first frame of point cloud data in the at least two frames of point cloud data with points in a previous frame of point cloud data that have the same attribute features, so as to obtain a plurality of registration point pairs;

[0180] The posture determination module 540 is used to determine the posture feature information of the radar when collecting each frame of point cloud data based on the coordinate information of each point in each registration point pair and the attribute characteristics of each registration point pair.

[0181] In some embodiments, before determining the attribute features of each point in each frame of point cloud data, the attribute determination module 520 is further configured to:

[0182] Eliminate points in each frame of point cloud data whose distance from the radar is greater than a first set threshold;

[0183] The points in each frame of point cloud data whose reflectivity is less than the second set threshold are downsampled.

[0184] In some embodiments, the attribute feature includes whether it is a ground point;

[0185] When determining the attribute features of each point in each frame of point cloud data, the attribute determination module 520 is used to:

[0186] Determine the preset segmentation area to which each point in the frame belongs according to the coordinate information of each point;

[0187] For each preset segmentation area, if the distances between points in other preset segmentation areas adjacent to the preset segmentation area and the ground are all smaller than the shortest distance between the point in the preset segmentation area and the ground, it is determined that the point in the preset segmentation area is not a ground point.

[0188] In some embodiments, when the attribute feature of any point includes a feature of a non-ground point, the feature of the non-ground point includes a shape feature, and the shape feature is any one of a linear feature, a planar feature, and a spherical feature;

[0189] When determining the attribute features of each point in each frame of point cloud data, the attribute determination module 520 is further used to:

[0190] For any point that is not a ground point, a plurality of points whose distances from the any point are within a preset distance range are obtained; and based on the obtained plurality of points, a three-dimensional feature value of the any point is determined;

[0191] Based on the three-dimensional eigenvalues, determining the linear eigenvalue, the planar eigenvalue and the spherical eigenvalue corresponding to the any point;

[0192] Based on the linearity eigenvalues, the planarity eigenvalues ​​and the sphericality eigenvalues, a shape feature corresponding to any point cloud is determined.

[0193] In some embodiments, when determining the shape feature corresponding to any point based on the linearity feature value, the planarity feature value and the sphericality feature value, the attribute determination module 520 is used to:

[0194] If the linear feature value of a point is greater than the fourth set threshold, it is determined that the point has a linear feature.

[0195] In some embodiments, when determining the shape feature corresponding to any point cloud based on the linearity feature value, the surface feature value and the sphericity feature value, the attribute determination module 520 is further configured to:

[0196] If the linear feature value of a point is less than or equal to the fourth set threshold, and the surface feature value of the point is greater than the fifth set threshold, it is determined that the point has the surface feature.

[0197] In some embodiments, when determining the shape feature corresponding to any point cloud based on the linearity feature value, the surface feature value and the sphericity feature value, the attribute determination module 520 is further configured to:

[0198] If the linear feature value of a point is less than or equal to the fourth set threshold, the surface feature value of the point is less than or equal to the fifth set threshold, and the spherical feature value of the point is greater than the sixth set threshold, it is determined that the point cloud has a spherical feature.

[0199] In some embodiments, the posture determination module 540, when determining the posture feature information of the radar when collecting each frame of point cloud data based on the coordinate information of each point in each registration point pair and the attribute characteristics of each registration point pair, is used to:

[0200] Based on the coordinate information of the points in each registration point pair and the attribute characteristics of each registration point pair, the error metric information of each registration point pair is determined respectively;

[0201] Based on the error measurement information of each registration point pair, determine the pose transformation matrix between each two adjacent frames of point cloud data;

[0202] Based on the pose conversion matrix between each two adjacent frames of point cloud data, the pose feature information of the radar when collecting each frame of point cloud data is determined.

[0203] In some embodiments, the attribute feature includes whether it is a ground point or a non-ground point, the feature of the non-ground point includes a shape feature, and the shape feature includes a spherical feature; the error metric information includes Euclidean distance error information;

[0204] The pose determination module 540 is used to determine the error metric information of each registration point pair based on the coordinate information of the points in each registration point pair and the attribute characteristics of each registration point pair:

[0205] For a registration point pair having a non-ground spherical feature, the Euclidean distance error information of the registration point pair is determined based on the coordinate information of each point in the registration point pair.

[0206] In some embodiments, the attribute feature includes whether it is a ground point or a non-ground point, the feature of the non-ground point includes a shape feature, and the shape feature includes a linear feature; the error metric information includes point-line distance error information;

[0207] When determining the error metric information of each registration point pair based on the coordinate information of the points in each registration point pair and the attribute characteristics of each registration point pair, the posture determination module 540 is further used to:

[0208] For a registration point pair with non-ground linear features, point-line distance error information of the registration point pair is determined based on the coordinate information of each point in the registration point pair and the vector of the straight line where one of the points is located.

[0209] In some embodiments, the attribute feature includes whether it is a ground point or a non-ground point, the feature of the non-ground point includes a shape feature, and the shape feature includes a surface feature; the error metric information includes point-to-surface distance error information;

[0210] When determining the error metric information of each registration point pair based on the coordinate information of the points in each registration point pair and the attribute characteristics of each registration point pair, the posture determination module 540 is further used to:

[0211] For a pair of registration points with non-ground planar features or a pair of registration points with ground features, the point-to-surface distance error information of the pair of registration points is determined based on the coordinate information of each point in the pair of registration points and the normal vector of the surface where the point in the pair of registration points belongs to the next frame of point cloud data.

[0212] In some embodiments, the posture determination module 540, when determining the posture feature information of the radar when collecting each frame of point cloud data based on the posture conversion matrix between each two adjacent frames of point cloud data, is used to:

[0213] Based on the pose conversion matrix between each two adjacent frames of point cloud data, the pose transformation constraint information of the radar when collecting each two adjacent frames of point cloud data is determined;

[0214] Pre-integrate multiple measurement pose information of the radar when the radar collects each two adjacent frames of point cloud data, and determine pre-integration constraint information of the radar when collecting each two adjacent frames of point cloud data;

[0215] Based on the pose transformation constraint information and pre-integration constraint information, the pose feature information of the radar when collecting each frame of point cloud data is determined.

[0216] In some embodiments, before determining the pose feature information of the radar when collecting each frame of point cloud data based on the pose transformation constraint information and the pre-integration constraint information, the pose determination module 540 is further used to:

[0217] Determining a first weight corresponding to the posture transformation constraint information based on a covariance corresponding to the posture transformation matrix;

[0218] Pre-integrate multiple measurement pose information of the radar when the radar collects each two adjacent frames of point cloud data to obtain a second weight corresponding to the pre-integration constraint information;

[0219] The posture determination module 540 is used to determine the posture feature information of the radar when collecting each frame of point cloud data based on the posture transformation constraint information and the pre-integration constraint information:

[0220] Based on the posture transformation constraint information, the pre-integration constraint information, the first weight and the second weight, the posture feature information of the radar when collecting each frame of point cloud data is determined.

[0221] In some embodiments, before determining the pose feature information of the radar when collecting each frame of point cloud data, the pose determination module 540 is further used to:

[0222] Based on the global positioning information of the radar when the radar collects each frame of point cloud data, the estimated posture information of the radar when collecting each frame of point cloud data is determined;

[0223] Determine the global positioning transformation constraint information based on the estimated pose information of the radar when collecting each two adjacent frames of point cloud data;

[0224] The posture determination module 540 is used to determine the posture feature information of the radar when collecting each frame of point cloud data based on the posture transformation constraint information and the pre-integration constraint information:

[0225] The method determines the posture feature information of the radar when collecting each frame of point cloud data based on the posture transformation constraint information, the pre-integration constraint information and the global positioning transformation constraint information.

[0226] In some embodiments, a mileage determination module 550 is also included, which is used to determine the movement mileage information of the radar based on the posture feature information of the radar when collecting multiple frames of continuous point cloud data.

[0227] The present disclosure discloses an electronic device, such as Figure 6 As shown, it includes: a processor 601, a memory 602 and a bus 603. The memory 602 stores machine-readable instructions executable by the processor 601. When the electronic device is running, the processor 601 communicates with the memory 602 through the bus 603.

[0228] When the machine-readable instructions are executed by the processor 601, the following steps of the posture determination method are performed:

[0229] Obtain at least two frames of point cloud data collected by the radar;

[0230] Determine the attribute characteristics of each point in each frame of point cloud data;

[0231] For each frame of point cloud data except the first frame of point cloud data in the at least two frames of point cloud data, matching points in the frame of point cloud data with points in a previous frame of point cloud data having the same attribute features to obtain a plurality of registration point pairs;

[0232] Based on the coordinate information of each point in each registration point pair and the attribute characteristics of each registration point pair, the position and posture characteristic information of the radar when collecting each frame of point cloud data is determined.

[0233] In addition, when the machine-readable instructions are executed by the processor 601, the method content in any implementation described in the above method part can also be executed, which will not be repeated here.

[0234] The embodiments of the present disclosure also provide a computer program product corresponding to the above-mentioned method and apparatus, including a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the method in the previous method embodiment. The specific implementation can be found in the method embodiment, which will not be repeated here.

[0235] The above description of various embodiments tends to emphasize the differences between the various embodiments. The same or similar aspects can be referenced to each other, and for the sake of brevity, they will not be repeated herein.

[0236] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system and device described above can refer to the corresponding process in the method embodiment, and will not be repeated in this disclosure. In the several embodiments provided in the present disclosure, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0237] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0238] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0239] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.

[0240] The above are only specific embodiments of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present disclosure, which should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A method for determining a posture, characterized in that: include: Acquire at least two frames of point cloud data collected by the radar, multiple measurement pose information of each frame of point cloud data collected by the radar, the measurement time of each measurement pose information, and the time when the radar transmits each batch of scan lines; For each batch of scan lines transmitted by the radar, based on the time when the radar transmits the batch of scan lines and the measurement time of each measurement pose information, the measurement pose information before the radar transmits the batch of scan lines and the measurement pose information after the radar transmits the batch of scan lines are filtered out from multiple measurement pose information, and interpolation is performed based on the filtered two pieces of measurement pose information to obtain the pose information when the radar transmits the batch of scan lines; For the point cloud data collected by non-first batch scanning lines in each frame of point cloud data, based on the posture information of the radar when the batch of scanning lines is emitted, the coordinates of the point cloud data collected by the batch of scanning lines are converted to the coordinate system of the radar corresponding to the emission time of the first batch of scanning lines in the frame of point cloud data; Determine the attribute characteristics of each point in each frame of point cloud data; For each frame of point cloud data except the first frame of point cloud data in the at least two frames of point cloud data, matching points in the frame of point cloud data with points in a previous frame of point cloud data having the same attribute features to obtain a plurality of registration point pairs; Based on the coordinate information of each point in each registration point pair and the attribute characteristics of each registration point pair, the position and posture characteristic information of the radar when collecting each frame of point cloud data is determined.

2. The method for determining a posture according to claim 1, characterized in that: Before determining the attribute characteristics of each point in each frame of point cloud data, it also includes: Eliminate points in each frame of point cloud data whose distance from the radar is greater than a first set threshold; The points in each frame of point cloud data whose reflectivity is less than the second set threshold are downsampled.

3. The method for determining a posture according to claim 1 or 2, characterized in that: The attribute features include whether it is a ground point; Determining the attribute features of each point in each frame of point cloud data includes: According to the coordinate information of each point, determine the preset segmentation area to which each point in a frame belongs; For each preset segmentation area, if the distances between points in other preset segmentation areas adjacent to the preset segmentation area and the ground are all smaller than the shortest distance between the point in the preset segmentation area and the ground, it is determined that the point in the preset segmentation area is not a ground point.

4. The method for determining a posture according to claim 1 or 2, characterized in that: In the case where the attribute feature of any point includes a feature of a non-ground point, the feature of the non-ground point includes a shape feature, and the shape feature is any one of a linear feature, a planar feature, and a spherical feature; The determining of the attribute features of each point in each frame of point cloud data also includes: For any point that is not a ground point, a plurality of points whose distances from the any point are within a preset distance range are obtained; and based on the obtained plurality of points, a three-dimensional feature value of the any point is determined; Based on the three-dimensional eigenvalues, determining the linear eigenvalue, the planar eigenvalue and the spherical eigenvalue corresponding to the any point; Based on the linearity eigenvalues, the planarity eigenvalues ​​and the sphericality eigenvalues, a shape feature corresponding to any point cloud is determined.

5. The method for determining the position and posture according to claim 4, characterized in that: The determining of the shape feature corresponding to any point based on the linear feature value, the planar feature value and the spherical feature value comprises: If the linear feature value of a point is greater than the fourth set threshold, it is determined that the point has a linear feature.

6. The method for determining the position and posture according to claim 5, characterized in that: The determining of the shape feature corresponding to any point cloud based on the linearity feature value, the planarity feature value and the sphericality feature value further includes: If the linear feature value of a point is less than or equal to the fourth set threshold, and the surface feature value of the point is greater than the fifth set threshold, it is determined that the point has the surface feature.

7. The method for determining a position and posture according to claim 5, characterized in that: The determining of the shape feature corresponding to any point cloud based on the linearity feature value, the planarity feature value and the sphericality feature value further includes: If the linear feature value of a point is less than or equal to the fourth set threshold, the surface feature value of the point is less than or equal to the fifth set threshold, and the spherical feature value of the point is greater than the sixth set threshold, it is determined that the point cloud has a spherical feature.

8. The method for determining a position and posture according to claim 1 or 2, characterized in that: The determining, based on the coordinate information of each point in each registration point pair and the attribute characteristics of each registration point pair, the position and posture characteristic information of the radar when collecting each frame of point cloud data includes: Based on the coordinate information of the points in each registration point pair and the attribute characteristics of each registration point pair, the error metric information of each registration point pair is determined respectively; Based on the error measurement information of each registration point pair, determine the pose transformation matrix between each two adjacent frames of point cloud data; Based on the pose conversion matrix between each two adjacent frames of point cloud data, the pose feature information of the radar when collecting each frame of point cloud data is determined.

9. The method for determining a position and posture according to claim 8, characterized in that: The attribute features include features of whether it is a ground point or a non-ground point, the features of the non-ground point include shape features, and the shape features include spherical features; the error metric information includes Euclidean distance error information; The step of determining the error metric information of each registration point pair based on the coordinate information of the points in each registration point pair and the attribute characteristics of each registration point pair includes: For a registration point pair having a non-ground spherical feature, the Euclidean distance error information of the registration point pair is determined based on the coordinate information of each point in the registration point pair.

10. The method for determining position and posture according to claim 8, characterized in that: The attribute features include features of whether it is a ground point or a non-ground point, the features of the non-ground point include shape features, and the shape features include linear features; the error measurement information includes point-line distance error information; The method of determining the error metric information of each registration point pair based on the coordinate information of the points in each registration point pair and the attribute characteristics of each registration point pair also includes: For a registration point pair with non-ground linear features, point-line distance error information of the registration point pair is determined based on the coordinate information of each point in the registration point pair and the vector of the straight line where one of the points is located.

11. The method for determining a position and posture according to claim 8, characterized in that: The attribute features include features of whether it is a ground point or a non-ground point, the features of the non-ground point include shape features, and the shape features include surface features; the error measurement information includes point-to-surface distance error information; The method of determining the error metric information of each registration point pair based on the coordinate information of the points in each registration point pair and the attribute characteristics of each registration point pair also includes: For a pair of registration points with non-ground planar features or a pair of registration points with ground features, the point-to-surface distance error information of the pair of registration points is determined based on the coordinate information of each point in the pair of registration points and the normal vector of the surface where the point in the pair of registration points belongs to the next frame of point cloud data.

12. The method for determining a position and posture according to claim 8, characterized in that: The determining of the pose characteristic information of the radar when collecting each frame of point cloud data based on the pose conversion matrix between each two adjacent frames of point cloud data includes: Based on the pose conversion matrix between each two adjacent frames of point cloud data, the pose transformation constraint information of the radar when collecting each two adjacent frames of point cloud data is determined; Pre-integrate multiple measurement pose information of the radar when the radar collects each two adjacent frames of point cloud data, and determine pre-integration constraint information of the radar when collecting each two adjacent frames of point cloud data; Based on the pose transformation constraint information and pre-integration constraint information, the pose feature information of the radar when collecting each frame of point cloud data is determined.

13. The method for determining a position and posture according to claim 12, characterized in that: Before determining the pose feature information of the radar when collecting each frame of point cloud data based on the pose transformation constraint information and the pre-integration constraint information, it also includes: Determining a first weight corresponding to the posture transformation constraint information based on a covariance corresponding to the posture transformation matrix; Pre-integrate multiple measurement pose information of the radar when the radar collects each two adjacent frames of point cloud data to obtain a second weight corresponding to the pre-integration constraint information; The determining of the posture feature information of the radar when collecting each frame of point cloud data based on the posture transformation constraint information and the pre-integration constraint information includes: Based on the posture transformation constraint information, the pre-integration constraint information, the first weight and the second weight, the posture feature information of the radar when collecting each frame of point cloud data is determined.

14. The method for determining a position and posture according to claim 12, characterized in that: Before determining the position and posture feature information of the radar when collecting each frame of point cloud data, it also includes: Based on the global positioning information of the radar when the radar collects each frame of point cloud data, the estimated posture information of the radar when collecting each frame of point cloud data is determined; Determine the global positioning transformation constraint information based on the estimated pose information of the radar when collecting each two adjacent frames of point cloud data; The determining of the posture feature information of the radar when collecting each frame of point cloud data based on the posture transformation constraint information and the pre-integration constraint information includes: The method determines the posture feature information of the radar when collecting each frame of point cloud data based on the posture transformation constraint information, the pre-integration constraint information and the global positioning transformation constraint information.

15. The method for determining a position and posture according to claim 1, characterized in that: Also includes: Based on the position and posture feature information of the radar when collecting continuous multi-frame point cloud data, the movement mileage information of the radar is determined.

16. A posture determination device, characterized in that: include: A data acquisition module, used to obtain at least two frames of point cloud data collected by the radar, multiple measurement pose information of each frame of point cloud data collected by the radar, the measurement time of each measurement pose information, and the time when the radar transmits each batch of scan lines; For each batch of scan lines transmitted by the radar, based on the time when the radar transmits the batch of scan lines and the measurement time of each measurement pose information, the measurement pose information before the radar transmits the batch of scan lines and the measurement pose information after the radar transmits the batch of scan lines are filtered out from multiple measurement pose information, and interpolation is performed based on the filtered two pieces of measurement pose information to obtain the pose information when the radar transmits the batch of scan lines; For the point cloud data collected by non-first batch scanning lines in each frame of point cloud data, based on the posture information of the radar when the batch of scanning lines is emitted, the coordinates of the point cloud data collected by the batch of scanning lines are converted to the coordinate system of the radar corresponding to the emission time of the first batch of scanning lines in the frame of point cloud data; An attribute determination module is used to determine the attribute characteristics of each point in each frame of point cloud data; a point pair registration module, for matching points in each frame of point cloud data except the first frame of point cloud data in the at least two frames of point cloud data with points in a previous frame of point cloud data that have the same attribute features, to obtain a plurality of registration point pairs; The posture determination module is used to determine the posture feature information of the radar when collecting each frame of point cloud data based on the coordinate information of each point in each registration point pair and the attribute characteristics of each registration point pair.

17. An electronic device, characterized in that: include: A processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the storage medium communicate via the bus, and the processor executes the machine-readable instructions to perform the posture determination method as described in any one of claims 1 to 15.

18. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the posture determination method according to any one of claims 1 to 15 is executed.

Images