A roadside radar position recognition method and device

By generating virtual viewpoints and calculating similarity distances, and automatically identifying the roadside radar position, the problem of difficult to determine the initial value in the prior art is solved, and the safety and efficiency of the road traffic system are improved.

CN118707470BActive Publication Date: 2024-11-15WUHAN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411205377.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-30
Publication Date
2024-11-15
Estimated Expiration
2044-08-30

AI Technical Summary

Technical Problem

The existing roadside radar calibration methods rely on GPS signals or only focus on inter-sensor calibration, which makes it difficult to determine the initial value, cannot meet all scenario needs, and increases human resource consumption.

Method used

By obtaining radar scanning data and map data of roadside radar, multiple virtual viewpoints are generated, virtual radar parameters are calculated, radar scanning data and virtual scan characteristics are extracted, similarity distance is calculated, and the target virtual viewpoint is the position of the roadside radar.

Benefits of technology

It realizes automatic identification of the radar position on the roadside, solves the problem of difficult to determine the initial value, is suitable for the integration of vehicle-road collaborative data in urban areas, and improves the safety and efficiency of the road traffic system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118707470B_ABST
    Figure CN118707470B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and device for identifying the position of a roadside radar, the method comprising: obtaining radar scanning data of a roadside radar at an intersection and map data of the intersection; generating the positions of multiple virtual viewpoints based on the map data; obtaining virtual radar parameters based on the radar scanning data; converting the map data into multiple virtual scans based on the positions of multiple virtual viewpoints and virtual radar parameters; extracting the features of the radar scanning data and multiple virtual scans respectively; calculating the first distances between the radar scanning data and multiple virtual scans respectively based on the features of the radar scanning data and the features of multiple virtual scans; determining a target virtual viewpoint based on the values ​​of multiple first distances, the position of the target virtual viewpoint being the position of the roadside radar. The present invention solves the problem that it is difficult to determine the initial value during the calibration of the roadside radar.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of vehicle-road cooperative technology, and in particular to a roadside radar position recognition method and device. Background Art

[0002] In intelligent transportation systems, roadside radar systems are one of the key technologies for achieving vehicle-road collaboration. They provide support for traffic management and decision-making by collecting vehicle location, speed, and flow information in real time. They are widely used to monitor traffic flow, detect traffic incidents, and provide data on road usage. Accurate calibration of roadside radars is a prerequisite for achieving efficient collaborative perception in practical applications. Current calibration methods have limitations: some methods rely on GPS signals to provide initial positioning information for radars, while others focus only on calibration between sensors. In practical applications, equipping each radar system with a GPS signal not only increases the consumption of human resources, but also calibration between sensors alone may not meet the needs of all scenarios. Therefore, it is particularly urgent to develop a method that can automatically complete the position identification of roadside radars. Summary of the invention

[0003] The present invention provides a method and device for identifying the position of a roadside radar, so as to solve the problem that an initial value is difficult to determine during the calibration process of the roadside radar.

[0004] According to one aspect of the present invention, a method for identifying a roadside radar position is provided, comprising:

[0005] Obtain radar scanning data of the roadside radar at the intersection, as well as map data of the intersection;

[0006] generating positions of a plurality of virtual viewpoints based on the map data;

[0007] Acquire virtual radar parameters based on the radar scanning data;

[0008] Based on the positions of the plurality of virtual viewpoints and the virtual radar parameters, converting the map data into virtual scans respectively corresponding to the positions of the plurality of virtual viewpoints;

[0009] respectively extracting features of the radar scan data and a plurality of the virtual scans;

[0010] Based on the features of the radar scan data and the features of the plurality of virtual scans, calculating first distances between the radar scan data and the plurality of virtual scans respectively; the first distances are used to express the similarity between the radar scan data and the virtual scans;

[0011] A target virtual viewpoint is determined based on multiple values ​​of the first distances, and the location of the target virtual viewpoint is the location of the roadside radar.

[0012] Optionally, generating the positions of multiple virtual viewpoints based on the map data includes:

[0013] Obtaining a vector group for each road in the map data;

[0014] Linear interpolation is performed on each vector in the vector group, and a plurality of virtual viewpoints are generated on each vector.

[0015] Optionally, the obtaining of a vector group for each road in the map data includes:

[0016]

[0017] Among them, R i The vector group representing the i-th road, V i,j represents the jth vector in the vector group that makes up the i-th road, n i represents the number of coordinate points on the i-th road, P i,j represents the jth coordinate point on the i-th road;

[0018] The performing linear interpolation on each vector in the vector group, and generating a plurality of virtual viewpoints on each vector, comprises:

[0019]

[0020] Among them, P i (k) represents the kth virtual viewpoint generated on the jth vector on the i-th road, m represents the number of virtual viewpoints that are expected to be generated on the jth vector, represents the starting point of the jth vector in the i-th road, Represents the end point of the jth vector in the i-th road.

[0021] Optionally, the acquiring virtual radar parameters based on the radar scanning data includes:

[0022] Calculate a second distance, a first horizontal angle, and a first vertical angle from each point in the radar scanning data to the roadside radar;

[0023] Acquire the number of virtual radar lines based on the number of the first vertical angles;

[0024] Acquire a measurement range of the virtual radar based on a second distance of each point;

[0025] Acquire a horizontal angle interval and a vertical angle interval of the virtual radar based on the first horizontal angle and the first vertical angle of each point;

[0026] Taking the maximum and minimum vertical angle intervals as the vertical field angle of the virtual radar;

[0027] A vertical angle set of virtual radar laser rays is calculated based on the vertical angle interval and the number of virtual radar lines, and a horizontal angle set of the virtual radar laser rays is calculated based on the horizontal angle interval.

[0028] Optionally, the converting the map data into virtual scans respectively corresponding to the locations of the multiple virtual viewpoints based on the locations of the multiple virtual viewpoints and the virtual radar parameters includes:

[0029] Clipping the map data based on the position and parameters of the virtual viewpoint;

[0030] Calculating a third distance, a second horizontal angle, and a second vertical angle from each point in the cropped map data to the virtual viewpoint;

[0031] If the second horizontal angle belongs to the horizontal angle set, and the second vertical angle belongs to the vertical set, retaining the point corresponding to the second horizontal angle;

[0032] The point in the map data that is closest to the virtual viewpoint is retained to obtain the virtual scan corresponding to the virtual viewpoint, and then to obtain multiple virtual scans corresponding to multiple virtual viewpoints.

[0033] Optionally, the extracting the features of the radar scan data and the plurality of virtual scans respectively includes:

[0034] The radar scanning data and the plurality of virtually scanned point cloud data are binned according to horizontal direction, horizontal angle and vertical angle;

[0035] Encode the points in each bin;

[0036] By accumulating the encoding functions of all boxes in a unified cross section, a feature descriptor of the point cloud is calculated, where the feature descriptor includes features corresponding to all boxes along different horizontal directions and different horizontal angles.

[0037] Optionally, encoding the points in each box includes:

[0038]

[0039] Where P i,j,k Display box The set of midpoints, N r represents the number of boxes along the horizontal direction, N h The number of bins representing the horizontal angle, N v The number of bins representing the vertical angle; Represents the number of points in the corresponding set, Nth Indicates the quantity threshold; represents the encoding function, represents the judgment function;

[0040] The feature descriptor of the point cloud is calculated by accumulating the encoding functions of all boxes in the unified cross section, including:

[0041]

[0042] F ij represents the features corresponding to all boxes along the i-th horizontal direction and the j-th horizontal angle, and F represents the feature descriptor of the entire point cloud.

[0043] Optionally, the calculating, based on the feature of the radar scan data and the features of the multiple virtual scans, first distances between the radar scan data and the multiple virtual scans respectively includes:

[0044]

[0045] In the formula, and The number of columns representing the radar scan data and the feature descriptors of the plurality of virtual scans respectively; i represents the distance between the virtual scan and the radar scan data after being shifted i columns to the right, dis represents the first distance between the virtual scan and the radar scan data, and g(i, j) represents the index of the jth column in the feature descriptor after being shifted i columns to the right; data of the g(i, j)th column of the feature descriptor representing the virtual scan, The data of the j-th column of the feature descriptor representing the radar scan data.

[0046] Optionally, the determining a target virtual viewpoint based on a plurality of values ​​of the first distances, wherein the location of the target virtual viewpoint is the location of a roadside radar, includes:

[0047] Sort the first distances to obtain top K candidate virtual viewpoints;

[0048] Divide the candidate virtual viewpoints into multiple clusters, and calculate the fourth distance between the elements in each cluster;

[0049] A distance matrix is ​​constructed based on the first distance and the fourth distance; each element in the first row of the distance matrix is ​​a first distance between the radar scan data and the virtual scans of the first K candidate virtual viewpoints, and other rows except the first row are fourth distances between any two virtual scans of the first K candidate virtual viewpoints;

[0050] The fifth distance between each row in the distance matrix and the first row is calculated, and the target virtual viewpoint is determined according to the virtual scan corresponding to the row having the shortest fifth distance.

[0051] According to another aspect of the present invention, there is provided a roadside radar position recognition device, comprising:

[0052] A first acquisition unit, used to acquire radar scanning data of a roadside radar at an intersection and map data of the intersection;

[0053] A generating unit, configured to generate positions of a plurality of virtual viewpoints based on the map data;

[0054] A second acquisition unit, configured to acquire virtual radar parameters based on the radar scanning data;

[0055] A conversion unit, configured to convert the map data into virtual scans respectively corresponding to the locations of the multiple virtual viewpoints based on the locations of the multiple virtual viewpoints and the virtual radar parameters;

[0056] An extraction unit, used to extract the features of the radar scan data and the plurality of virtual scans respectively;

[0057] a calculation unit, configured to calculate first distances between the radar scan data and the plurality of virtual scans respectively based on the features of the radar scan data and the features of the plurality of virtual scans; the first distances being used to express the similarity between the radar scan data and the virtual scans;

[0058] A position identification unit is used to determine a target virtual viewpoint based on multiple values ​​of the first distances, and the position of the target virtual viewpoint is the position of the roadside radar.

[0059] The technical solution of the embodiment of the present invention generates the positions of multiple virtual viewpoints through map data, and obtains virtual radar parameters through radar scanning data, thereby generating a virtual scan of a virtual pilot point; by extracting the features of the radar scanning data and the multiple virtual scans, the first distances of the radar scanning data and the multiple virtual scans are calculated using the feature values, and the target virtual viewpoint is determined according to the value of the first distance, thereby determining the position of the roadside radar, which solves the problem that the initial value is difficult to determine during the roadside radar calibration process. The present invention can be applied to data integration work for urban vehicle-road collaboration, and helps to achieve a safer and more efficient road traffic system.

[0060] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0062] Figure 1 It is a flow chart of an embodiment of a roadside radar position identification method provided by the present invention;

[0063] Figure 2 A flowchart of a sub-embodiment of a roadside radar position identification method provided by the present invention;

[0064] Figure 3 A flowchart of a sub-embodiment of a roadside radar position identification method provided by the present invention;

[0065] Figure 4 A flowchart of another sub-embodiment of a roadside radar position identification method provided by the present invention;

[0066] Figure 5 A flowchart of another sub-embodiment of a roadside radar position identification method provided by the present invention;

[0067] Figure 6 It is a schematic diagram of the structure of a roadside radar position recognition device implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0068] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0069] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0070] Figure 1 A flow chart of a method for identifying a roadside radar position is provided for the first embodiment of the present invention. Figure 1 As shown, the method includes:

[0071] S101, obtaining radar scanning data of a roadside radar at an intersection and map data of the intersection;

[0072] Roadside radar is specifically roadside laser radar, which is a high-precision laser radar technology, also known as mobile laser radar. It uses laser beams to scan the ground, can obtain 3D environmental information around the vehicle, and realize the positioning, perception and control of the vehicle.

[0073] This embodiment can obtain radar data collected by multiple different roadside radars on the vehicle in different directions at multiple intersections, and can obtain map data of the intersection. The radar scanning data can be point cloud data collected by the roadside laser radar. The map data can be high-precision map data.

[0074] S102, generating positions of multiple virtual viewpoints based on the map data;

[0075] In this embodiment, the positions of different points in the map data can be selected as the positions of the virtual viewpoints. There can be many selection methods, and this embodiment does not make any specific restrictions.

[0076] S103, acquiring virtual radar parameters based on the radar scanning data;

[0077] For the acquired radar scanning data, the relevant parameters of the virtual radar used when acquiring the virtual scan can be determined. For example, for any point selected in the radar scanning data, the distance, horizontal angle, vertical angle, etc. from each point to the point can be calculated based on the point cloud data collected by the roadside laser radar, and then the distance, horizontal angle, vertical angle, etc. from each point in the radar scanning data to the point can be used to determine the virtual radar parameters used when the point is used as a virtual viewpoint.

[0078] S104, based on the positions of the plurality of virtual viewpoints and virtual radar parameters, converting the map data into a plurality of virtual scans;

[0079] Combined with the selected virtual pilot point and the virtual radar parameter information used at the point, the map data can be converted into virtual scans corresponding to multiple different virtual pilot points. The virtual scan of each virtual pilot point corresponds to the virtual radar scanning data scanned by the virtual radar with the point as the center. That is, this embodiment finds the virtual scan closest to the radar scanning data from multiple virtual scans, and its corresponding virtual viewpoint can be used as the actual radar location.

[0080] S105, extracting features of the radar scan data and the multiple virtual scans respectively;

[0081] The features extracted in this embodiment may be features representing the similarity between the radar scan data and the virtual scan, and there are many ways to extract the features, which are not specifically limited in this embodiment. The virtual scan closest to the radar scan data may be selected by calculating the similarity between the radar scan data and the virtual scan.

[0082] S106. Based on the characteristics of the radar scan data and the characteristics of the multiple virtual scans, calculate the first distances between the radar scan data and the multiple virtual scans respectively; the first distances are used to express the similarity between the radar scan data and the virtual scans.

[0083] S107: Determine a target virtual viewpoint based on multiple values ​​of the first distances, where the location of the target virtual viewpoint is the location of a roadside radar.

[0084] In this embodiment, the first distance can be expressed as the similarity between the radar scan data and the virtual scan, that is, by calculating the first distance between the radar scan data and multiple virtual scans, it can be used to find the virtual scan closest to the radar scan data from multiple virtual scans, and the virtual viewpoint corresponding to the virtual scan is found as the target virtual test point, so that the location of the virtual test point can be used as the location of the roadside radar. For example, the virtual viewpoint corresponding to the minimum value of the first distance can be used as the target virtual viewpoint, and the virtual scan of the virtual viewpoint is closest to the real radar scan data.

[0085] The technical solution of the embodiment of the present invention generates the positions of multiple virtual viewpoints through map data, and obtains the parameters of multiple virtual viewpoints through radar scanning data, thereby generating a virtual scan of a virtual pilot point; by extracting the characteristics of the radar scanning data and the multiple virtual scans, the first distance of the radar scanning data and the multiple virtual scans is calculated with the characteristic value, and the target virtual pilot point is determined according to the value of the first distance, thereby determining the position of the roadside radar, which solves the problem that the initial value is difficult to determine during the roadside radar calibration process. The present invention can be applied to data integration work of urban vehicle-road collaboration, and helps to achieve a safer and more efficient road traffic system.

[0086] In one embodiment, if Figure 2 As shown, step S102 includes:

[0087] S1021, obtaining a vector group of each road in the map data;

[0088] This embodiment can obtain road information from map data and use a series of vectors to represent each road, so that each road has a corresponding vector group.

[0089] In one embodiment, the vector group of each road in the map data is obtained, specifically:

[0090]

[0091] Among them, R i The vector group representing the i-th road, V i,j Represents the jth vector of the vector group that makes up the i-th road, n i represents the number of coordinate points on the i-th road, P i,j represents the jth coordinate point on the i-th road;

[0092] S1022. Perform linear interpolation on each vector in the vector group, and generate a plurality of virtual viewpoints on each vector.

[0093] Linear interpolation is performed on each vector in the vector group, and multiple virtual viewpoints are generated on each vector, including:

[0094]

[0095] Among them, P i (k, j) represents the kth virtual viewpoint generated on the jth vector on the i-th road, m represents the number of virtual viewpoints that are expected to be generated on the jth vector, represents the starting point of the jth vector in the i-th road, Represents the end point of the jth vector in the i-th road.

[0096] In one embodiment, step S103 obtains virtual radar parameters based on the radar scanning data.

[0097] Since the installation height of the roadside radar is relatively fixed, the Z axis of the corresponding virtual radar can be uniformly set to the same height. For the attitude information of the virtual radar, considering the rotation invariance of the feature descriptor (the feature descriptor is a matrix representing the radar scanning data and the feature set of the virtual scan), the yaw angle, roll angle and pitch angle of the virtual radar can be set to 0. That is, the virtual radar parameters are obtained based on the radar scanning data, which may include the number of virtual radar lines, the measurement range of the virtual radar, the horizontal angle interval, the vertical angle interval, the vertical field of view angle, the vertical angle set and the horizontal angle set.

[0098] Specifically, the second distance, the first horizontal angle, and the first vertical angle from each point in the radar scanning data to the roadside radar may be calculated.

[0099] In one embodiment, the method for calculating the second distance, the first horizontal angle, and the first vertical angle is:

[0100]

[0101] in, represents the i-th point in the radar scan data, Represent the x, y, and z coordinates of point i, respectively. i represents the second distance of the i-th point, H i Represents the first horizontal angle of the i-th point, V i Represents the first vertical angle of the i-th point.

[0102] The number of virtual radar lines is obtained based on the number of the first vertical angles.

[0103] A measurement range of the virtual radar is obtained based on the second distance of each point.

[0104] Acquire a horizontal angle interval and a vertical angle interval of the virtual radar based on a first horizontal angle and a first vertical angle of each point;

[0105] Taking the maximum and minimum vertical angle intervals as the vertical field angle of the virtual radar;

[0106] The vertical set of virtual radar laser rays is calculated based on the vertical angle interval and the number of virtual radar lines, and the horizontal angle set of virtual radar laser rays is calculated based on the horizontal angle interval.

[0107] In one embodiment, the number of the first vertical angles may be used as the number of virtual radar lines N rays; The maximum distance among the second distances from each point in the radar scanning data to the roadside radar is used as the measurement range of the virtual radar; according to the first horizontal angle and the first vertical angle of each point obtained above, the minimum first horizontal angle interval between points with the same first vertical angle can be selected as the horizontal angle interval; the maximum and minimum vertical angle intervals As the vertical field of view angle of the virtual radar. In this embodiment, considering the limitations of common parameters of real laser radars, the number of virtual radar lines can also be constrained to common parameters such as 16, 32, 64, 128, and the measurement range and vertical field of view of the virtual radar can also be rounded to obtain the final virtual radar estimation parameters. Finally, the vertical angle set of the virtual radar laser ray can be calculated based on the vertical angle interval and the number of virtual radar lines by interpolation, and the horizontal angle set of the virtual radar laser ray can be calculated based on the horizontal angle interval. Specifically, the calculation method of the vertical angle set and the horizontal angle set can be:

[0108]

[0109] Where N rays Indicates the number of lines of the virtual radar; is the vertical field of view of the virtual radar; is the horizontal angle interval of the virtual radar; A vn represents the set of vertical angles, A vi A represents the vertical angle of the i-th point; hn represents the horizontal angle set, A hi Represents the horizontal angle of the i-th point.

[0110] In one embodiment, if Figure 3 As shown, step S104 includes:

[0111] S1041. Crop the map data based on the position of the virtual viewpoint and the measurement range of the virtual radar.

[0112] Since the sensing range of the radar is limited, this embodiment can take the position of the virtual viewpoint as the center and the measurement range parameter of the virtual radar as a constraint to crop the map data. Of course, the map data can also be cropped with other ranges as the measurement radius, for example, 0.9 times the measurement range parameter can be selected as the measurement radius. That is, this embodiment can obtain map data as the data of the virtual radar at a virtual viewpoint.

[0113] S1042: Calculate a third distance, a second horizontal angle, and a second vertical angle from each point in the cropped map data to the virtual viewpoint.

[0114] The third distance, second horizontal angle and second vertical angle from each point in the cropped map data to the virtual viewpoint can be calculated as the third distance, second horizontal angle and second vertical angle from each point in the point cloud image centered on the virtual viewpoint to the virtual viewpoint. The calculation method of the third distance, second horizontal angle and second vertical angle is the same as the calculation method of the second distance, first horizontal angle and first vertical angle.

[0115] S1043: Generate a virtual radar scan line according to the second horizontal angle and the second vertical angle.

[0116] A virtual radar scan line may be generated according to the second horizontal angle and the second vertical angle of each point in the cropped map data to the virtual viewpoint. For example, when there are N points in the cropped map data, and M of the points have the same second horizontal angle or second vertical angle as other points, then NM radar scan lines may be generated. Of course, the above is only an exemplary description, and other methods may be used to generate virtual radar scan lines.

[0117] S1044: If there is a first point whose second horizontal angle belongs to the horizontal angle set, and if the second vertical angle of the first point belongs to the vertical angle set, the first point is on the scanning line of the virtual radar, and the first point is retained.

[0118] If the second horizontal angle of the first point belongs to the horizontal angle set obtained in step S1022, and the second vertical angle belongs to the vertical set obtained in step S1022, it means that the point is located on the radar scanning line, and the point can be retained as a scanning point for virtual scanning.

[0119] S1045. Keep the point on the scan line of the virtual radar that is closest to the virtual viewpoint, obtain the virtual scan corresponding to the virtual viewpoint, and then obtain the virtual scans corresponding to the multiple virtual viewpoints respectively.

[0120] Since the real radar does not have penetration during the scanning process, in this embodiment, only the point closest to the virtual viewpoint on each scan line in the cropped map data needs to be retained, so that the virtual scan corresponding to each virtual viewpoint can be obtained, that is, each scan line in the virtual scan of the virtual viewpoint only contains one point. The virtual scan obtained in the above manner can greatly reduce the amount of calculation when calculating the first distance later.

[0121] In one embodiment, if Figure 4 As shown, step S105 includes:

[0122] S1051. Binning the radar scanning data and the point cloud data of the multiple virtual scans according to the horizontal direction, horizontal angle and vertical angle.

[0123] It should be noted that in this embodiment, the radar scan data and the point cloud data of multiple virtual scans can be binned according to the horizontal direction, horizontal angle and vertical angle. Specifically, the point cloud data can be divided into azimuth and radial bins at equal intervals, and then further divided along the vertical angle. After the division is completed, it can be obtained boxes, where N r represents the number of boxes along the horizontal direction, N h The number of bins representing the horizontal angle, N v Indicates the number of boxes for vertical angles. In addition, since the horizontal field of view of the real roadside scan is usually small, while the horizontal field of view of the generated virtual scan is 360 degrees, the two scans are not the same.

[0124] S1052, encoding the points in each box;

[0125] The points in each box can be encoded, and then the feature descriptor of the point cloud can be calculated by accumulating the encoding functions of all boxes in the unified cross section.

[0126] In one embodiment, encoding the points in each box is specifically as follows:

[0127]

[0128] Where P i,j,k Display box The set of midpoints, N r represents the number of boxes along the horizontal direction, N h The number of bins representing the horizontal angle, N v The number of bins representing the vertical angle; Represents the number of points in the corresponding set, N th Indicates the quantity threshold; represents the encoding function, represents the judgment function;

[0129] S1053. By accumulating the encoding functions of all boxes in the unified cross section, a feature descriptor of the point cloud is calculated, where the feature descriptor includes features corresponding to all boxes along different horizontal directions and different horizontal angles.

[0130] In one embodiment, the feature descriptor of the point cloud is calculated by accumulating the encoding functions of all boxes in the unified cross section, including:

[0131]

[0132] F ijrepresents the features corresponding to all boxes along the i-th horizontal direction and the j-th horizontal angle, and F represents the feature descriptor of the entire point cloud.

[0133] In one embodiment, step S106 calculates the first distances between the radar scan data and the plurality of virtual scans based on the features of the radar scan data and the features of the plurality of virtual scans, specifically:

[0134] Considering that the virtual radar field of view is set to 360 degrees, while the real radar field of view is usually less than 360 degrees, the dimensions of the feature descriptors of the two are not the same. Therefore, each column in the feature can be regarded as a discrete distribution, by continuously shifting the virtual scan to the right, and using cosine similarity to calculate the distance between the corresponding columns after each shift, and taking the minimum distance as the final distance, and taking the product of the number of shifts and the horizontal angle gap as the yaw angle. A virtual scan of a virtual viewpoint is a frame of virtual scan. Then the distance dis between the radar scan data and a virtual scan can be calculated as:

[0135]

[0136] In the formula, and The number of columns representing the radar scan data and the feature descriptors of the plurality of virtual scans respectively; i represents the distance between the virtual scan and the radar scan data after being shifted i columns to the right, dis represents the first distance between the virtual scan and the radar scan data, and g(i, j) represents the index of the jth column in the feature descriptor after being shifted i columns to the right; data of the g(i, j)th column of the feature descriptor representing the virtual scan, The data of the j-th column of the feature descriptor representing the radar scan data.

[0137] In one embodiment, if Figure 5 As shown, step S107 includes:

[0138] S1071. Sort the multiple first distances to obtain top K candidate virtual viewpoints.

[0139] S1072: Divide the candidate virtual viewpoints into multiple clusters, and calculate a fourth distance between elements in each cluster.

[0140] It should be noted that the first distance represents the similarity between the virtual scan and the radar scan data, so the virtual scans can be sorted according to the first distance. For example, the first 3 or 5 virtual scans with the largest first distance (largest similarity) can be selected as candidate scans, so as to select the virtual scan most similar to the radar scan data as the target virtual scan.

[0141] This embodiment can divide the candidate scans into multiple clusters based on the Euclidean distance, and calculate the fourth distance between the elements in each cluster, that is, calculate the similarity between two candidate virtual viewpoints in each cluster.

[0142] S1073. Construct a distance matrix based on the first distance and the fifth distance; each element in the first row of the distance matrix is ​​the first distance between the radar scan data and the virtual scans of the first K candidate virtual viewpoints, and the other rows except the first row are the fourth distances between any two of the virtual scans of the first K candidate virtual viewpoints.

[0143] In one embodiment, the similarity between each element in the cluster is calculated, and the similarity between the radar scanning data and each element in the cluster is added to construct a distance matrix. The specific formula is as follows:

[0144]

[0145] Among them, dis Li represents the distance between the radar scan data and the ith virtual scan, dis ij represents the distance between the ith virtual scan and the jth virtual scan.

[0146] S1074. Calculate the fifth distance between each row in the distance matrix and the first row, and determine the target virtual viewpoint according to the virtual scan corresponding to the row having the shortest fifth distance.

[0147] Based on the above distance matrix, the distance between each row and the first row in the matrix can be calculated as the fifth distance between any two virtual scans, and the fifth distance between any virtual scan and the real radar scan data. Specifically, each row can be regarded as a unit vector, and the Euclidean distance between each row and other rows can be calculated. For example, in the above formula, the first row can be regarded as the unit vector of the real radar scan data, the second row can be regarded as the unit vector of the first candidate virtual scan, and the kth row can be regarded as the unit vector of the k-1th candidate virtual scan. The similarity between the i-th candidate scan and the real radar scan is determined by calculating the Euclidean distance between the i-th row and the first row. The smallest value in the fifth distance between the unit vector of the candidate virtual scan and the unit vector of the real radar scan data is determined, which means that the candidate virtual scan is the most similar to the real radar scan data. This method can be used to verify the error caused by a data error in a candidate virtual scan, thereby ensuring that the final candidate virtual scan is the target virtual scan.

[0148] The present invention is based on high-precision map data, introduces virtual viewpoint technology, realizes position recognition of roadside radar based on feature extraction, and solves the problem that initial value is difficult to determine during roadside radar calibration. The present invention can be applied to data integration work of urban vehicle-road collaboration, and helps to realize a safer and more efficient road traffic system.

[0149] Figure 6 The following is a schematic diagram of the structure of a roadside radar position recognition device provided by an embodiment of the present invention. Figure 6 As shown, the device comprises:

[0150] A first acquisition unit 601 is used to acquire radar scanning data of a roadside radar at an intersection and map data of the intersection;

[0151] A generating unit 602, configured to generate a position of a virtual radar based on the map data;

[0152] A second acquisition unit 603 is used to acquire parameters of a plurality of virtual viewpoints based on the radar scanning data;

[0153] A conversion unit 604, configured to convert the map data into virtual scans corresponding to the locations of the multiple virtual viewpoints based on the locations of the multiple virtual viewpoints and the virtual radar parameters;

[0154] An extraction unit 605, configured to extract features of the radar scan data and the plurality of virtual scans respectively;

[0155] A calculation unit 606 is used to calculate first distances between the radar scan data and the plurality of virtual scans respectively based on the features of the radar scan data and the features of the plurality of virtual scans; the first distances are used to express the similarity between the radar scan data and the virtual scans;

[0156] The position identification unit 607 is used to determine a target virtual viewpoint based on multiple values ​​of the first distances, and the position of the target virtual viewpoint is the position of the roadside radar.

[0157] In one embodiment, the generating unit 602 is further configured to obtain a vector group for each road in the map data; perform linear interpolation on each vector in the vector group, and generate a plurality of virtual viewpoints on each vector.

[0158] In one embodiment, the second acquisition unit 603 is further used to calculate a second distance, a first horizontal angle, and a first vertical angle from each point in the radar scanning data to the roadside radar;

[0159] Calculate a second distance, a first horizontal angle, and a first vertical angle from each point in the radar scanning data to the roadside radar;

[0160] Acquire the number of virtual radar lines based on the number of the first vertical angles;

[0161] Acquire a measurement range of the virtual radar based on a second distance of each point;

[0162] Acquire a horizontal angle interval and a vertical angle interval of the virtual radar based on the first horizontal angle and the first vertical angle of each point;

[0163] Taking the maximum and minimum vertical angle intervals as the vertical field angle of the virtual radar;

[0164] A vertical angle set of virtual radar laser rays is calculated based on the vertical angle interval and the number of virtual radar lines, and a horizontal angle set of the virtual radar laser rays is calculated based on the horizontal angle interval.

[0165] In one embodiment, the conversion unit 604 is further configured to clip the map data based on the position of the virtual viewpoint and the measurement range of the virtual radar;

[0166] Calculating a third distance, a second horizontal angle, and a second vertical angle from each point in the cropped map data to the virtual viewpoint;

[0167] generating a virtual radar scan line from the second horizontal angle and the second vertical angle;

[0168] If there is a first point whose second horizontal angle belongs to the horizontal angle set, and whose second vertical angle belongs to the vertical angle set, the first point is on the scanning line of the virtual radar, and the first point is retained;

[0169] The point on the scan line of the virtual radar that is closest to the virtual viewpoint is retained to obtain the virtual scan corresponding to the virtual viewpoint, and then the virtual scans corresponding to the multiple virtual viewpoints are obtained.

[0170] In one embodiment, the extraction unit 605 is further used to bin the radar scanning data and the point cloud data of the plurality of virtual scans according to the horizontal direction, the horizontal angle and the vertical angle;

[0171] Encode the points in each bin;

[0172] By accumulating the encoding functions of all boxes in a unified cross section, a feature descriptor of the point cloud is calculated, where the feature descriptor includes features corresponding to all boxes along different horizontal directions and different horizontal angles.

[0173] In one embodiment, the position identification unit 607 is further used to sort the plurality of the first distances to obtain the first K candidate virtual viewpoints;

[0174] Divide the candidate virtual viewpoints into multiple clusters, and calculate the fourth distance between the elements in each cluster;

[0175] A distance matrix is ​​constructed based on the first distance and the fifth distance; each element in the first row of the distance matrix is ​​a first distance between the radar scan data and the virtual scans of the first K candidate virtual viewpoints, and other rows except the first row are fourth distances between any two virtual scans in the virtual scans of the first K candidate virtual viewpoints;

[0176] The fifth distance between each row in the distance matrix and the first row is calculated, and the target virtual viewpoint is determined according to the virtual scan corresponding to the row having the shortest fifth distance.

[0177] The roadside radar position identification device provided in the embodiment of the present invention can execute the roadside radar position identification method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0178] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0179] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A roadside radar position recognition method, characterized in that: include: Obtain radar scanning data of the roadside radar at the intersection, as well as map data of the intersection; generating positions of a plurality of virtual viewpoints based on the map data; Acquire virtual radar parameters based on the radar scanning data; Based on the positions of the plurality of virtual viewpoints and the virtual radar parameters, converting the map data into virtual scans respectively corresponding to the positions of the plurality of virtual viewpoints; respectively extracting features of the radar scan data and a plurality of the virtual scans; Based on the features of the radar scan data and the features of the plurality of virtual scans, calculating first distances between the radar scan data and the plurality of virtual scans respectively; the first distances are used to express the similarity between the radar scan data and the virtual scans; Determine a target virtual viewpoint based on multiple values ​​of the first distances, where the location of the target virtual viewpoint is the location of the roadside radar; The extracting the features of the radar scan data and the plurality of virtual scans respectively includes: The radar scanning data and the plurality of virtually scanned point cloud data are binned according to horizontal direction, horizontal angle and vertical angle; Encode the points in each bin; By accumulating the encoding functions of all boxes in the unified cross section, a feature descriptor of the point cloud is calculated, wherein the feature descriptor includes features corresponding to all boxes along different horizontal directions and different horizontal angles; The encoding of the points in each box includes: Where P i,j,k Display box The set of midpoints, , N r represents the number of boxes along the horizontal direction, N h The number of bins representing the horizontal angle, N v The number of bins representing the vertical angle; Represents the number of points in the corresponding set, N th Indicates the quantity threshold; represents the encoding function, Represents the judgment function.

2. The method for identifying the position of a roadside radar according to claim 1, characterized in that: The generating of the positions of a plurality of virtual viewpoints based on the map data comprises: Obtaining a vector group for each road in the map data; Linear interpolation is performed on each vector in the vector group, and a plurality of virtual viewpoints are generated on each vector.

3. The roadside radar position recognition method according to claim 2, characterized in that: The obtaining of a vector group for each road in the map data comprises: Among them, R i The vector group representing the i-th road, V i,j represents the jth vector in the vector group that makes up the i-th road, n i represents the number of coordinate points on the i-th road, P i,j represents the jth coordinate point on the i-th road; The performing linear interpolation on each vector in the vector group, and generating a plurality of virtual viewpoints on each vector, comprises: Among them, P i (k, j) represents the kth virtual viewpoint generated on the jth vector on the i-th road, m represents the number of virtual viewpoints that are expected to be generated on the jth vector, represents the starting point of the jth vector in the i-th road, Represents the end point of the jth vector in the i-th road.

4. The method for identifying the position of a roadside radar according to claim 1, characterized in that: The acquiring virtual radar parameters based on the radar scanning data comprises: Calculate a second distance, a first horizontal angle, and a first vertical angle from each point in the radar scanning data to the roadside radar; obtaining the number of virtual radar lines based on the number of the first vertical angles; Acquire a measurement range of the virtual radar based on a second distance of each point; Acquire a horizontal angle interval and a vertical angle interval of the virtual radar based on the first horizontal angle and the first vertical angle of each point; Taking the maximum and minimum vertical angle intervals as the vertical field angle of the virtual radar; A vertical angle set of virtual radar laser rays is calculated based on the vertical angle interval and the number of virtual radar lines, and a horizontal angle set of the virtual radar laser rays is calculated based on the horizontal angle interval.

5. The method for identifying the position of a roadside radar according to claim 4, characterized in that: The converting the map data into virtual scans respectively corresponding to the locations of the multiple virtual viewpoints based on the locations of the multiple virtual viewpoints and the virtual radar parameters comprises: Clipping the map data based on the position of the virtual viewpoint and the measurement range of the virtual radar; Calculating a third distance, a second horizontal angle, and a second vertical angle from each point in the cropped map data to the virtual viewpoint; generating a virtual radar scan line from the second horizontal angle and the second vertical angle; If there is a first point whose second horizontal angle belongs to the horizontal angle set, and whose second vertical angle belongs to the vertical angle set, the first point is on the scanning line of the virtual radar, and the first point is retained; The point on the scan line of the virtual radar that is closest to the virtual viewpoint is retained to obtain the virtual scan corresponding to the virtual viewpoint, and then the virtual scans corresponding to the multiple virtual viewpoints are obtained.

6. The roadside radar position recognition method according to claim 1, characterized in that: The feature descriptor of the point cloud is calculated by accumulating the encoding functions of all boxes in the unified cross section, including: F ij represents the features corresponding to all boxes along the i-th horizontal direction and the j-th horizontal angle, and F represents the feature descriptor of the entire point cloud.

7. The roadside radar position recognition method according to claim 1, characterized in that: The calculating, based on the feature of the radar scan data and the features of the plurality of virtual scans, first distances between the radar scan data and the plurality of virtual scans, respectively, comprises: In the formula, and The number of columns representing the radar scan data and the feature descriptors of the plurality of virtual scans respectively; i represents the distance between the virtual scan and the radar scan data after being shifted i columns to the right, dis represents the first distance between the virtual scan and the radar scan data, and g(i, j) represents the index of the jth column in the feature descriptor after being shifted i columns to the right; data of the g(i, j)th column of the feature descriptor representing the virtual scan, The data of the j-th column of the feature descriptor representing the radar scan data.

8. The method for identifying the position of a roadside radar according to claim 7, characterized in that: The determining of a target virtual viewpoint based on the values ​​of the plurality of first distances, wherein the location of the target virtual viewpoint is the location of a roadside radar, comprises: Sort the first distances to obtain top K candidate virtual viewpoints; Divide the candidate virtual viewpoints into multiple clusters, and calculate the fourth distance between the elements in each cluster; A distance matrix is ​​constructed based on the first distance and the fourth distance; each element in the first row of the distance matrix is ​​a first distance between the radar scan data and the virtual scans of the first K candidate virtual viewpoints, and other rows except the first row are fourth distances between any two virtual scans of the first K candidate virtual viewpoints; The fifth distance between each row in the distance matrix and the first row is calculated, and the target virtual viewpoint is determined according to the virtual scan corresponding to the row having the shortest fifth distance.

9. A roadside radar position recognition device, characterized in that: include: A first acquisition unit, used to acquire radar scanning data of a roadside radar at an intersection and map data of the intersection; A generating unit, configured to generate positions of a plurality of virtual viewpoints based on the map data; A second acquisition unit, configured to acquire virtual radar parameters based on the radar scanning data; A conversion unit, configured to convert the map data into virtual scans respectively corresponding to the locations of the multiple virtual viewpoints based on the locations of the multiple virtual viewpoints and the virtual radar parameters; An extraction unit, used to extract the features of the radar scan data and the plurality of virtual scans respectively; a calculation unit, configured to calculate first distances between the radar scan data and the plurality of virtual scans respectively based on the features of the radar scan data and the features of the plurality of virtual scans; the first distances being used to express the similarity between the radar scan data and the virtual scans; a position identification unit, configured to determine a target virtual viewpoint based on a plurality of values ​​of the first distances, wherein the position of the target virtual viewpoint is the position of the roadside radar; The extraction unit is further used to bin the radar scanning data and the plurality of virtual scanned point cloud data according to horizontal direction, horizontal angle and vertical angle; Encode the points in each bin; By accumulating the encoding functions of all boxes in the unified cross section, a feature descriptor of the point cloud is calculated, wherein the feature descriptor includes features corresponding to all boxes along different horizontal directions and different horizontal angles; The encoding of the points in each box includes: Where P i,j,k Display box The set of midpoints, , N r represents the number of boxes along the horizontal direction, N h The number of bins representing the horizontal angle, N v The number of bins representing the vertical angle; Represents the number of points in the corresponding set, N th Indicates the quantity threshold; represents the encoding function, Represents the judgment function.