Processing method and device for road side sensing track evaluation

Abstract

The embodiment of the invention relates to a processing method and device for road side sensing track evaluation. The method comprises the following steps: receiving a first task request; acquiring a drive test trajectory set, a sensing trajectory set and a local high-precision map from the cloud; unifying a coordinate system for all track points of the two track sets; carrying out track cutting on the two track sets; segmenting the road test tracks based on the track point curvature, selecting a matching track set for each segmented track based on a DTW algorithm, merging all the matching track sets corresponding to each road test track, and performing sensing track splicing based on each merged track set according to a time-space continuity rule; calculating the effective sensing track length of each drive test-sensing track pair according to a high-low speed self-adaptive invalid track length identification method, and calculating a corresponding sensing coverage rate; and taking the average value of all the sensing coverage rates as the roadside sensing coverage rate. According to the invention, the evaluation accuracy of the roadside perception capability can be effectively improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a processing method and apparatus for roadside sensing trajectory evaluation. Background Technology

[0002] With the rapid development of vehicle-road cooperative and intelligent transportation systems, roadside units (RSUs) and roadside sensors (such as cameras, millimeter-wave radar, and lidar) play a crucial role in vehicle target detection and trajectory tracking in intersection environments. Accurate assessment of their perception capabilities is fundamental to improving system reliability.

[0003] Under normal circumstances, the technical principle for evaluating the perception capability of an intersection or its Roadside Unit (RSU) is as follows: A road test vehicle (referred to as the road test vehicle) drives through all lanes or all designated areas of the current intersection for one round, collecting high-precision motion trajectories output by the vehicle during the journey and saving the collected results as true reference trajectories (also called road test trajectories). At the same time, all roadside perception trajectories generated at the current intersection are collected. Then, based on the road test-perception trajectories of the same target (the same road test vehicle), the perception coverage is calculated. The higher the coverage, the stronger the perception capability; conversely, the higher the coverage, the weaker the perception capability. The perception coverage is calculated as: (effective length of perception trajectory / true trajectory length of road test trajectory) × 100%.

[0004] It is not difficult to see from the above technical principles that in order to improve the accuracy of roadside perception capability assessment, it is necessary to ensure that the road test trajectory and the perception trajectory can achieve high-precision spatiotemporal alignment, effectively identify and eliminate invalid measurement segments in the perception trajectory, and reliably stitch together multiple perception trajectories when the target is briefly lost. However, current evaluation methods for roadside sensing trajectories still have significant shortcomings in the aforementioned key aspects: 1) Insufficient trajectory alignment accuracy: Existing methods often use hard timestamp alignment or simple nearest neighbor matching, ignoring changes in vehicle motion patterns such as acceleration, deceleration, and turning at intersections, resulting in large trajectory comparison deviations and affecting evaluation accuracy; 2) Difficulty in removing invalid sensing segments: Sensing trajectories often generate local error segments due to occlusion, mistracking, and target jumps. Specifically, this manifests as sudden position changes at high speeds and position jitter at low speeds. Existing methods lack adaptive invalid length recognition mechanisms for different speed scenarios, leading to inaccurate calculation of effective sensing length; 3) Difficulty in stitching together multi-segment sensing trajectories: Roadside devices may generate multiple segmented trajectories due to temporary target loss. Existing methods lack multi-segment trajectory stitching rules based on spatiotemporal continuity, motion consistency, and matching consistency, making it difficult to reconstruct complete trajectories and affecting the complete assessment of coverage.

[0005] To address the aforementioned issues, there is an urgent need for a roadside perception trajectory evaluation method that can achieve flexible trajectory alignment, adaptive invalid segment removal in high and low speed scenarios, and reliable splicing of multiple trajectory segments, so as to more accurately assess the roadside perception capabilities of intersections or intersection RSUs. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing a processing method, apparatus, electronic device, and computer-readable storage medium for roadside sensing trajectory evaluation. This invention calculates the curvature of trajectory points for each roadside trajectory and segments the trajectory based on the calculation results to obtain corresponding segmented trajectory sequences. Using the DTW algorithm, it selects matching trajectory sets from numerous sensing trajectories that intersect with each segmented trajectory in time intervals. It then merges and deduplicates all matching trajectory sets corresponding to each roadside trajectory to obtain a merged trajectory set. Following spatiotemporal continuity rules, it stitches the trajectories based on the merged trajectory sets corresponding to each roadside trajectory to obtain the corresponding sensing stitched trajectory. Finally, it forms corresponding trajectory pairs from each roadside trajectory and the sensing stitched trajectory. Using a high-low speed adaptive invalid trajectory length identification method, it calculates the effective sensing trajectory length of each trajectory pair. Based on the effective sensing trajectory length plus the roadside trajectory length of each trajectory pair, it calculates the corresponding sensing coverage rate and uses the average of all sensing coverage rates as the corresponding roadside sensing coverage rate. This invention employs curvature-based roadside trajectory segmentation and a DTW flexible matching algorithm, improving trajectory alignment accuracy by allowing nonlinear time warping. Furthermore, it utilizes a high-low speed adaptive invalid trajectory length identification method, employing an error time propagation-based elimination strategy for high-speed segments and a grid jitter subtraction strategy for low-speed segments, thereby enhancing the accuracy of invalid sensing segment identification and elimination. Additionally, it leverages a four-fold spatiotemporal continuity rule encompassing space, time, motion, and matching consistency to reliably stitch together multiple sensing trajectories generated due to temporary target loss, achieving complete reconstruction of the target sensing trajectory. This invention not only solves problems such as insufficient trajectory alignment accuracy, difficulty in eliminating invalid sensing segments, and difficulty in stitching together multiple sensing trajectories in conventional solutions, but also effectively improves the assessment accuracy of roadside sensing capabilities.

[0007] To achieve the above objectives, a first aspect of the present invention provides a processing method for roadside sensing trajectory evaluation, the method comprising:

[0008] The received roadside perception trajectory evaluation task request is recorded as the corresponding first task request; the first task request includes a first task identifier, a first evaluation time period, and a first intersection identifier;

[0009] Based on the first evaluation period and the first intersection signage requested by the first task, the road test trajectory set, perception trajectory set, and local high-precision map for the current intersection and current period are obtained from the preset road test trajectory cloud, roadside perception cloud, and high-precision map cloud; the coordinates of all trajectory points in the two trajectory sets are transformed to the high-precision map coordinate system corresponding to the local high-precision map; and the two trajectory sets are clipped according to the local high-precision map based on the preset trajectory clipping mode; the road test trajectory set includes multiple road test trajectories; the perception trajectory set includes multiple perception trajectories; the map elements of the local high-precision map include ROI elements, lane line elements, and stop line elements;

[0010] The curvature of the trajectory points of each road test trajectory is calculated, and the trajectory is segmented based on the calculation results to obtain the corresponding segmented trajectory sequence. A set of trajectory to be matched is formed from the sensing trajectories in the sensing trajectory set that intersect with each segmented trajectory in terms of trajectory time intervals. A matching trajectory set is selected from the corresponding set of trajectory to be matched for each segmented trajectory based on the DTW algorithm. Trajectories are merged and deduplicated for all matching trajectory sets corresponding to each segmented trajectory sequence to obtain the corresponding merged trajectory set. Finally, the trajectory is stitched together based on each merged trajectory set according to a preset spatiotemporal continuity rule to obtain the corresponding sensing stitched trajectory.

[0011] Each road test trajectory and its corresponding sensing stitched trajectory form a corresponding road test-sensing trajectory pair; the effective sensing trajectory length of each road test-sensing trajectory pair is calculated according to the high-low speed adaptive invalid trajectory length identification method; the corresponding sensing coverage rate is calculated based on the effective sensing trajectory length and road test trajectory length of each road test-sensing trajectory pair; and the average value of all the sensing coverage rates is taken as the corresponding roadside sensing coverage rate.

[0012] Each of the roadside perception trajectory pairs and its corresponding perception coverage rate constitutes a corresponding roadside perception evaluation item; all the roadside perception evaluation items constitute a corresponding roadside perception evaluation table; and the first task request, the roadside perception evaluation table, and the roadside perception coverage rate constitute a corresponding roadside perception evaluation report and are saved.

[0013] Preferably, the road test trajectory cloud stores multiple first intersection trajectory sets in the cloud; each first intersection trajectory set corresponds to an intersection identifier; each first intersection trajectory set consists of multiple first intersection trajectories; each first intersection trajectory is a segment of vehicle motion trajectory collected when a road test vehicle passes through the current intersection within the current trajectory time period; the trajectory information of each first intersection trajectory consists of the vehicle identifier of the current road test vehicle, the unique trajectory identifier of the current trajectory, and the trajectory point sequence of the current trajectory; each trajectory point of the first intersection trajectory corresponds to a set of trajectory point identifiers, trajectory point coordinates, timestamps, speeds, and heading angles;

[0014] The roadside sensing cloud stores multiple second intersection trajectory sets in the cloud; each second intersection trajectory set corresponds to an intersection identifier; each second intersection trajectory set consists of multiple second intersection trajectories; each second intersection trajectory is a target sensing trajectory obtained by the roadside unit device at the current intersection through roadside target sensing technology within the current trajectory time period; the trajectory information of each second intersection trajectory consists of the unique target identifier of the currently sensed target and the trajectory point sequence of the current trajectory; each trajectory point of the second intersection trajectory corresponds to a set of trajectory point identifiers, trajectory point coordinates, timestamps, speeds, and heading angles;

[0015] The high-precision map cloud stores high-precision maps of all intersections in the cloud; the map coverage of each high-precision map of an intersection includes not only all the internal areas of the current intersection, but also a section of the outer edge of each road leading into or out of the intersection; the outer edge length of each outer edge road is a preset first length threshold.

[0016] The road test trajectory includes multiple first trajectory points; each first trajectory point includes a trajectory point identifier, trajectory point coordinates, timestamp, speed, and heading angle.

[0017] The sensing trajectory includes multiple second trajectory points; each second trajectory point includes a trajectory point identifier, trajectory point coordinates, timestamp, speed, and heading angle.

[0018] The number of the ROI element, the lane line element, or the stop line element is greater than or equal to 1;

[0019] Each ROI element corresponds to a region of interest within the current intersection; the ROI element includes the region identifier, shape type, and geometric dimension parameters of the corresponding region; the shape type includes polygon and circle, and when it is a polygon, the corresponding geometric dimension parameters are the set of vertex coordinates of the current region, and when it is a circle, the corresponding geometric dimension parameters include the coordinates of the center point of the current region and the region radius;

[0020] Each lane line element corresponds to a driving lane entering or exiting the current intersection, or a driving lane within the current intersection; the lane line element includes a lane identifier and a lane line coordinate set; the lane line coordinate set consists of multiple lane line point coordinates;

[0021] Each stop line element corresponds to a stop line at the entry or exit point of the current intersection, or a stop line within the current intersection; each stop line element includes a stop line identifier and a stop line coordinate set; the stop line coordinate set consists of multiple stop line point coordinates.

[0022] The segmented trajectory sequence corresponds one-to-one with the road test trajectory and is formed by sequentially sorting one or more segmented trajectories; each segmented trajectory corresponds to a segment type; the segment type includes straight, left turn, and right turn; each segmented trajectory includes multiple segmented trajectory points; each segmented trajectory point includes trajectory point identifier, trajectory point coordinates, timestamp, speed, and heading angle;

[0023] The sensing and stitching trajectory includes multiple third trajectory points; each third trajectory point includes a trajectory point identifier, trajectory point coordinates, timestamp, speed, and heading angle.

[0024] Preferably, the step of obtaining the current intersection's current time period, the perception trajectory set, and the local high-precision map from the preset road test trajectory cloud, roadside perception cloud, and high-precision map cloud based on the first evaluation period and the first intersection signage in the first task request specifically includes:

[0025] The first and second intersection trajectories on the road test trajectory cloud and the roadside perception cloud that have a time intersection with the first evaluation period corresponding to the first intersection sign are respectively recorded as the first and second candidate trajectories; the trajectory points in each of the first and second candidate trajectories that are within the first evaluation period are extracted to form the corresponding road test trajectory and the perception trajectory; and the corresponding road test trajectory set and perception trajectory set are formed by all the obtained road test trajectories and all the obtained perception trajectories; and the intersection high-precision map on the high-precision map cloud that corresponds to the first intersection sign is used as the corresponding local high-precision map.

[0026] Preferably, the trajectory clipping mode includes a first mode, a second mode, and a third mode;

[0027] When the trajectory clipping mode is the first mode, the overall clipping boundary is the boundary of the combined area of ​​the current intersection and all its corresponding entrance and exit outer roads.

[0028] When the trajectory clipping mode is the second mode, the boundaries of each region of interest at the current intersection are used as a corresponding local clipping boundary.

[0029] When the trajectory clipping mode is the third mode, the area boundary of the outer edge of the entrance and exit roads outside each stop line of the current intersection is used as a corresponding local clipping boundary.

[0030] Preferably, the trajectory clipping based on the preset trajectory clipping mode performs trajectory clipping on the two trajectory sets according to the local high-precision map, specifically including:

[0031] The trajectory clipping pattern is identified;

[0032] If the trajectory clipping mode is the first mode, then the boundary of the overall joint area of ​​the current intersection and all its corresponding entrance and exit outer roads is identified according to the local high-precision map, and the identification result is used as the corresponding overall clipping boundary; then all the road test trajectories are traversed once, and during this round of traversal, the trajectory points of the currently traversed road test trajectory outside the overall clipping boundary are clipped to obtain a new road test trajectory; then all the perception trajectories are traversed once, and during this round of traversal, the trajectory points of the currently traversed perception trajectory outside the overall clipping boundary are clipped to obtain a new perception trajectory;

[0033] If the trajectory clipping mode is the second mode or the third mode, then when the current mode is the second mode, the boundary of the region of interest corresponding to each ROI element of the local high-precision map is used as a corresponding local clipping boundary; when the current mode is the third mode, the boundary of the area of ​​the outer edge of the entrance and exit road outside the stop line corresponding to each stop line element of the local high-precision map is used as a corresponding local clipping boundary; and a round of traversal is performed on all the road test trajectories, and during this round of traversal, the sub-trajectories of the currently traversed road test trajectory within each of the local clipping boundaries are extracted as a new road test trajectory; and a round of traversal is performed on all the perception trajectories, and during this round of traversal, the sub-trajectories of the currently traversed perception trajectory within each of the local clipping boundaries are extracted as a new perception trajectory;

[0034] After the two rounds of traversal are completed, a new set of road test trajectories is formed by all the new road test trajectories, and a new set of perception trajectories is formed by all the new perception trajectories.

[0035] Preferably, the step of calculating the curvature of the trajectory points of each of the road test trajectories and segmenting the trajectory based on the calculation results to obtain the corresponding segmented trajectory sequence specifically includes:

[0036] Step 61: Count the total number Q of the trajectory points of the road test trajectory;

[0037] Step 62, based on the timestamp t of each of the first trajectory points of the road test trajectory. q Speed ​​v q Heading angle θ q Calculate the corresponding curvature k q Based on a preset straight curvature threshold range, and according to each curvature k... q Determine the corresponding trajectory direction type s q ;

[0038] Where 1 ≤ index q ≤ Q;

[0039] The curvature k q The calculation method is as follows:

[0040]

[0041] The straight curvature threshold range is [k] min ,k max ], k min k max For the corresponding minimum and maximum boundary thresholds; k min <0<k max ;

[0042] The trajectory direction type s q Including left turn, straight ahead, and right turn, the method for determining them is as follows:

[0043]

[0044] Step 63: Divide each segment of the road test trajectory into U or more consecutive trajectory direction types s. q A sequence of trajectory points consisting of the first trajectory points that are all left turns, straight ahead, or right turns is a segmented trajectory whose segmentation type is left turn, straight ahead, or right turn; and all the obtained segmented trajectories are sorted in chronological order to form the corresponding segmented trajectory sequence;

[0045] The preset minimum total number of continuous trajectory points U is a positive integer.

[0046] Preferably, the step of selecting a matching trajectory set from the corresponding set of trajectories to be matched for each segmented trajectory based on the DTW algorithm specifically includes:

[0047] Step 71: Record the segmented trajectory as the corresponding trajectory X; record each of the perceived trajectories in the trajectory set to be matched as the corresponding trajectory Y; and count the total number of trajectory points of trajectory X and trajectory Y to obtain the corresponding total number N and M respectively; and construct a cumulative cost matrix D of shape N×M for trajectory X and trajectory Y.

[0048] Wherein, each segment trajectory point of the trajectory X is denoted as trajectory point x. i Each of the aforementioned trajectory points x i The trajectory point coordinates are marked as the corresponding position coordinates. 1 ≤ index i ≤ N; each of the second trajectory points of the trajectory Y is denoted as trajectory point y. j Each of the said trajectory points y j The trajectory point coordinates are marked as the corresponding position coordinates. 1 ≤ index j ≤ M;

[0049] The rows of the cumulative cost matrix D and the trajectory point x i One-to-one correspondence, matrix columns and the trajectory points y j One-to-one correspondence; the starting position of both the row index and column index is the lower left vertex cell position; the cumulative cost matrix D consists of N×M matrix cells (i,j), and the cell data of each matrix cell (i,j) is the corresponding cumulative cost d. i,j ;

[0050] Step 72: Based on the DTW algorithm, calculate the cumulative cost matrix for all the cumulative costs d in the cumulative cost matrix D. i,j The calculation is as follows:

[0051] First, calculate the cumulative cost d of the matrix element at the bottom left vertex. i＝1,j＝1 :

[0052]

[0053] Next, calculate the cumulative cost d for the remaining M-1 rows of the matrix, excluding the bottom-left vertex. i＝1,1<j≤M And calculate the cumulative cost d of the remaining N-1 columns of the matrix, excluding the bottom left vertex. 1<i≤N,j＝1 :

[0054]

[0055] Next, calculate the cumulative cost d of the remaining (N-1)×(M-1) matrix elements. 2≤i≤N,2≤j≤M :

[0056]

[0057] Among them, f dis () is the Euclidean distance calculation function, used to calculate the Euclidean distance between two position coordinates input to the function;

[0058] Step 73, the cumulative cost d corresponding to the upper right vertex matrix element (N,M) in the cumulative cost matrix D is... i＝N,j＝MThe DTW distance between the trajectory X and the current trajectory Y;

[0059] Step 74: Based on the regularized path backtracking method of the DTW algorithm, backtrack the path points from the upper right vertex matrix unit (N,M) to the lower left vertex matrix unit (1,1) and use the reverse path of the backtracked path as the corresponding regularized path W. Specifically:

[0060] Step 741: Initialize the backtracking point index b to 1; and set the upper right corner vertex matrix unit (N, M) as the current matrix unit (i). now ,j now );

[0061] Among them, i now j now These are the row index and column index of the current cell, respectively;

[0062] Step 742, set the current matrix unit (i) now ,j now Mark it as a corresponding backtracking point. And for the current matrix unit (i) now ,j now If the left-lower vertex matrix unit (1,1) is identified, proceed to step 744; otherwise, proceed to step 743.

[0063] Step 743, set the current matrix unit (i) now ,j now The left neighbor matrix element (i) now ,j now -1), Lower neighbor matrix unit (i now -1,j now ), left-sloping lower matrix unit (i now -1,j now -1) corresponds to the three cumulative costs The minimum value in the matrix is ​​taken as the current minimum cost; and the matrix element corresponding to the current minimum cost is taken as the new current matrix element (i). now ,j now ); and increment the backtracking point index b by 1; and return to step 742;

[0064] Step 744: Use the index value of the current backtracking point index b as the corresponding path length L; and set each backtracking point... Let w be the corresponding path point. l＝L-b+1 And from the obtained L path points w l The sequential arrangement forms the corresponding regular path W;

[0065] The regularized path W consists of multiple path points w l Composition, 1 ≤ index l ≤ L; each of the path points w l A corresponding matrix element is denoted as the corresponding matrix element (i). l ,j l ), i l j l The path points w are respectively l The row and column indices of the corresponding matrix elements;

[0066] Step 75: Based on the regularized path W of each trajectory Y, calculate the trajectory error E between the trajectory X and each trajectory Y, specifically as follows:

[0067] The regularized path W of each trajectory Y is taken as the current regularized path; and for all trajectory points x of the trajectory X... i Perform one round of traversal; and during this round of traversal, change the current traversed trajectory point x. i And its corresponding index i is denoted as the current road test point and the current index i. * ; and in the current normalized path, the row index i of the corresponding matrix unit l With the current index i * Equal path points w l All are used as corresponding alignment points; and the index j in the current trajectory Y is compared with the column index j corresponding to each alignment point. l The matched trajectory point y j All are denoted as aligned sensing points; the Euclidean distance between the current road test point and each of the aligned sensing points is calculated, and the calculation result is used as the corresponding road test-sensing point error; the mean of all the road test-sensing point errors corresponding to the current road test point is calculated to obtain the corresponding road test point mean error; at the end of this round of traversal, the mean of all the road test point mean errors obtained in this round of traversal is calculated, and the calculation result is used as the trajectory error E corresponding to the current trajectory Y;

[0068] Step 76: A first data group is formed by the DTW distance and the trajectory error E corresponding to each trajectory Y; and the first data group whose DTW distance is less than a preset first distance threshold and whose trajectory error E is less than a preset first error threshold is recorded as the second data group.

[0069] Among them, the first data group corresponding to the current segmented trajectory corresponds one-to-one with the sensing trajectory in the set of trajectory to be matched corresponding to the current segmented trajectory;

[0070] Step 77: Count the total number of data groups in the second data group; if the total number of data groups is zero, set the corresponding matching trajectory set to empty; if the total number of data groups is greater than zero, then the corresponding matching trajectory set is composed of all the sensing trajectories corresponding to the second data group.

[0071] Preferably, the step of stitching together the trajectories based on each of the merged trajectory sets according to a preset spatiotemporal continuity rule to obtain the corresponding perceptual stitched trajectory specifically includes:

[0072] Step 81: Take each of the merged trajectory sets as the current merged trajectory set;

[0073] Step 82: Identify whether the current merged trajectory set is empty; if yes, set the corresponding perceived stitching trajectory to empty and proceed to step 87; if no, proceed to step 83.

[0074] Step 83: Initialize an empty trajectory sequence for the current merged trajectory set as the corresponding first sequence; add the sensing trajectory with the earliest start time in the current merged trajectory set to the first sequence; and after successful addition, delete the sensing trajectory added at this time from the current merged trajectory set.

[0075] Step 84: Take the last sensing trajectory in the first sequence as the current tail trajectory;

[0076] Step 85: According to the spatiotemporal continuity rule, select a sensing trajectory from the current merged trajectory set that satisfies the spatiotemporal continuity relationship with the current tail trajectory as the corresponding next segment trajectory;

[0077] The spatiotemporal continuity rule stipulates that the next segment of trajectory that satisfies the spatiotemporal continuity relationship with the current tail trajectory should simultaneously satisfy the preset spatial continuity sub-rule, temporal continuity sub-rule, motion continuity sub-rule, and matching consistency sub-rule.

[0078] The spatial continuity sub-rule requires that the Euclidean distance between the last second trajectory point of the current tail trajectory and the first second trajectory point of the next segment trajectory should be lower than a preset second distance threshold.

[0079] The time continuity sub-rule requires that: the time of the last second trajectory point of the current tail trajectory should be earlier than the time of the first second trajectory point of the next segment trajectory, and the time interval between the first second trajectory point of the next segment trajectory and the last second trajectory point of the current tail trajectory should be less than a preset first duration threshold.

[0080] The motion continuity sub-rule requires that: the uniformly accelerated motion model be based on the timestamp t of the last second trajectory point of the current tail trajectory. pre Speed ​​v pre And the timestamp t of the first second trajectory point of the next trajectory segment nxt Speed ​​v nxt Calculate the corresponding acceleration a = (v nxt -v pre ) / (t nxt -t pre The absolute value of the acceleration a, |a|, should be less than a preset threshold value for absolute acceleration.

[0081] The matching consistency sub-rule requires that: the segmented trajectories that match the current tail trajectory and the next segment trajectory are recorded as the corresponding first segment trajectory and second segment trajectory; if the first and second segment trajectories are different, the current matching consistency sub-rule is assumed to be satisfied; if the first and second segment trajectories are the same, the absolute difference |ΔE| between the two trajectories E corresponding to the current tail trajectory, the next segment trajectory and the current segment trajectory should be less than a preset absolute difference threshold.

[0082] If the next trajectory segment is empty, it means that there is no sensing trajectory in the current merged trajectory set that can simultaneously satisfy the four sub-rules of the spatiotemporal continuity rule.

[0083] Step 86: Identify whether the next segment of the trajectory is empty; if yes, then sequentially splice all the sensing trajectories in the first sequence into the corresponding sensing spliced ​​trajectory, and proceed to step 87; if no, then add the current next segment of the trajectory to the first sequence, and after successful addition, delete the current next segment of the trajectory from the current merged trajectory set, and return to step 84.

[0084] Step 87: Output the obtained perception stitching trajectory as the result of this trajectory stitching.

[0085] Preferably, the method for identifying invalid trajectory lengths based on high and low speed adaptive calculations specifically includes:

[0086] Step 91: Take each of the aforementioned road test-sensing trajectory pairs as the current trajectory pair;

[0087] Step 92: Take the road test trajectory and the perception stitching trajectory of the current trajectory pair as the corresponding current road test trajectory and current perception trajectory; and identify whether the current perception trajectory is empty; if yes, set the length of the corresponding effective perception trajectory to 0 and go to step 99; if no, go to step 93.

[0088] Step 93: Perform an initial calculation on the trajectory length of the current perceived trajectory to obtain the corresponding initial trajectory length, specifically as follows:

[0089] The trajectory travel distance between every two third trajectory points of the current sensing trajectory is estimated to obtain the corresponding first distance; and the sum of all the first distances is used as the corresponding initial trajectory length.

[0090] Step 94: Take the first trajectory point on the current road test trajectory that is closest to each of the third trajectory points on the current sensing trajectory as the pairing point of the current third trajectory point, and calculate the Euclidean distance between each third trajectory point and its corresponding pairing point, and use the calculation result as the corresponding pairing point error;

[0091] Step 95: Based on preset high-speed and low-speed thresholds, identify high-speed and low-speed segments on the current sensing trajectory, specifically as follows:

[0092] The third trajectory point on the current sensing trajectory with a speed greater than the high-speed threshold is recorded as a high-speed point, and the third trajectory point with a speed less than the low-speed threshold is recorded as a low-speed point; and each trajectory segment composed of G or more consecutive high-speed points is recorded as a corresponding high-speed segment, and each trajectory segment composed of H or more consecutive low-speed points is recorded as a corresponding low-speed segment.

[0093] Where 0 < low speed threshold < high speed threshold; the preset total number of continuous trajectory points G and H are each a positive integer;

[0094] Step 96: Identify large-error invalid time periods based on all the high-speed segments and calculate the corresponding first invalid length based on the identification results, specifically as follows:

[0095] A traversal is performed on all the high-speed segments; during this traversal, the currently traversed high-speed segment is taken as the current segment; and a poll is performed on all the third trajectory points of the current segment, with the currently polled third trajectory point taken as the current point. The pairing point error corresponding to the current point is checked to see if it exceeds a preset third distance threshold. If so, the timestamp of the current point is taken as the corresponding current time t. mid And based on the preset duration r, a corresponding invalid time period [t] is set. mid -r,t mid+r]; and at the end of this round of traversal, merge all the invalid time periods obtained to obtain one or more merged time periods; mark the third trajectory points on the current sensing trajectory that are in any of the merged time periods as invalid points; and take the first distance between every two adjacent invalid points as a corresponding second distance; and take the sum of all the second distances as the corresponding first invalid length;

[0096] Step 97: Based on all the low-speed segments, calculate the length of the invalid trajectory with large errors and adjust the length of the valid trajectory with small errors to obtain the corresponding second invalid length and first adjusted length, specifically as follows:

[0097] Step 971: Perform a traversal of all the low-speed segments; during this traversal, the currently traversed low-speed segment is taken as the current segment; the rectangular area of ​​the road traversed by the current segment on the local high-precision map is taken as the current region; the current region is gridded according to a preset square grid size to obtain multiple corresponding region grids; the region grids that intersect with the trajectory of the current segment are recorded as the corresponding passing grids; the average error of the pairing points of all the third trajectory points of each passing grid is calculated and the calculation result is taken as the corresponding grid error; and it is identified whether the grid error of each passing grid is greater than the third distance threshold. If it is, the current grid is recorded as an invalid grid; otherwise, the current grid is recorded as a valid grid.

[0098] Step 972: After this round of traversal, the first distance between each pair of adjacent third trajectory points in each invalid grid is summed and the result is used as the corresponding invalid grid trajectory length. The lengths of all the obtained invalid grid trajectories are summed and the result is used as the corresponding second invalid length.

[0099] Step 973: Sum the first distances between adjacent third trajectory points in each of the effective grids and use the result as the corresponding original grid trajectory length; use the side length of the square grid of the square grid size as the unified network trajectory length; set the corresponding grid optimized trajectory length = max(0, original grid trajectory length - unified network trajectory length) based on each of the original grid trajectory lengths and the unified network trajectory length; and sum all the obtained grid optimized trajectory lengths and use the result as the corresponding first adjustment length.

[0100] Step 98: The effective sensing trajectory length is obtained by subtracting the sum of the first invalid length, the second invalid length, and the first adjusted length from the initial trajectory length;

[0101] Wherein, the effective sensing trajectory length = initial trajectory length - (first invalid length + second invalid length + first adjustment length);

[0102] Step 99: The obtained effective sensing trajectory length is output as the result of this calculation.

[0103] Preferably, the step of calculating the corresponding sensing coverage rate based on the effective sensing trajectory length and the driving trajectory length of each of the drive-sensing trajectory pairs specifically includes:

[0104] Each of the road-sense trajectory pairs is taken as the current trajectory pair; the road-sense trajectory of the current trajectory pair is taken as the corresponding current trajectory; the trajectory driving distance between every two first trajectory points of the current trajectory is estimated to obtain the corresponding third distance; the sum of all the third distances is taken as the corresponding road-sense trajectory length; and the corresponding sensing coverage rate is calculated, sensing coverage rate = (effective sensing trajectory length / road-sense trajectory length) × 100%.

[0105] A second aspect of the present invention provides an apparatus for implementing the processing method for roadside sensing trajectory evaluation described in the first aspect above, the apparatus comprising: an evaluation task receiving module, a data preparation module, a trajectory processing module, a coverage evaluation module, and an evaluation report processing module;

[0106] The evaluation task receiving module is used to receive roadside perception trajectory evaluation task requests, denoted as the corresponding first task request; the first task request includes a first task identifier, a first evaluation time period, and a first intersection identifier;

[0107] The data preparation module obtains the current intersection's current time period, perception trajectory set, and local high-precision map from the preset road test trajectory cloud, roadside perception cloud, and high-precision map cloud based on the first evaluation period and the first intersection identifier requested by the first task. It then converts the coordinates of all trajectory points in both trajectory sets to the high-precision map coordinate system corresponding to the local high-precision map. Based on a preset trajectory clipping mode, it performs trajectory clipping on both trajectory sets according to the local high-precision map. The road test trajectory set includes multiple road test trajectories; the perception trajectory set includes multiple perception trajectories; and the local high-precision map includes ROI elements, lane line elements, and stop line elements.

[0108] The trajectory processing module is used to calculate the curvature of the trajectory points of each road test trajectory and segment the trajectory based on the calculation results to obtain the corresponding segmented trajectory sequence; and form a corresponding set of unmatched trajectories by the sensing trajectories in the sensing trajectory set that have trajectory time intersections with each segmented trajectory; and select the corresponding matching trajectory set for each segmented trajectory from its corresponding set of unmatched trajectories based on the DTW algorithm; and perform trajectory merging and deduplication processing on all the matching trajectory sets corresponding to each segmented trajectory sequence to obtain the corresponding merged trajectory set; and perform trajectory splicing based on each merged trajectory set according to a preset spatiotemporal continuity rule to obtain the corresponding sensing spliced ​​trajectory;

[0109] The coverage assessment module is used to form corresponding road-sensor trajectory pairs from each of the road-test trajectories and their corresponding sensing splicing trajectories; calculate the effective sensing trajectory length of each of the road-test-sensor trajectory pairs according to the high-low speed adaptive invalid trajectory length identification method; calculate the corresponding sensing coverage rate based on the effective sensing trajectory length and road-test trajectory length of each of the road-test-sensor trajectory pairs; and take the average of all the sensing coverage rates as the corresponding roadside sensing coverage rate.

[0110] The evaluation report processing module is used to form a corresponding roadside perception evaluation item by each of the road test-perception trajectory pairs and their corresponding perception coverage rates; and to form a corresponding roadside perception evaluation table by all the roadside perception evaluation items; and to form and save a corresponding roadside perception evaluation report by the first task request, the roadside perception evaluation table and the roadside perception coverage rate.

[0111] A third aspect of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0112] The processor is used to couple with the memory, read and execute instructions in the memory to implement the steps of the method described in the first aspect above;

[0113] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0114] A fourth aspect of the present invention provides a computer-readable storage medium storing computer instructions that, when executed by a computer, cause the computer to perform the instructions described in the first aspect.

[0115] This invention provides a processing method, apparatus, electronic device, and computer-readable storage medium for roadside sensing trajectory evaluation. As described above, this invention calculates the curvature of trajectory points for each roadside trajectory and segments the trajectory based on the calculation results to obtain corresponding segmented trajectory sequences. Then, based on the DTW algorithm, it selects corresponding matching trajectory sets from numerous sensing trajectories that intersect with each segmented trajectory in time periods. Finally, it merges and deduplicates all matching trajectory sets corresponding to each roadside trajectory to obtain a merged trajectory set. Following spatiotemporal continuity rules, it splices the trajectories based on the merged trajectory sets corresponding to each roadside trajectory to obtain the corresponding sensing spliced ​​trajectory. Next, each roadside trajectory plus the sensing spliced ​​trajectory forms a corresponding trajectory pair. The effective sensing trajectory length of each trajectory pair is calculated using a high-low speed adaptive invalid trajectory length identification method. The corresponding sensing coverage rate is calculated based on the effective sensing trajectory length plus the roadside trajectory length of each trajectory pair, and the average of all sensing coverage rates is taken as the corresponding roadside sensing coverage rate. This invention employs a curvature-based roadside trajectory segmentation and DTW flexible matching algorithm, improving trajectory alignment accuracy by allowing nonlinear time warping. Furthermore, it utilizes a high-low speed adaptive invalid trajectory length identification method, employing an error time propagation-based elimination strategy for high-speed segments and a grid jitter subtraction strategy for low-speed segments, thereby enhancing the accuracy of invalid sensing segment identification and elimination. Additionally, it leverages a four-fold spatiotemporal continuity rule encompassing space, time, motion, and matching consistency to reliably stitch together multiple sensing trajectories generated due to temporary target loss, achieving complete reconstruction of the target sensing trajectory. This invention not only solves the problems of insufficient trajectory alignment accuracy, difficulty in eliminating invalid sensing segments, and difficulty in stitching together multiple sensing trajectories in conventional solutions, but also effectively improves the assessment accuracy of roadside sensing capabilities. Attached Figure Description

[0116] Figure 1 This is a schematic diagram of a processing method for roadside sensing trajectory evaluation provided in Embodiment 1 of the present invention;

[0117] Figure 2 This is a schematic diagram of three trajectory clipping modes provided in Embodiment 1 of the present invention;

[0118] Figure 3 This is a schematic diagram of the cumulative cost matrix and the normalized path provided in Embodiment 1 of the present invention;

[0119] Figure 4 This is a module structure diagram of a processing device for roadside sensing trajectory evaluation provided in Embodiment 2 of the present invention;

[0120] Figure 5 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. Detailed Implementation

[0121] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0122] Embodiment 1 of the present invention provides a processing method for roadside sensing trajectory evaluation, such as... Figure 1 The schematic diagram of a processing method for roadside sensing trajectory evaluation provided in Embodiment 1 of the present invention includes the following main steps:

[0123] Step 1: Receive the roadside perception trajectory evaluation task request and record it as the corresponding first task request.

[0124] Here, the first task request in this embodiment of the invention includes a first task identifier, a first evaluation period, and a first intersection identifier. The sender of the first task request in this embodiment of the invention can be any sender, such as an administrator, or other service interfaces, service platforms, or systems. The purpose of the first task request is to request an evaluation of the roadside perception capability of the intersection corresponding to the current intersection identifier. The data required for the evaluation is obtained from three preset cloud environments. The three preset cloud environments in this embodiment of the invention are a road test trajectory cloud, a roadside perception cloud, and a high-precision map cloud.

[0125] The road test trajectory cloud stores multiple first intersection trajectory sets in the cloud; each first intersection trajectory set corresponds to an intersection identifier; each first intersection trajectory set consists of multiple first intersection trajectories; each first intersection trajectory is a segment of vehicle motion trajectory collected when a road test vehicle passes through the current intersection within the current trajectory time period; the trajectory information of each first intersection trajectory consists of the vehicle identifier of the current road test vehicle, the unique trajectory identifier of the current trajectory, and the trajectory point sequence of the current trajectory; each trajectory point of the first intersection trajectory corresponds to a set of trajectory point identifiers, trajectory point coordinates, timestamps, speeds, and heading angles.

[0126] The roadside sensing cloud stores multiple second intersection trajectory sets in the cloud; each second intersection trajectory set corresponds to an intersection identifier; each second intersection trajectory set consists of multiple second intersection trajectories; each second intersection trajectory is a target perception trajectory obtained by the roadside unit device at the current intersection through roadside target perception technology within the current trajectory time period; the trajectory information of each second intersection trajectory consists of the unique target identifier of the currently perceived target and the trajectory point sequence of the current trajectory; each trajectory point of the second intersection trajectory corresponds to a set of trajectory point identifiers, trajectory point coordinates, timestamps, speeds, and heading angles.

[0127] The high-precision map cloud stores high-precision maps of all intersections in the cloud. The map coverage of each intersection's high-precision map includes not only the entire internal area of ​​the current intersection but also a section of the outer edge of each road leading into or out of the intersection. The outer edge length of each outer edge road is a preset first length threshold. Here, in this embodiment of the invention, the first length threshold is a preset length threshold, for example, 90 meters.

[0128] Step 2: Based on the first evaluation period and the first intersection identifier requested by the first task, obtain the current intersection's current period's road test trajectory set, perception trajectory set, and local high-precision map from the preset road test trajectory cloud, roadside perception cloud, and high-precision map cloud; transform the coordinates of all trajectory points in the two trajectory sets to the high-precision map coordinate system corresponding to the local high-precision map; and perform trajectory clipping on the two trajectory sets according to the local high-precision map based on the preset trajectory clipping mode.

[0129] Here, the road test trajectory set in this embodiment of the invention includes multiple road test trajectories; each road test trajectory includes multiple first trajectory points. Each first trajectory point includes a trajectory point identifier, trajectory point coordinates, timestamp, speed, and heading angle.

[0130] The sensing trajectory set in this embodiment of the invention includes multiple sensing trajectories; each sensing trajectory includes multiple second trajectory points. Each second trajectory point includes a trajectory point identifier, trajectory point coordinates, a timestamp, speed, and heading angle.

[0131] The map elements of the local high-precision map in this embodiment of the invention include ROI elements, lane line elements, and stop line elements. The number of ROI elements, lane line elements, or stop line elements is greater than or equal to 1.

[0132] In this embodiment of the invention, each ROI element corresponds to a region of interest within the current intersection. The ROI element includes the region identifier, shape type, and geometric dimension parameters of the corresponding region. The shape type includes polygons and circles. When it is a polygon, the corresponding geometric dimension parameters are the set of vertex coordinates of the current region. When it is a circle, the corresponding geometric dimension parameters include the coordinates of the center point of the current region and the region radius.

[0133] In this embodiment of the invention, each lane line element corresponds to a driving lane entering or exiting the current intersection, or a driving lane within the current intersection; the lane line element includes a lane identifier and a lane line coordinate set; the lane line coordinate set consists of multiple lane line point coordinates.

[0134] Each stop line element in this embodiment of the invention corresponds to a stop line at the entry or exit position of the current intersection, or a stop line within the current intersection; the stop line element includes a stop line identifier and a stop line coordinate set; the stop line coordinate set consists of multiple stop line point coordinates.

[0135] The current step 2 specifically includes:

[0136] Step 21: Based on the first evaluation period and the first intersection identifier requested by the first task, obtain the current intersection's current period's road test trajectory set, perception trajectory set, and local high-precision map from the preset road test trajectory cloud, roadside perception cloud, and high-precision map cloud.

[0137] Step 21 specifically includes: recording the first and second intersection trajectories on the road test trajectory cloud and roadside perception cloud that have a time overlap with the first evaluation period as the corresponding first and second candidate trajectories; extracting the trajectory points in each of the first and second candidate trajectories that are within the first evaluation period to form the corresponding road test trajectory and perception trajectory; and forming the corresponding road test trajectory set and perception trajectory set by all the obtained road test trajectories and all the obtained perception trajectories; and using the high-precision map on the high-precision map cloud that corresponds to the first intersection sign as the corresponding local high-precision map.

[0138] Step 22, and convert the coordinates of all trajectory points in the two trajectory sets to the coordinate system of the high-precision map corresponding to the local high-precision map.

[0139] Step 23, and based on the preset trajectory clipping mode, perform trajectory clipping on the two trajectory sets respectively according to the local high-precision map.

[0140] Here, the trajectory clipping modes in this embodiment of the invention include a first mode, a second mode, and a third mode; when the trajectory clipping mode is the first mode, the overall combined region boundary of the current intersection and all its corresponding entrance and exit outer edges is used as the overall clipping boundary; when the trajectory clipping mode is the second mode, the boundary of each region of interest of the current intersection is used as a corresponding local clipping boundary; when the trajectory clipping mode is the third mode, the region boundary of the entrance and exit outer edges outside each stop line of the current intersection is used as a corresponding local clipping boundary; as shown... Figure 2 The diagram shows three modes of trajectory clipping provided in Embodiment 1 of the present invention.

[0141] Step 23 specifically includes:

[0142] Step 231: Identify the trajectory clipping pattern;

[0143] Step 232: If the trajectory clipping mode is the first mode, then the boundary of the overall joint area of ​​the current intersection and all its corresponding entrance and exit outer roads is identified based on the local high-precision map, and the identification result is used as the corresponding overall clipping boundary; then all road test trajectories are traversed once, and during this round of traversal, the trajectory points of the currently traversed road test trajectory outside the overall clipping boundary are clipped to obtain a new road test trajectory; then all perception trajectories are traversed once, and during this round of traversal, the trajectory points of the currently traversed perception trajectory outside the overall clipping boundary are clipped to obtain a new perception trajectory;

[0144] Step 233: If the trajectory clipping mode is the second mode or the third mode, then when the current mode is the second mode, the boundary of the region of interest corresponding to each ROI element of the local high-precision map is used as a corresponding local clipping boundary; when the current mode is the third mode, the boundary of the area outside the stop line of the entrance and exit road corresponding to each stop line element of the local high-precision map is used as a corresponding local clipping boundary; and a round of traversal is performed on all road test trajectories, and during this round of traversal, the sub-trajectories of the currently traversed road test trajectory within each local clipping boundary are extracted as new road test trajectories; and a round of traversal is performed on all perception trajectories, and during this round of traversal, the sub-trajectories of the currently traversed perception trajectory within each local clipping boundary are extracted as new perception trajectories.

[0145] Step 234, and after the two rounds of traversal are completed, a new set of road test trajectories is formed by all the new road test trajectories, and a new set of perception trajectories is formed by all the new perception trajectories.

[0146] Step 3: Calculate the curvature of the trajectory points of each road test trajectory and segment the trajectory based on the calculation results to obtain the corresponding segmented trajectory sequence; form the corresponding set of trajectory to be matched by the sensing trajectories in the sensing trajectory set that have trajectory time intersections with each segmented trajectory; select the corresponding matching trajectory set for each segmented trajectory from its corresponding set of trajectory to be matched based on the DTW algorithm; perform trajectory merging and deduplication processing on all matching trajectory sets corresponding to each segmented trajectory sequence to obtain the corresponding merged trajectory set; and stitch the trajectories based on each merged trajectory set according to the preset spatiotemporal continuity rules to obtain the corresponding sensing stitched trajectory.

[0147] Step 3 specifically includes:

[0148] Step 31: Calculate the curvature of the trajectory points of each road test trajectory and segment the trajectory based on the calculation results to obtain the corresponding segmented trajectory sequence.

[0149] Here, the segmented trajectory sequence of this invention corresponds one-to-one with the road test trajectory, and is formed by sequentially sorting one or more segmented trajectories; each segmented trajectory corresponds to a segment type; the segment types include straight, left turn, and right turn; each segmented trajectory includes multiple segmented trajectory points; the segmented trajectory points include trajectory point identifier, trajectory point coordinates, timestamp, speed, and heading angle.

[0150] Step 31 specifically includes:

[0151] Step 311: Count the total number of trajectory points Q of the road test trajectory;

[0152] Step 312, based on the timestamp t of each first trajectory point of the road test trajectory q Speed ​​v q Heading angle θ q Calculate the corresponding curvature k q Based on a preset straight curvature threshold range, and according to each curvature k... q Determine the corresponding trajectory direction type s q ;

[0153] Here, 1 ≤ index q ≤ Q;

[0154] Curvature k in the embodiment of the present invention q The calculation method is as follows:

[0155]

[0156] The straight curvature threshold range of this invention embodiment is [k] min ,k max ], where k min k max For the corresponding minimum and maximum boundary thresholds; k min <0<k max ;

[0157] Trajectory direction type s in this embodiment of the invention q Including left turn, straight ahead, and right turn, the method for determining them is as follows:

[0158]

[0159] Step 313: Divide each segment on the road test trajectory into U or more consecutive trajectory direction types s. q The sequence of trajectory points, consisting of the first trajectory points that are all left turns, straight ahead, or right turns, is a segmented trajectory with a segmentation type of left turn, straight ahead, or right turn; and all the obtained segmented trajectories are sorted in chronological order to form the corresponding segmented trajectory sequence;

[0160] The preset minimum total number of continuous trajectory points U is a positive integer.

[0161] Step 32, and the corresponding set of tracks to be matched is formed by the perceptual tracks in the perceptual track set that have intersections with the time periods of each segmented track.

[0162] Step 33, and based on the DTW algorithm, select the corresponding matching trajectory set from the corresponding unmatched trajectory set for each segmented trajectory.

[0163] Specifically, it includes: step 331, recording the segmented trajectory as the corresponding trajectory X; recording each perceived trajectory of the trajectory set to be matched as the corresponding trajectory Y; and counting the total number of trajectory points of trajectory X and trajectory Y to obtain the corresponding total number N and M respectively; and constructing a cumulative cost matrix D of shape N×M for trajectory X and trajectory Y;

[0164] Here, the trajectory points of each segment of trajectory X in this embodiment of the invention are denoted as trajectory points x. i Each trajectory point x i The trajectory point coordinates are marked as the corresponding position coordinates. 1 ≤ index i ≤ N; each second trajectory point of trajectory Y is denoted as trajectory point y. j Each trajectory point y j The trajectory point coordinates are marked as the corresponding position coordinates. 1 ≤ index j ≤ M;

[0165] The matrix rows of the cumulative cost matrix D and the trajectory point x in this embodiment of the invention i One-to-one correspondence, matrix columns and trajectory points y j One-to-one correspondence; the starting position of both the row index and column index is the bottom left vertex cell; the cumulative cost matrix D consists of N×M matrix cells (i,j), and the cell data of each matrix cell (i,j) is the corresponding cumulative cost d. i,j ;like Figure 3 The diagram shows the cumulative cost matrix and regularization path provided in Embodiment 1 of the present invention.

[0166] Step 332: Based on the DTW algorithm, calculate the cumulative cost matrix for all cumulative costs d in the cumulative cost matrix D. i,j Perform calculations;

[0167] Specifically, this includes: Step 3321, first calculating the cumulative cost d of the matrix element at the lower left vertex of the matrix. i＝1,j＝1 :

[0168]

[0169] Step 3322: Calculate the cumulative cost d for the remaining M-1 rows of the matrix in the first row, excluding the bottom left vertex. i＝1,1<j≤M And calculate the cumulative cost d of the remaining N-1 columns of the matrix, excluding the bottom left vertex.1<i≤N,j＝1 :

[0170]

[0171] Step 3323, then calculate the cumulative cost d of the remaining (N-1)×(M-1) matrix elements. 2≤i≤N,2≤j≤M :

[0172]

[0173] Here, f in the embodiment of the present invention dis () is the Euclidean distance calculation function, used to calculate the Euclidean distance between the coordinates of two positions input to the function.

[0174] Step 333: Calculate the cumulative cost d corresponding to the top-right vertex matrix element (N, M) in the cumulative cost matrix D. i＝N,j＝M This is the DTW distance between trajectory X and the current trajectory Y.

[0175] Step 334: Based on the regularized path backtracking method of the DTW algorithm, backtrack the path points from the upper right vertex matrix unit (N,M) to the lower left vertex matrix unit (1,1) and take the reverse path of the backtracked path as the corresponding regularized path W.

[0176] Specifically, this includes: Step 3341, initializing the backtracking point index b to 1; and setting the upper right corner vertex matrix element (N, M) as the current matrix element (i). now ,j now );

[0177] Among them, i now j now These are the row index and column index of the current cell, respectively;

[0178] Step 3342, set the current matrix element (i) now ,j now Mark it as a corresponding backtracking point. And for the current matrix element (i) now ,j now If the left-bottom vertex matrix cell (1,1) is identified, proceed to step 3344 if yes, otherwise proceed to step 3343.

[0179] Step 3343, set the current matrix element (i) now ,j now The left neighbor matrix element (i) now ,j now -1), Lower neighbor matrix unit (i now -1,j now ), left-sloping lower matrix unit (i now -1,jnow -1) corresponds to the three cumulative costs The minimum value in the matrix is ​​taken as the current minimum cost; and the matrix element corresponding to the current minimum cost is taken as the new current matrix element (i). now ,j now ); and increment the backtracking point index b by 1; and return to step 3342;

[0180] Step 3344: Use the index value of the current backtracking point index b as the corresponding path length L; and set each backtracking point... Let w be the corresponding path point. l＝L-b+1 And from the obtained L path points w l The sequential arrangement forms the corresponding regular path W;

[0181] Here, the regularized path W in this embodiment of the invention consists of multiple path points w l Composition, 1 ≤ index l ≤ L; each path point w l A corresponding matrix element is denoted as the corresponding matrix element (i). l ,j l ), i l j l respectively, path point w l The row index and column index of the corresponding matrix element; such as Figure 3 As shown.

[0182] Step 335: Based on the regularized path W of each trajectory Y, calculate the trajectory error E between trajectory X and each trajectory Y.

[0183] Specifically, this includes: taking the normalized path W of each trajectory Y as the current normalized path; and processing all trajectory points x of trajectory X. i Perform one round of traversal; and during this round of traversal, change the current traversed trajectory point x. i And its corresponding index i is denoted as the current road test point and the current index i. * ; and in the current normalized path, the row index i of the corresponding matrix unit l With current index i * Equal path points w l All are used as corresponding alignment points; and the index j in the current trajectory Y is compared with the column index j corresponding to each alignment point. l Matched trajectory point y jAll are denoted as aligned sensing points; the Euclidean distance between the current road test point and each aligned sensing point is calculated and the calculation result is used as the corresponding road test-sensing point error; the mean of all road test-sensing point errors corresponding to the current road test point is calculated to obtain the corresponding road test point mean error; at the end of this round of traversal, the mean of all road test point mean errors obtained in this round of traversal is calculated and the calculation result is used as the trajectory error E corresponding to the current trajectory Y.

[0184] Step 336: A first data group is formed by the DTW distance and trajectory error E corresponding to each trajectory Y; and the first data group with DTW distance less than a preset first distance threshold and trajectory error E less than a preset first error threshold is recorded as the second data group.

[0185] In this embodiment, the first data group corresponding to the current segmented trajectory corresponds one-to-one with the perceived trajectories in the trajectory set to be matched corresponding to the current segmented trajectory. The first distance threshold and the first error threshold in this embodiment are two pre-set threshold parameters.

[0186] Step 337: Count the total number of data groups in the second data group; if the total number of data groups is zero, set the corresponding matching trajectory set to empty; if the total number of data groups is greater than zero, then the corresponding matching trajectory set is composed of all the sensing trajectories corresponding to the second data group.

[0187] Here, in this embodiment of the invention, the matching trajectory set corresponds one-to-one with the segmented trajectory. When the matching trajectory set is not empty, it consists of one or more sensing trajectories that match the current segmented trajectory.

[0188] Step 34 involves merging and deduplicating all matching trajectory sets corresponding to each segment trajectory sequence to obtain the corresponding merged trajectory set.

[0189] Step 35: Based on the preset spatiotemporal continuity rules, the trajectory is spliced ​​to obtain the corresponding perception spliced ​​trajectory.

[0190] Specifically, this includes step 351, which takes each merged trajectory set as the current merged trajectory set.

[0191] Step 352: Identify whether the current merged trajectory set is empty; if yes, set the corresponding perceptual stitching trajectory to empty and proceed to step 357; if no, proceed to step 353.

[0192] Step 353: Initialize an empty trajectory sequence for the current merged trajectory set as the corresponding first sequence; add the sensing trajectory with the earliest start time in the current merged trajectory set to the first sequence; and after successful addition, delete the added sensing trajectory from the current merged trajectory set.

[0193] Step 354: Take the last perception trajectory in the first sequence as the current tail trajectory.

[0194] Step 355: According to the spatiotemporal continuity rule, select a perception trajectory from the current merged trajectory set that satisfies the spatiotemporal continuity relationship with the current tail trajectory as the corresponding next segment trajectory.

[0195] Here, the spatiotemporal continuity rule of this invention stipulates that the next segment of the trajectory that satisfies the spatiotemporal continuity relationship with the current tail trajectory should simultaneously satisfy preset spatial continuity sub-rules, temporal continuity sub-rules, motion continuity rules, and matching consistency sub-rules; wherein...

[0196] 1) Spatial continuity sub-rule requirement: The Euclidean distance between the last second trajectory point of the current tail trajectory and the first second trajectory point of the next segment should be lower than the preset second distance threshold;

[0197] Here, the second distance threshold in this embodiment of the invention is a pre-set threshold parameter;

[0198] 2) The time continuity sub-rule requires that the time of the last second trajectory point of the current tail trajectory should be earlier than the time of the first second trajectory point of the next segment trajectory, and the time interval between the first second trajectory point of the next segment trajectory and the last second trajectory point of the current tail trajectory should be less than the preset first duration threshold.

[0199] Here, the first duration threshold in this embodiment of the invention is a pre-set duration threshold parameter;

[0200] 3) Motion continuity sub-rule requirement: The uniformly accelerated motion model is based on the timestamp t of the last second trajectory point of the current tail trajectory. pre Speed ​​v pre And the timestamp t of the first and second trajectory points of the next trajectory segment. nxt Speed ​​v nxt Calculate the corresponding acceleration a = (v nxt -v pre ) / (t nxt -t pre The absolute value of the acceleration a, |a|, should be less than the preset absolute value threshold for acceleration.

[0201] Here, the absolute acceleration threshold of this embodiment of the invention is a preset acceleration threshold parameter with a value greater than or equal to 0;

[0202] 4) Matching consistency sub-rule requirements: Record the segmented trajectories that match the current tail trajectory and the next segment trajectory as the corresponding first segment trajectory and second segment trajectory; if the first and second segment trajectories are different, the current matching consistency sub-rule is assumed to be satisfied; if the first and second segment trajectories are the same, the absolute difference |△E| between the two trajectories E corresponding to the current tail trajectory, the next segment trajectory and the current segment trajectory should be less than the preset absolute difference threshold.

[0203] It should be noted that if the next trajectory obtained in the embodiment of the present invention is empty, it means that there is no sensing trajectory in the current merged trajectory set that can simultaneously satisfy the four sub-rules of the spatiotemporal continuity rule.

[0204] Step 356: Identify whether the next trajectory segment is empty; if so, sequentially splice all the perceptual trajectories in the first sequence into the corresponding perceptual spliced ​​trajectory, and proceed to step 357; if not, add the current next trajectory segment to the first sequence, and after successful addition, delete the current next trajectory segment from the current merged trajectory set, and return to step 354.

[0205] Step 357: Output the obtained perceptual stitching trajectory as the result of this trajectory stitching.

[0206] Here, when the sensing stitching trajectory in this embodiment of the invention is not empty, it includes multiple third trajectory points; the third trajectory points include trajectory point identifier, trajectory point coordinates, timestamp, speed, and heading angle.

[0207] Step 4: Each road test trajectory and its corresponding sensing stitching trajectory form a corresponding road test-sensing trajectory pair; the effective sensing trajectory length of each road test-sensing trajectory pair is calculated according to the high-low speed adaptive invalid trajectory length identification method; the corresponding sensing coverage rate is calculated based on the effective sensing trajectory length and road test trajectory length of each road test-sensing trajectory pair; and the average value of all sensing coverage rates is taken as the corresponding roadside sensing coverage rate.

[0208] Specifically, this includes: Step 41, which involves forming corresponding road test-sensor trajectory pairs from each road test trajectory and its corresponding sensing splicing trajectory.

[0209] Step 42, and calculate the effective sensing trajectory length of each road test-sensing trajectory pair according to the high and low speed adaptive invalid trajectory length identification method.

[0210] Specifically, step 421 involves taking each road test-sensing trajectory pair as the current trajectory pair.

[0211] Step 422: Take the road test trajectory and the perception stitching trajectory of the current trajectory pair as the corresponding current road test trajectory and current perception trajectory; and identify whether the current perception trajectory is empty; if yes, set the length of the corresponding valid perception trajectory to 0 and go to step 429; if no, go to step 423.

[0212] Step 423: Perform initial calculation on the trajectory length of the current sensing trajectory to obtain the corresponding initial trajectory length;

[0213] Specifically, this includes: estimating the trajectory travel distance between every two third trajectory points of the current perceived trajectory to obtain the corresponding first distance; and using the sum of all first distances as the corresponding initial trajectory length.

[0214] Step 424: Take the first trajectory point on the current road test trajectory that is closest to each of the third trajectory points on the current sensing trajectory as the pairing point of the current third trajectory point, and calculate the Euclidean distance between each third trajectory point and its corresponding pairing point, and use the calculation result as the corresponding pairing point error.

[0215] Step 425: Identify high-speed segments and low-speed segments on the current sensing trajectory based on preset high-speed thresholds and low-speed thresholds;

[0216] Specifically, this includes: marking the third trajectory point on the current sensing trajectory with a speed greater than the high-speed threshold as a high-speed point and the third trajectory point with a speed less than the low-speed threshold as a low-speed point; and marking each trajectory segment consisting of G or more consecutive high-speed points as a corresponding high-speed segment and each trajectory segment consisting of H or more consecutive low-speed points as a corresponding low-speed segment.

[0217] Here, the low-speed threshold and high-speed threshold in this embodiment of the invention are two preset speed thresholds, wherein 0 < low-speed threshold < high-speed threshold. The preset total number of continuous trajectory points G and H are each a positive integer.

[0218] Step 426: Identify large-error invalid time periods based on all high-speed segments and calculate the corresponding first invalid length based on the identification results;

[0219] Specifically, this includes: performing a round of traversal on all high-speed segments; during this round of traversal, designating the currently traversed high-speed segment as the current segment; and polling all third trajectory points of the current segment in a round of polling, designating the currently polled third trajectory point as the current point, and identifying whether the pairing point error corresponding to the current point is greater than a preset third distance threshold. If so, the timestamp of the current point is used as the corresponding current time t. mid And based on the preset duration r, a corresponding invalid time period [t] is set. mid -r,tmid +r]; and at the end of this round of traversal, merge all the invalid time periods obtained to obtain one or more merged time periods; mark the third trajectory point on the current perception trajectory that is in any merged time period as an invalid point; take the first distance between every two adjacent invalid points as a corresponding second distance; and take the sum of all second distances as the corresponding first invalid length.

[0220] Here, the third distance threshold in this embodiment of the invention is a pre-set error distance threshold parameter; the duration r in this embodiment of the invention is a pre-set time length parameter, such as 0.3 seconds.

[0221] Step 427: Calculate the length of the invalid trajectory with large error and adjust the length of the valid trajectory with small error based on all low-speed segments to obtain the corresponding second invalid length and first adjusted length;

[0222] Specifically, this includes: Step 4271, performing a traversal of all low-speed segments; during this traversal, the currently traversed low-speed segment is taken as the current segment; the rectangular area of ​​the road traversed by the current segment on the local high-precision map is taken as the current region; the current region is gridded according to the preset square grid size to obtain multiple corresponding region grids; the region grids intersecting with the trajectory route of the current segment are recorded as the corresponding passing grids; the average error of the pairing points of all third trajectory points of each passing grid is calculated and the calculation result is taken as the corresponding grid error; and whether the grid error of each passing grid is greater than the third distance threshold is identified. If it is, the current grid is recorded as an invalid grid; otherwise, the current grid is recorded as a valid grid.

[0223] Here, the square grid size in this embodiment of the invention is a pre-set square grid size parameter, specifically composed of the corresponding square side lengths;

[0224] Step 4272: After this round of traversal, the first distance between each pair of adjacent third trajectory points in each invalid grid is summed and the result is used as the corresponding invalid grid trajectory length. The lengths of all invalid grid trajectories are summed and the result is used as the corresponding second invalid length.

[0225] Step 4273: Sum the first distances between adjacent third trajectory points in each effective grid and use the result as the corresponding original grid trajectory length; use the side length of the square grid of the square grid size as the unified network trajectory length; set the corresponding grid optimized trajectory length = max(0, original grid trajectory length - unified network trajectory length) based on each original grid trajectory length and the unified network trajectory length; and sum all the obtained grid optimized trajectory lengths and use the result as the corresponding first adjustment length.

[0226] Step 428: The effective sensing trajectory length is obtained by subtracting the sum of the first invalid length, the second invalid length, and the first adjusted length from the initial trajectory length;

[0227] Wherein, the effective sensing trajectory length = initial trajectory length - (first invalid length + second invalid length + first adjustment length).

[0228] Step 429: The obtained effective sensing trajectory length is output as the result of this calculation.

[0229] Step 43, and calculate the corresponding sensing coverage based on the effective sensing trajectory length and driving trajectory length of each road test-sensing trajectory pair;

[0230] Specifically, this includes: taking each road-sensing trajectory pair as the current trajectory pair; taking the road-sensing trajectory of the current trajectory pair as the corresponding current trajectory; estimating the trajectory travel distance between every two first trajectory points of the current trajectory to obtain the corresponding third distance; taking the sum of all third distances as the corresponding road-sensing trajectory length; and calculating the corresponding sensing coverage rate, where sensing coverage rate = (effective sensing trajectory length / road-sensing trajectory length) × 100%.

[0231] Step 44, and take the average of all perception coverage as the corresponding roadside perception coverage.

[0232] Step 5: Each roadside perception trajectory pair and its corresponding perception coverage rate constitute a corresponding roadside perception evaluation item; all roadside perception evaluation items constitute a corresponding roadside perception evaluation table; and the first task request, the roadside perception evaluation table, and the roadside perception coverage rate constitute a corresponding roadside perception evaluation report and save it.

[0233] Here, in this embodiment of the invention, while saving the roadside perception evaluation report, the current roadside perception evaluation report can also be sent back to the sender of the current first task request.

[0234] Figure 4This is a module structure diagram of a processing device for roadside perception trajectory evaluation provided in Embodiment 2 of the present invention. This device can be a terminal device or server implementing the aforementioned method embodiments, or it can be a device that enables the aforementioned terminal device or server to implement the aforementioned method embodiments. For example, the device can be a device or chip system of the aforementioned terminal device or server. Figure 4 As shown, the processing device for roadside perception trajectory evaluation provided in Embodiment 2 of the present invention includes: an evaluation task receiving module 201, a data preparation module 202, a trajectory processing module 203, a coverage evaluation module 204, and an evaluation report processing module 205.

[0235] The evaluation task receiving module 201 is used to receive roadside perception trajectory evaluation task requests, which are denoted as the corresponding first task request; the first task request includes the first task identifier, the first evaluation period, and the first intersection identifier.

[0236] The data preparation module 202 obtains the current intersection's current time period, perception trajectory set, and local high-precision map from the preset road test trajectory cloud, roadside perception cloud, and high-precision map cloud based on the first evaluation period and the first intersection identifier requested by the first task. It then transforms the coordinates of all trajectory points in the two trajectory sets to the high-precision map coordinate system corresponding to the local high-precision map. Based on the preset trajectory clipping mode, it performs trajectory clipping on the two trajectory sets according to the local high-precision map. The road test trajectory set includes multiple road test trajectories; the perception trajectory set includes multiple perception trajectories; and the map elements of the local high-precision map include ROI elements, lane line elements, and stop line elements.

[0237] The trajectory processing module 203 is used to calculate the curvature of the trajectory points of each road test trajectory and segment the trajectory based on the calculation results to obtain the corresponding segmented trajectory sequence; and form the corresponding set of trajectory to be matched by the sensing trajectories in the sensing trajectory set that have trajectory time intersections with each segmented trajectory; and select the corresponding matching trajectory set for each segmented trajectory from its corresponding set of trajectory to be matched based on the DTW algorithm; and perform trajectory merging and deduplication processing on all matching trajectory sets corresponding to each segmented trajectory sequence to obtain the corresponding merged trajectory set; and perform trajectory splicing based on each merged trajectory set according to the preset spatiotemporal continuity rules to obtain the corresponding sensing spliced ​​trajectory.

[0238] The coverage assessment module 204 is used to form corresponding road-sensor trajectory pairs from each road-test trajectory and its corresponding sensing stitching trajectory; calculate the effective sensing trajectory length of each road-sensor trajectory pair according to the high-low speed adaptive invalid trajectory length identification method; calculate the corresponding sensing coverage rate based on the effective sensing trajectory length and road-test trajectory length of each road-sensor trajectory pair; and take the average of all sensing coverage rates as the corresponding roadside sensing coverage rate.

[0239] The evaluation report processing module 205 is used to form a corresponding roadside perception evaluation item by each road test-perception trajectory pair and its corresponding perception coverage; and to form a corresponding roadside perception evaluation table by all roadside perception evaluation items; and to form and save a corresponding roadside perception evaluation report by the first task request, the roadside perception evaluation table and the roadside perception coverage.

[0240] The processing device for roadside sensing trajectory evaluation provided in this embodiment of the invention can execute the method steps in the above method embodiment. Its implementation principle and technical effect are similar, and will not be repeated here.

[0241] It should be noted that the division of the various modules in the above device is merely a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, these modules can be implemented entirely in software via processing element calls; they can be fully implemented in hardware; or some modules can be implemented by processing element calls to software, while others are implemented in hardware. For example, the evaluation task receiving module can be a separate processing element, or it can be integrated into a chip in the above device. Alternatively, it can be stored as program code in the memory of the above device, and called and executed by a processing element of the device. The implementation of other modules is similar. Moreover, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed through integrated logic circuits in the hardware of the processor element or through software instructions.

[0242] For example, these modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more Digital Signal Processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). As another example, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a System-on-a-Chip (SOC).

[0243] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. This computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the foregoing method embodiments are generated. The computer described above can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The aforementioned computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the aforementioned computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, Bluetooth, microwave, etc.) means. The aforementioned computer-readable storage medium can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The aforementioned available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0244] Figure 5 This is a schematic diagram of an electronic device provided in Embodiment 3 of the present invention. This electronic device can be a terminal device or server implementing the methods of the aforementioned embodiments, or it can be a terminal device or server connected to the aforementioned terminal device or server implementing the methods of the aforementioned embodiments. Figure 5 As shown, the electronic device may include: a processor 301 (e.g., CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transmission and reception operations of the transceiver 303. The memory 302 may store various instructions for performing various processing functions and implementing the processing steps described in the foregoing embodiments. Preferably, the electronic device involved in the embodiments of the present invention further includes: a power supply 304, a system bus 305, and a communication port 306. The system bus 305 is used to realize communication connections between components. The communication port 306 is used for communication between the electronic device and other peripherals.

[0245] exist Figure 5The system bus 305 mentioned can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This system bus can be divided into address bus, data bus, control bus, etc. For ease of representation, it is represented by only one thick line in the figure, but this does not indicate that there is only one bus or one type of bus. The communication interface is used to enable communication between the database access device and other devices (e.g., clients, read-write libraries, and read-only libraries). Memory may include Random Access Memory (RAM) and may also include non-volatile memory, such as at least one disk storage device.

[0246] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), graphics processing units (GPUs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0247] It should be noted that the embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when run on a computer, cause the computer to perform the methods and processes provided in the above embodiments.

[0248] This invention provides a processing method, apparatus, electronic device, and computer-readable storage medium for roadside sensing trajectory evaluation. As described above, this invention calculates the curvature of trajectory points for each roadside trajectory and segments the trajectory based on the calculation results to obtain corresponding segmented trajectory sequences. Then, based on the DTW algorithm, it selects corresponding matching trajectory sets from numerous sensing trajectories that intersect with each segmented trajectory in time intervals. Finally, it merges and deduplicates all matching trajectory sets corresponding to each roadside trajectory to obtain a merged trajectory set. Following spatiotemporal continuity rules, it splices the trajectories based on the merged trajectory sets corresponding to each roadside trajectory to obtain the corresponding sensing spliced ​​trajectory. Next, each roadside trajectory plus the sensing spliced ​​trajectory forms a corresponding trajectory pair. The effective sensing trajectory length of each trajectory pair is calculated using a high-low speed adaptive invalid trajectory length identification method. The corresponding sensing coverage rate is calculated based on the effective sensing trajectory length plus the roadside trajectory length of each trajectory pair, and the average of all sensing coverage rates is taken as the corresponding roadside sensing coverage rate. This invention employs a curvature-based roadside trajectory segmentation and DTW flexible matching algorithm, improving trajectory alignment accuracy by allowing nonlinear time warping. Furthermore, it utilizes a high-low speed adaptive invalid trajectory length identification method, employing an error time propagation-based elimination strategy for high-speed segments and a grid jitter subtraction strategy for low-speed segments, thereby enhancing the accuracy of invalid sensing segment identification and elimination. Additionally, it leverages a four-fold spatiotemporal continuity rule encompassing space, time, motion, and matching consistency to reliably stitch together multiple sensing trajectories generated due to temporary target loss, achieving complete reconstruction of the target sensing trajectory. This invention not only solves the problems of insufficient trajectory alignment accuracy, difficulty in eliminating invalid sensing segments, and difficulty in stitching together multiple sensing trajectories in conventional solutions, but also effectively improves the assessment accuracy of roadside sensing capabilities.

[0249] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0250] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A processing method for roadside sensing trajectory evaluation, characterized in that, The method includes: The received roadside perception trajectory evaluation task request is recorded as the corresponding first task request; the first task request includes a first task identifier, a first evaluation time period, and a first intersection identifier; Based on the first evaluation period and the first intersection signage requested by the first task, the road test trajectory set, perception trajectory set, and local high-precision map for the current intersection and current period are obtained from the preset road test trajectory cloud, roadside perception cloud, and high-precision map cloud; the coordinates of all trajectory points in the two trajectory sets are transformed to the high-precision map coordinate system corresponding to the local high-precision map; and the two trajectory sets are clipped according to the local high-precision map based on the preset trajectory clipping mode; the road test trajectory set includes multiple road test trajectories; the perception trajectory set includes multiple perception trajectories; the map elements of the local high-precision map include ROI elements, lane line elements, and stop line elements; The curvature of the trajectory points of each road test trajectory is calculated, and the trajectory is segmented based on the calculation results to obtain the corresponding segmented trajectory sequence. A set of trajectory to be matched is formed from the sensing trajectories in the sensing trajectory set that intersect with each segmented trajectory in terms of trajectory time intervals. A matching trajectory set is selected from the corresponding set of trajectory to be matched for each segmented trajectory based on the DTW algorithm. Trajectories are merged and deduplicated for all matching trajectory sets corresponding to each segmented trajectory sequence to obtain the corresponding merged trajectory set. Finally, the trajectory is stitched together based on each merged trajectory set according to a preset spatiotemporal continuity rule to obtain the corresponding sensing stitched trajectory. Each road test trajectory and its corresponding sensing stitched trajectory form a corresponding road test-sensing trajectory pair; the effective sensing trajectory length of each road test-sensing trajectory pair is calculated according to the high-low speed adaptive invalid trajectory length identification method; the corresponding sensing coverage rate is calculated based on the effective sensing trajectory length and road test trajectory length of each road test-sensing trajectory pair; and the average value of all the sensing coverage rates is taken as the corresponding roadside sensing coverage rate. Each of the roadside perception trajectory pairs and its corresponding perception coverage rate constitutes a corresponding roadside perception evaluation item; all the roadside perception evaluation items constitute a corresponding roadside perception evaluation table; and the first task request, the roadside perception evaluation table, and the roadside perception coverage rate constitute a corresponding roadside perception evaluation report and are saved.

2. The processing method for roadside sensing trajectory evaluation according to claim 1, characterized in that, The road test trajectory cloud stores multiple first intersection trajectory sets in the cloud; each first intersection trajectory set corresponds to an intersection identifier; each first intersection trajectory set consists of multiple first intersection trajectories; each first intersection trajectory is a segment of vehicle motion trajectory collected when a road test vehicle passes through the current intersection within the current trajectory time period; the trajectory information of each first intersection trajectory consists of the vehicle identifier of the current road test vehicle, the unique trajectory identifier of the current trajectory, and the trajectory point sequence of the current trajectory; each trajectory point of the first intersection trajectory corresponds to a set of trajectory point identifiers, trajectory point coordinates, timestamps, speeds, and heading angles; The roadside sensing cloud stores multiple second intersection trajectory sets in the cloud; each second intersection trajectory set corresponds to an intersection identifier; each second intersection trajectory set consists of multiple second intersection trajectories; each second intersection trajectory is a target sensing trajectory obtained by the roadside unit device at the current intersection through roadside target sensing technology within the current trajectory time period; the trajectory information of each second intersection trajectory consists of the unique target identifier of the currently sensed target and the trajectory point sequence of the current trajectory; each trajectory point of the second intersection trajectory corresponds to a set of trajectory point identifiers, trajectory point coordinates, timestamps, speeds, and heading angles; The high-precision map cloud stores high-precision maps of all intersections in the cloud; the map coverage of each high-precision map of an intersection includes not only all the internal areas of the current intersection, but also a section of the outer edge of each road leading into or out of the intersection; the outer edge length of each outer edge road is a preset first length threshold. The road test trajectory includes multiple first trajectory points; each first trajectory point includes a trajectory point identifier, trajectory point coordinates, timestamp, speed, and heading angle. The sensing trajectory includes multiple second trajectory points; each second trajectory point includes a trajectory point identifier, trajectory point coordinates, timestamp, speed, and heading angle. The number of the ROI element, the lane line element, or the stop line element is greater than or equal to 1; Each ROI element corresponds to a region of interest within the current intersection; the ROI element includes the region identifier, shape type, and geometric dimension parameters of the corresponding region; the shape type includes polygon and circle, and when it is a polygon, the corresponding geometric dimension parameters are the set of vertex coordinates of the current region, and when it is a circle, the corresponding geometric dimension parameters include the coordinates of the center point of the current region and the region radius; Each lane line element corresponds to a driving lane entering or exiting the current intersection, or a driving lane within the current intersection; the lane line element includes a lane identifier and a lane line coordinate set; the lane line coordinate set consists of multiple lane line point coordinates; Each stop line element corresponds to a stop line at the entry or exit point of the current intersection, or a stop line within the current intersection; each stop line element includes a stop line identifier and a stop line coordinate set; the stop line coordinate set consists of multiple stop line point coordinates. The segmented trajectory sequence corresponds one-to-one with the road test trajectory and is formed by sequentially sorting one or more segmented trajectories; each segmented trajectory corresponds to a segment type; the segment type includes straight, left turn, and right turn; each segmented trajectory includes multiple segmented trajectory points; each segmented trajectory point includes trajectory point identifier, trajectory point coordinates, timestamp, speed, and heading angle; The sensing and stitching trajectory includes multiple third trajectory points; each third trajectory point includes a trajectory point identifier, trajectory point coordinates, timestamp, speed, and heading angle.

3. The processing method for roadside sensing trajectory evaluation according to claim 2, characterized in that, The process of obtaining the current intersection's current time period, perception trajectory set, and local high-precision map from the preset road test trajectory cloud, roadside perception cloud, and high-precision map cloud based on the first evaluation period and the first intersection identifier, specifically includes: The first and second intersection trajectories on the road test trajectory cloud and the roadside perception cloud that have a time intersection with the first evaluation period corresponding to the first intersection sign are respectively recorded as the first and second candidate trajectories; the trajectory points in each of the first and second candidate trajectories that are within the first evaluation period are extracted to form the corresponding road test trajectory and the perception trajectory; and the corresponding road test trajectory set and perception trajectory set are formed by all the obtained road test trajectories and all the obtained perception trajectories; and the intersection high-precision map on the high-precision map cloud that corresponds to the first intersection sign is used as the corresponding local high-precision map.

4. The processing method for roadside sensing trajectory evaluation according to claim 1, characterized in that, The trajectory clipping modes include a first mode, a second mode, and a third mode; When the trajectory clipping mode is the first mode, the overall clipping boundary is the boundary of the combined area of ​​the current intersection and all its corresponding entrance and exit outer roads. When the trajectory clipping mode is the second mode, the boundaries of each region of interest at the current intersection are used as a corresponding local clipping boundary. When the trajectory clipping mode is the third mode, the area boundary of the outer edge of the entrance and exit roads outside each stop line of the current intersection is used as a corresponding local clipping boundary.

5. The processing method for roadside sensing trajectory evaluation according to claim 4, characterized in that, The preset trajectory clipping mode performs trajectory clipping on the two trajectory sets according to the local high-precision map, specifically including: The trajectory clipping pattern is identified; If the trajectory clipping mode is the first mode, then the boundary of the overall joint area of ​​the current intersection and all its corresponding entrance and exit outer roads is identified according to the local high-precision map, and the identification result is used as the corresponding overall clipping boundary; then all the road test trajectories are traversed once, and during this round of traversal, the trajectory points of the currently traversed road test trajectory outside the overall clipping boundary are clipped to obtain a new road test trajectory; then all the perception trajectories are traversed once, and during this round of traversal, the trajectory points of the currently traversed perception trajectory outside the overall clipping boundary are clipped to obtain a new perception trajectory; If the trajectory clipping mode is the second mode or the third mode, then when the current mode is the second mode, the boundary of the region of interest corresponding to each ROI element of the local high-precision map is used as a corresponding local clipping boundary; when the current mode is the third mode, the boundary of the area of ​​the outer edge of the entrance and exit road outside the stop line corresponding to each stop line element of the local high-precision map is used as a corresponding local clipping boundary; and a round of traversal is performed on all the road test trajectories, and during this round of traversal, the sub-trajectories of the currently traversed road test trajectory within each of the local clipping boundaries are extracted as a new road test trajectory; and a round of traversal is performed on all the perception trajectories, and during this round of traversal, the sub-trajectories of the currently traversed perception trajectory within each of the local clipping boundaries are extracted as a new perception trajectory; After the two rounds of traversal are completed, a new set of road test trajectories is formed by all the new road test trajectories, and a new set of perception trajectories is formed by all the new perception trajectories.

6. The processing method for roadside sensing trajectory evaluation according to claim 2, characterized in that, The calculation of the curvature of the trajectory points of each of the aforementioned road test trajectories and the segmentation of the trajectory based on the calculation results to obtain the corresponding segmented trajectory sequence specifically includes: Step 61: Count the total number Q of the trajectory points of the road test trajectory; Step 62, based on the timestamp t of each of the first trajectory points of the road test trajectory. q Speed ​​v q Heading angle θ q Calculate the corresponding curvature k q Based on a preset straight curvature threshold range, and according to each curvature k... q Determine the corresponding trajectory direction type s q ; Where 1 ≤ index q ≤ Q; The curvature k q The calculation method is as follows: The straight curvature threshold range is [k] min ,k max ], k min k max For the corresponding minimum and maximum boundary thresholds; k min <0<k max ; The trajectory direction type s q Including left turn, straight ahead, and right turn, the method for determining them is as follows: Step 63: Divide each segment of the road test trajectory into U or more consecutive trajectory direction types s. q A sequence of trajectory points consisting of the first trajectory points that are all left turns, straight ahead, or right turns is a segmented trajectory whose segmentation type is left turn, straight ahead, or right turn; and all the obtained segmented trajectories are sorted in chronological order to form the corresponding segmented trajectory sequence; The preset minimum total number of continuous trajectory points U is a positive integer.

7. The processing method for roadside sensing trajectory evaluation according to claim 2, characterized in that, The DTW algorithm-based selection of a matching trajectory set from the corresponding set of unmatched trajectories for each segmented trajectory specifically includes: Step 71: Record the segmented trajectory as the corresponding trajectory X; record each of the perceived trajectories in the trajectory set to be matched as the corresponding trajectory Y; and count the total number of trajectory points of trajectory X and trajectory Y to obtain the corresponding total number N and M respectively; and construct a cumulative cost matrix D of shape N×M for trajectory X and trajectory Y. Wherein, each segment trajectory point of the trajectory X is denoted as trajectory point x. i Each of the aforementioned trajectory points x i The trajectory point coordinates are marked as the corresponding position coordinates. 1 ≤ index i ≤ N; each of the second trajectory points of the trajectory Y is denoted as trajectory point y. j Each of the said trajectory points y j The trajectory point coordinates are marked as the corresponding position coordinates. 1 ≤ index j ≤ M; The rows of the cumulative cost matrix D and the trajectory point x i One-to-one correspondence, matrix columns and the trajectory points y j One-to-one correspondence; the starting position of both the row index and column index is the lower left vertex cell position; the cumulative cost matrix D consists of N×M matrix cells (i,j), and the cell data of each matrix cell (i,j) is the corresponding cumulative cost d. i,j ; Step 72: Based on the DTW algorithm, calculate the cumulative cost matrix for all the cumulative costs d in the cumulative cost matrix D. i,j The calculation is as follows: First, calculate the cumulative cost d of the matrix element at the bottom left vertex. i＝1,j＝1 : Next, calculate the cumulative cost d for the remaining M-1 rows of the matrix, excluding the bottom-left vertex. i＝1,1<j≤M And calculate the cumulative cost d of the remaining N-1 columns of the matrix, excluding the bottom left vertex. 1<i≤N,j＝1 : Next, calculate the cumulative cost d of the remaining (N-1)×(M-1) matrix elements. 2≤i≤N,2≤j≤M : Among them, f dis () is the Euclidean distance calculation function, used to calculate the Euclidean distance between two position coordinates input to the function; Step 73, the cumulative cost d corresponding to the upper right vertex matrix element (N,M) in the cumulative cost matrix D is... i＝N,j＝M The DTW distance between the trajectory X and the current trajectory Y; Step 74: Based on the regularized path backtracking method of the DTW algorithm, backtrack the path points from the upper right vertex matrix unit (N,M) to the lower left vertex matrix unit (1,1) and use the reverse path of the backtracked path as the corresponding regularized path W. Specifically: Step 741: Initialize the backtracking point index b to 1; and set the upper right corner vertex matrix unit (N, M) as the current matrix unit (i). now ,j now ); Among them, i now j now These are the row index and column index of the current cell, respectively; Step 742, set the current matrix unit (i) now ,j now Mark it as a corresponding backtracking point. And for the current matrix unit (i) now ,j now If the left-lower vertex matrix unit (1,1) is identified, proceed to step 744; otherwise, proceed to step 743. Step 743, set the current matrix unit (i) now ,j now The left neighbor matrix element (i) now ,j now -1), Lower neighbor matrix unit (i now -1,j now ), left-sloping lower matrix unit (i now -1,j now -1) corresponds to the three cumulative costs The minimum value in the matrix is ​​taken as the current minimum cost; and the matrix element corresponding to the current minimum cost is taken as the new current matrix element (i). now ,j now ); and increment the backtracking point index b by 1; and return to step 742; Step 744: Use the index value of the current backtracking point index b as the corresponding path length L; and set each backtracking point... Let w be the corresponding path point. l＝L-b+1 And from the obtained L path points w l The sequential arrangement forms the corresponding regular path W; The regularized path W consists of multiple path points w l Composition, 1 ≤ index l ≤ L; each of the path points w l A corresponding matrix element is denoted as the corresponding matrix element (i). l ,j l ), i l j l The path points w are respectively l The row and column indices of the corresponding matrix elements; Step 75: Based on the regularized path W of each trajectory Y, calculate the trajectory error E between the trajectory X and each trajectory Y, specifically as follows: The regularized path W of each trajectory Y is taken as the current regularized path; and for all trajectory points x of the trajectory X... i Perform one round of traversal; and during this round of traversal, change the current traversed trajectory point x. i And its corresponding index i is denoted as the current road test point and the current index i. * ; and in the current normalized path, the row index i of the corresponding matrix unit l With the current index i * Equal path points w l All are used as corresponding alignment points; and the index j in the current trajectory Y is compared with the column index j corresponding to each alignment point. l The matched trajectory point y j All are denoted as aligned sensing points; the Euclidean distance between the current road test point and each of the aligned sensing points is calculated, and the calculation result is used as the corresponding road test-sensing point error; the mean of all the road test-sensing point errors corresponding to the current road test point is calculated to obtain the corresponding road test point mean error; at the end of this round of traversal, the mean of all the road test point mean errors obtained in this round of traversal is calculated, and the calculation result is used as the trajectory error E corresponding to the current trajectory Y; Step 76: A first data group is formed by the DTW distance and the trajectory error E corresponding to each trajectory Y; and the first data group whose DTW distance is less than a preset first distance threshold and whose trajectory error E is less than a preset first error threshold is recorded as the second data group. Among them, the first data group corresponding to the current segmented trajectory corresponds one-to-one with the sensing trajectory in the set of trajectory to be matched corresponding to the current segmented trajectory; Step 77: Count the total number of data groups in the second data group; if the total number of data groups is zero, set the corresponding matching trajectory set to empty; if the total number of data groups is greater than zero, then the corresponding matching trajectory set is composed of all the sensing trajectories corresponding to the second data group.

8. The processing method for roadside sensing trajectory evaluation according to claim 2, characterized in that, The step of stitching together trajectories based on each of the merged trajectory sets according to preset spatiotemporal continuity rules to obtain the corresponding perceptual stitched trajectory specifically includes: Step 81: Take each of the merged trajectory sets as the current merged trajectory set; Step 82: Identify whether the current merged trajectory set is empty; if yes, set the corresponding perceived stitching trajectory to empty and proceed to step 87; if no, proceed to step 83. Step 83: Initialize an empty trajectory sequence for the current merged trajectory set as the corresponding first sequence; add the sensing trajectory with the earliest start time in the current merged trajectory set to the first sequence; and after successful addition, delete the sensing trajectory added at this time from the current merged trajectory set. Step 84: Take the last sensing trajectory in the first sequence as the current tail trajectory; Step 85: According to the spatiotemporal continuity rule, select a sensing trajectory from the current merged trajectory set that satisfies the spatiotemporal continuity relationship with the current tail trajectory as the corresponding next segment trajectory; The spatiotemporal continuity rule stipulates that the next segment of trajectory that satisfies the spatiotemporal continuity relationship with the current tail trajectory should simultaneously satisfy the preset spatial continuity sub-rule, temporal continuity sub-rule, motion continuity sub-rule, and matching consistency sub-rule. The spatial continuity sub-rule requires that the Euclidean distance between the last second trajectory point of the current tail trajectory and the first second trajectory point of the next segment trajectory should be lower than a preset second distance threshold. The time continuity sub-rule requires that: the time of the last second trajectory point of the current tail trajectory should be earlier than the time of the first second trajectory point of the next segment trajectory, and the time interval between the first second trajectory point of the next segment trajectory and the last second trajectory point of the current tail trajectory should be less than a preset first duration threshold. The motion continuity sub-rule requires that: the uniformly accelerated motion model be based on the timestamp t of the last second trajectory point of the current tail trajectory. pre Speed ​​v pre And the timestamp t of the first second trajectory point of the next trajectory segment nxt Speed ​​v nxt Calculate the corresponding acceleration a = (v nxt -v pre ) / (t nxt -t pre The absolute value of the acceleration a, |a|, should be less than a preset threshold value for absolute acceleration. The matching consistency sub-rule requires that: the segmented trajectories that match the current tail trajectory and the next segment trajectory are recorded as the corresponding first segment trajectory and second segment trajectory; if the first and second segment trajectories are different, the current matching consistency sub-rule is assumed to be satisfied; if the first and second segment trajectories are the same, the absolute difference |ΔE| between the two trajectories E corresponding to the current tail trajectory, the next segment trajectory and the current segment trajectory should be less than a preset absolute difference threshold. If the next trajectory segment is empty, it means that there is no sensing trajectory in the current merged trajectory set that can simultaneously satisfy the four sub-rules of the spatiotemporal continuity rule. Step 86: Identify whether the next segment of the trajectory is empty; if yes, then sequentially splice all the sensing trajectories in the first sequence into the corresponding sensing spliced ​​trajectory, and proceed to step 87; if no, then add the current next segment of the trajectory to the first sequence, and after successful addition, delete the current next segment of the trajectory from the current merged trajectory set, and return to step 84. Step 87: Output the obtained perception stitching trajectory as the result of this trajectory stitching.

9. The processing method for roadside sensing trajectory evaluation according to claim 2, characterized in that, The method for identifying invalid trajectory lengths based on high and low speed adaptive calculations calculates the valid sensing trajectory length for each of the road test-sensing trajectory pairs, specifically including: Step 91: Take each of the aforementioned road test-sensing trajectory pairs as the current trajectory pair; Step 92: Take the road test trajectory and the perception stitching trajectory of the current trajectory pair as the corresponding current road test trajectory and current perception trajectory; and identify whether the current perception trajectory is empty; if yes, set the length of the corresponding effective perception trajectory to 0 and go to step 99; if no, go to step 93. Step 93: Perform an initial calculation on the trajectory length of the current perceived trajectory to obtain the corresponding initial trajectory length, specifically as follows: The trajectory travel distance between every two third trajectory points of the current sensing trajectory is estimated to obtain the corresponding first distance; and the sum of all the first distances is used as the corresponding initial trajectory length. Step 94: Take the first trajectory point on the current road test trajectory that is closest to each of the third trajectory points on the current sensing trajectory as the pairing point of the current third trajectory point, and calculate the Euclidean distance between each third trajectory point and its corresponding pairing point, and use the calculation result as the corresponding pairing point error; Step 95: Based on preset high-speed and low-speed thresholds, identify high-speed and low-speed segments on the current sensing trajectory, specifically as follows: The third trajectory point on the current sensing trajectory with a speed greater than the high-speed threshold is recorded as a high-speed point, and the third trajectory point with a speed less than the low-speed threshold is recorded as a low-speed point; and each trajectory segment composed of G or more consecutive high-speed points is recorded as a corresponding high-speed segment, and each trajectory segment composed of H or more consecutive low-speed points is recorded as a corresponding low-speed segment. Where 0 < low speed threshold < high speed threshold; the preset total number of continuous trajectory points G and H are each a positive integer; Step 96: Identify large-error invalid time periods based on all the high-speed segments and calculate the corresponding first invalid length based on the identification results, specifically as follows: A traversal is performed on all the high-speed segments; during this traversal, the currently traversed high-speed segment is taken as the current segment; and a poll is performed on all the third trajectory points of the current segment, with the currently polled third trajectory point taken as the current point. The pairing point error corresponding to the current point is checked to see if it exceeds a preset third distance threshold. If so, the timestamp of the current point is taken as the corresponding current time t. mid And based on the preset duration r, a corresponding invalid time period [t] is set. mid -r,t mid +r]; and at the end of this round of traversal, merge all the invalid time periods obtained to obtain one or more merged time periods; mark the third trajectory points on the current sensing trajectory that are in any of the merged time periods as invalid points; and take the first distance between every two adjacent invalid points as a corresponding second distance; and take the sum of all the second distances as the corresponding first invalid length; Step 97: Based on all the low-speed segments, calculate the length of the invalid trajectory with large errors and adjust the length of the valid trajectory with small errors to obtain the corresponding second invalid length and first adjusted length, specifically as follows: Step 971: Perform a traversal of all the low-speed segments; during this traversal, the currently traversed low-speed segment is taken as the current segment; the rectangular area of ​​the road traversed by the current segment on the local high-precision map is taken as the current region; the current region is gridded according to a preset square grid size to obtain multiple corresponding region grids; the region grids that intersect with the trajectory of the current segment are recorded as the corresponding passing grids; the average error of the pairing points of all the third trajectory points of each passing grid is calculated and the calculation result is taken as the corresponding grid error; and it is identified whether the grid error of each passing grid is greater than the third distance threshold. If it is, the current grid is recorded as an invalid grid; otherwise, the current grid is recorded as a valid grid. Step 972: After this round of traversal, the first distance between each pair of adjacent third trajectory points in each invalid grid is summed and the result is used as the corresponding invalid grid trajectory length. The lengths of all the obtained invalid grid trajectories are summed and the result is used as the corresponding second invalid length. Step 973: Sum the first distances between adjacent third trajectory points in each of the effective grids and use the result as the corresponding original grid trajectory length; use the side length of the square grid of the square grid size as the unified network trajectory length; set the corresponding grid optimized trajectory length = max(0, original grid trajectory length - unified network trajectory length) based on each of the original grid trajectory lengths and the unified network trajectory length; and sum all the obtained grid optimized trajectory lengths and use the result as the corresponding first adjustment length. Step 98: The effective sensing trajectory length is obtained by subtracting the sum of the first invalid length, the second invalid length, and the first adjusted length from the initial trajectory length; Wherein, the effective sensing trajectory length = initial trajectory length - (first invalid length + second invalid length + first adjustment length); Step 99: The obtained effective sensing trajectory length is output as the result of this calculation.

10. The processing method for roadside sensing trajectory evaluation according to claim 2, characterized in that, The calculation of the corresponding sensing coverage rate based on the effective sensing trajectory length and the driving trajectory length of each of the drive-sensing trajectory pairs specifically includes: Each of the road-sense trajectory pairs is taken as the current trajectory pair; the road-sense trajectory of the current trajectory pair is taken as the corresponding current trajectory; the trajectory driving distance between every two first trajectory points of the current trajectory is estimated to obtain the corresponding third distance; the sum of all the third distances is taken as the corresponding road-sense trajectory length; and the corresponding sensing coverage rate is calculated, sensing coverage rate = (effective sensing trajectory length / road-sense trajectory length) × 100%.

11. An apparatus for performing the processing method for roadside sensing trajectory evaluation according to any one of claims 1-10, characterized in that, The device includes: an evaluation task receiving module, a data preparation module, a trajectory processing module, a coverage evaluation module, and an evaluation report processing module; The evaluation task receiving module is used to receive roadside perception trajectory evaluation task requests, denoted as the corresponding first task request; the first task request includes a first task identifier, a first evaluation time period, and a first intersection identifier; The data preparation module obtains the current intersection's current time period, perception trajectory set, and local high-precision map from the preset road test trajectory cloud, roadside perception cloud, and high-precision map cloud based on the first evaluation period and the first intersection identifier requested by the first task. It then converts the coordinates of all trajectory points in both trajectory sets to the high-precision map coordinate system corresponding to the local high-precision map. Based on a preset trajectory clipping mode, it performs trajectory clipping on both trajectory sets according to the local high-precision map. The road test trajectory set includes multiple road test trajectories; the perception trajectory set includes multiple perception trajectories; and the local high-precision map includes ROI elements, lane line elements, and stop line elements. The trajectory processing module is used to calculate the curvature of the trajectory points of each road test trajectory and segment the trajectory based on the calculation results to obtain the corresponding segmented trajectory sequence; and form a corresponding set of unmatched trajectories by the sensing trajectories in the sensing trajectory set that have trajectory time intersections with each segmented trajectory; and select the corresponding matching trajectory set for each segmented trajectory from its corresponding set of unmatched trajectories based on the DTW algorithm; and perform trajectory merging and deduplication processing on all the matching trajectory sets corresponding to each segmented trajectory sequence to obtain the corresponding merged trajectory set; and perform trajectory splicing based on each merged trajectory set according to a preset spatiotemporal continuity rule to obtain the corresponding sensing spliced ​​trajectory; The coverage assessment module is used to form corresponding road-sensor trajectory pairs from each of the road-test trajectories and their corresponding sensing splicing trajectories; calculate the effective sensing trajectory length of each of the road-test-sensor trajectory pairs according to the high-low speed adaptive invalid trajectory length identification method; calculate the corresponding sensing coverage rate based on the effective sensing trajectory length and road-test trajectory length of each of the road-test-sensor trajectory pairs; and take the average of all the sensing coverage rates as the corresponding roadside sensing coverage rate. The evaluation report processing module is used to form a corresponding roadside perception evaluation item by each of the road test-perception trajectory pairs and their corresponding perception coverage rates; and to form a corresponding roadside perception evaluation table by all the roadside perception evaluation items; and to form and save a corresponding roadside perception evaluation report by the first task request, the roadside perception evaluation table and the roadside perception coverage rate.

12. An electronic device, characterized in that, include: Memory, processor, and transceiver; The processor is configured to be coupled to the memory, read and execute instructions in the memory to implement the method according to any one of claims 1-10; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a computer, cause the computer to perform the method described in any one of claims 1-10.

Images