Map quality detection method, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a map quality detection method, electronic equipment and a storage medium, and the method comprises the steps: obtaining the trajectory data of a traffic participant in a preset scene, a predetermined true value map and a to-be-detected vectorized map, matching the trajectory data with a first map element in the true value map, obtaining labeled trajectory data, and storing the labeled trajectory data in a database; and determining a target trajectory where the first map element in the labeled trajectory data is matched with a second map element in the vectorized map, obtaining a trajectory subset of the second map element, performing space matching degree and time consistency detection on the trajectory subset of the second map element and the second map element, and obtaining a dynamic detection result of the vectorized map. And determining a quality detection result of the vectorized map according to the dynamic detection result. Quality detection is performed by integrating the dynamic characteristics of the map, and the accuracy of the map quality detection result is improved.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and in particular to a map quality detection method, electronic device, and storage medium. Background Technology

[0002] With the rapid evolution of intelligent driving technology, the importance of high-precision maps is becoming increasingly prominent, especially vectorized high-precision maps, which have become an indispensable part due to their compact data structure, rich semantic information, and ease of understanding and processing by autonomous driving systems.

[0003] Currently, map quality assessment is mainly conducted using the map evaluation protocol of the nuScenes dataset. Specifically, the map evaluation protocol evaluates the geometric deviations of elements in the map, such as lane lines and road boundaries. However, this quality assessment, which focuses only on the static geometric attributes of map elements, does not consider the impact of real-time and dynamically changing traffic flow on map elements and ignores the adaptability of the map to the actual interaction with dynamic traffic flow. This can easily lead to maps that meet the static geometric attribute standards but have a low degree of matching with actual traffic flow being misjudged as high-quality maps, affecting the accuracy of map quality detection and thus the reliability of autonomous driving system decisions. Summary of the Invention

[0004] This application provides a map quality detection method, electronic device, and storage medium, aiming to solve the problem that focusing solely on the static geometric attributes of map elements for quality assessment affects the accuracy of map quality detection, thereby impacting the decision-making reliability of autonomous driving systems, and achieving accurate and comprehensive map quality detection.

[0005] To address the aforementioned problems, this application discloses a map quality detection method, the method comprising: Acquire trajectory data of traffic participants in a preset scenario, a pre-determined ground truth map, and a vectorized map to be detected; The trajectory data is matched with a first map element in the truth map, and the matched first map element is marked in the trajectory data to obtain marked trajectory data; Determine the target trajectory that matches the first map element in the labeled trajectory data with the second map element in the vectorized map, and obtain the trajectory subset of the second map element; wherein, the first map element and the second map element are respectively elements in the ground truth map and the vectorized map used to represent the same traffic sign in the actual traffic environment; The spatial matching degree and temporal consistency of the trajectory subset of the second map element are detected with the second map element to obtain the dynamic detection result of the vectorized map; Based on the dynamic detection results, the quality detection result of the vectorized map is determined.

[0006] Based on the aforementioned technical means, by introducing trajectory data of traffic participants and associating and labeling trajectory data with map elements, the dynamic applicability of the map is evaluated from both spatial and temporal dimensions. Taking into account spatial constraints and temporal consistency, the applicability of the map in actual traffic flow is accurately assessed, resulting in quality inspection results that accurately reflect the dynamic characteristics of the map. This avoids the map being misjudged as a high-quality map due to a single static indicator evaluation, thus improving the accuracy of map quality inspection results and providing a reliable decision-making basis for autonomous driving systems, thereby enhancing the safety and reliability of autonomous driving.

[0007] Optionally, the method further includes: The second map element in the vectorized map is matched with the first map element in the ground truth map to obtain the map element matching relationship; Based on the map element matching relationship, the static detection result of the vectorized map is determined; wherein, the map element matching relationship includes a one-to-one matching of the second map element and the first map element; Based on the static detection results and the dynamic detection results, the quality detection result of the vectorized map is determined.

[0008] Optionally, the step of performing spatial matching and temporal consistency detection on the trajectory subset of the second map element and the second map element to obtain the dynamic detection result of the vectorized map includes: For each target trajectory in the trajectory subset of the second map element, the overlap between the trajectory points in the target trajectory and the second map element is detected to obtain the spatial matching degree of the second map element; Extract abnormal target trajectories from the trajectory subset of the second map element, obtain the abnormal time of the abnormal target trajectory and the update time of the second map element, determine the time difference between the abnormal time of the abnormal target trajectory and the update time, and obtain the time matching degree of the second map element. The spatial matching degree and the temporal matching degree of the second map element are weighted and summed to obtain the dynamic detection result of the vectorized map.

[0009] Based on the aforementioned technical means, the spatial and temporal matching degree between the trajectory and map elements is comprehensively evaluated. Taking into account both spatial constraints and temporal consistency, the dynamic applicability of the map is assessed from both spatial and temporal dimensions to ensure the accuracy of dynamic map detection.

[0010] Optionally, determining the quality inspection result of the vectorized map based on the static detection result and the dynamic detection result includes: Obtain the scene dynamic density parameters corresponding to the vectorized map; Based on the scene dynamic density parameter, determine the weight coefficients of the static detection result and the dynamic detection result; The static detection results and the dynamic detection results are weighted and summed according to the weighting coefficients to obtain the quality detection results of the vectorized map.

[0011] Based on the aforementioned technical means, the weights of static and dynamic detection results of the vectorized map are adjusted according to the dynamic characteristics of the scene to obtain a quality detection result that comprehensively reflects the static and dynamic characteristics of the map. This improves the accuracy of the map quality detection result, provides a reliable decision-making basis for the autonomous driving system, and thus enhances the safety and reliability of autonomous driving.

[0012] Optionally, the step of matching the trajectory data with a first map element in the truth map and marking the matching first map element in the trajectory data to obtain marked trajectory data includes: The trajectory data is converted to the coordinate system of the truth map, and the trajectory data and the truth map are spatiotemporally aligned. The spatiotemporally aligned trajectory data is matched with the first map element in the ground truth map to determine the first map element that matches the trajectory data. The first map element that matches the trajectory data is marked in the trajectory data to obtain marked trajectory data.

[0013] By incorporating the trajectory data of traffic participants and associating and labeling the trajectory data with map elements using the aforementioned technical means, the applicability of the map in actual traffic flow can be accurately assessed, providing effective data for subsequent dynamic detection and analysis.

[0014] Optionally, determining the target trajectory that matches the first map element in the labeled trajectory data with the second map element in the vectorized map, and obtaining the trajectory subset of the second map element, includes: For each second map element in the vectorized map, a first map element that matches the second map element is determined based on the map element matching relationship; Based on the labeled trajectory data, determine at least one target trajectory associated with the first map element; At least one of the target trajectories is determined as a subset of the trajectories of the second map element.

[0015] Based on the aforementioned technical means, the trajectory data associated with the vectorized map elements can be quickly identified through the matching and mapping relationship between dynamic trajectory data, ground truth map, and vectorized map elements. This provides a data foundation for subsequent spatial matching degree and temporal consistency detection, ensuring the comprehensiveness of dynamic detection.

[0016] Optionally, the step of matching the second map element in the vectorized map with the first map element in the ground truth map to obtain a map element matching relationship, and determining the static detection result of the vectorized map based on the map element matching relationship, includes: For each second map element in the vectorized map, it is matched with a first map element in the ground truth map to determine the first map element that matches the second map element, thus obtaining the map element matching relationship; Based on the map element matching relationship and the preset static index, calculate the static index value between each second map element and the first map element that matches the second map element; The static index values ​​of each second map element are averaged to obtain the static detection results of the vectorized map.

[0017] Based on the above technical means, multi-dimensional static index evaluation is performed on map elements in vectorized maps and map elements in ground truth maps to improve the static detection accuracy of vectorized maps, avoid the impact of anomalies in individual map elements on the whole, and ensure that the accuracy of the map in static aspects can support the analysis of dynamic traffic flow.

[0018] Optionally, the step of calculating the static index value between each second map element and the first map element that matches the second map element, based on the map element matching relationship and preset static indices, includes... Determine the geometric features of the first map element and the second map element; wherein, the geometric features include polygons and point sets; For each pair of first map elements that match the second map element, calculate the spatial overlap between the polygons, as well as the nearest neighbor distance and average symmetric distance between the point sets. The spatial overlap, nearest neighbor distance, and average symmetry distance of each second map element are weighted and summed to obtain the static index value between the second map element and the first map element that matches the second map element.

[0019] Based on the aforementioned technical means, the static detection accuracy of vectorized maps is improved by evaluating the spatial overlap, nearest neighbor distance, and average symmetric distance of map elements in vectorized maps and ground truth maps using multi-dimensional static indicators.

[0020] To address the aforementioned problems, this application also discloses a map quality inspection device, the device comprising: The data acquisition module is used to acquire trajectory data of traffic participants in a preset scenario, a pre-determined ground truth map, and a vectorized map to be detected. The first matching module is used to match the trajectory data with a first map element in the truth map, and mark the matching first map element in the trajectory data to obtain marked trajectory data. The second matching module is used to determine the target trajectory that matches the first map element in the labeled trajectory data with the second map element in the vectorized map, and to obtain a subset of the trajectory of the second map element; wherein the first map element and the second map element are elements in the ground truth map and the vectorized map, respectively, used to represent the same traffic sign in the actual traffic environment; The first detection module is used to perform spatial matching degree and temporal consistency detection on the trajectory subset of the second map element and the second map element to obtain the dynamic detection result of the vectorized map; The first result module is used to determine the quality detection result of the vectorized map based on the dynamic detection result.

[0021] To address the aforementioned problems, this application also discloses an electronic device, including a processor and a memory, wherein... Memory, used to store computer programs; The processor is used to execute the program stored in the memory to implement the map quality detection method described above.

[0022] To address the aforementioned issues, this application also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned map quality detection method.

[0023] To address the aforementioned issues, this application also discloses a vehicle designed to implement any of the map quality detection methods described above.

[0024] The map quality detection method provided in this application acquires trajectory data of traffic participants in a preset scenario, a pre-determined ground truth map, and a vectorized map to be detected. It then matches the trajectory data with a first map element in the ground truth map, marks the matching first map element in the trajectory data to obtain marked trajectory data, determines the target trajectory matching the first map element in the marked trajectory data with a second map element in the vectorized map, and obtains a subset of the trajectory of the second map element. The first map element and the second map element are elements in the ground truth map and the vectorized map, respectively, used to represent the same traffic sign in the actual traffic environment. The method performs spatial matching degree and temporal consistency detection on the subset of the trajectory of the second map element and the second map element to obtain the dynamic detection result of the vectorized map. Based on the dynamic detection result, it determines the quality detection result of the vectorized map. This application's embodiments introduce trajectory data of traffic participants, associate and label the trajectory data with map elements, and evaluate the dynamic applicability of the map from both spatial and temporal dimensions. By comprehensively considering spatial constraints and temporal consistency, the applicability of the map in actual traffic flow is accurately assessed, resulting in a quality inspection result that comprehensively reflects the dynamic characteristics of the map. This avoids the map being misjudged as a high-quality map due to a single static indicator evaluation, improves the accuracy of map quality inspection results, provides a reliable decision-making basis for autonomous driving systems, and thus improves the safety and reliability of autonomous driving.

[0025] Of course, any product implementing this application does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating the steps of a map quality detection method provided in an embodiment of this application; Figure 2 This is a flowchart of another map quality detection method provided in the embodiments of this application; Figure 3 This is a schematic flowchart of a map quality detection method provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a map quality detection device provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0027] To make the technical problems, technical solutions, and beneficial effects solved by this application clearer, the following detailed description is provided in conjunction with embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0028] Example 1 This application provides a map quality detection method. Please refer to the following embodiments. Figure 1 This includes the following steps: S110: Acquire trajectory data of traffic participants in a preset scenario, a pre-determined ground truth map, and a vectorized map to be detected.

[0029] In this embodiment, to address the problem that current quality assessments focus solely on the static geometric attributes of map elements, neglecting the impact of real-time, dynamically changing traffic flow on map elements and ignoring the adaptability of maps to actual interactions with dynamic traffic flow, maps with satisfactory static geometric attributes but low matching with actual traffic flow are often misjudged as high-quality maps, affecting the accuracy of map quality detection and thus the reliability of autonomous driving system decisions, this embodiment incorporates dynamic interaction consistency into the detection system. This enables more comprehensive and practical quality detection of vectorized high-precision maps, yielding quality detection results that comprehensively reflect both the static and dynamic characteristics of the map. This improves the accuracy of map quality detection results, provides a reliable decision-making basis for autonomous driving systems, and ultimately enhances the safety and reliability of autonomous driving.

[0030] It should be noted that the acquisition of traffic participant trajectory data across multiple scenarios utilizes a ground truth map as the detection benchmark, and performs multi-dimensional quality checks on the vectorized map to be detected. Traffic participant trajectory data refers to the actual movement trajectories of traffic participants (such as vehicles, pedestrians, and bicycles) in various scenarios. Examples include vehicle GPS (Global Positioning System) trajectories and pedestrian SLAM (Simultaneous Location and Mapping) trajectories. Trajectory data typically includes information such as trajectory points, timestamps, location coordinates, speed, and direction. Vehicle GPS trajectories refer to the motion data recorded by the GPS positioning system during vehicle operation. Pedestrian SLAM trajectories primarily rely on surrounding images, extracting image feature points and calculating trajectories based on image differences to determine their own position. The various scenarios include driving scenarios with varying degrees of vehicle movement, such as urban intersections, highways, and residential roads. They can also encompass environments like sunny days, rainy days, and nighttime. The ground truth map is a high-precision map, typically generated by professional surveying agencies or high-precision sensors, serving as a benchmark for evaluating other maps. Map elements in the ground truth map are considered accurate and include lane lines, traffic signs, curbs, and road boundaries. The vectorized map to be tested is the map requiring quality inspection. It is usually generated by autonomous driving systems or map providers. Map elements in the vectorized map need to be compared and matched with the ground truth map to assess its quality.

[0031] S120: Match the trajectory data with the first map element in the ground truth map, and mark the matching first map element in the trajectory data to obtain marked trajectory data.

[0032] In this embodiment, it is necessary to match the trajectory data of traffic participants with map elements in the ground truth map. Based on the matching results, the first matching map element is marked in the trajectory data to obtain marked trajectory data. The first map element refers to a map element in the ground truth map, including traffic sign elements such as lane lines, road elements, etc. Specifically, the trajectory data is converted to the coordinate system of the ground truth map, and the trajectory data and the ground truth map are spatiotemporally aligned. The spatiotemporally aligned trajectory data is then matched with the first map element in the ground truth map to determine the first map element that matches the trajectory data. This first matching map element is then marked in the trajectory data to obtain marked trajectory data.

[0033] In this embodiment, based on the location information in the trajectory data, it is determined which map element in the truth map each trajectory point is located on. The trajectory data is matched with the map elements in the truth map. After the matching is completed, the first map element that matches the trajectory data is marked in the trajectory data. That is, the map element corresponding to each trajectory is marked in the trajectory data to form marked trajectory data and stored.

[0034] S130: Determine the target trajectory that matches the first map element in the labeled trajectory data with the second map element in the vectorized map, and obtain the trajectory subset of the second map element.

[0035] The first map element and the second map element are elements used to represent the same traffic sign in the actual traffic environment in the ground truth map and the vectorized map, respectively.

[0036] In this embodiment, based on the first map element annotated in the labeled trajectory data and the second map element in the vectorized map, it is determined which trajectory data are associated with the second map element in the vectorized map. The vectorized map is described using basic geometric elements such as points, lines, and polygons and their attributes. The second map element refers to the map element in the vectorized map. The first map element and the second map element are elements in the ground truth map and the vectorized map, respectively, used to represent the same traffic sign in the actual traffic environment. The first map element and the second map element represent the same traffic sign in the actual traffic environment, such as lane lines, pedestrian crossings, road edge lines, arrows, etc. The second map element matches the first map element one-to-one. In other words, the first map element and the second map element are different representations of the same traffic sign in the ground truth map and the vectorized map, respectively. Specifically, the target trajectory that matches the first map element in the labeled trajectory data with the second map element in the vectorized map is determined, and a subset of the trajectory of the second map element is obtained. That is, the first map element in the ground truth map corresponding to the second map element in the vectorized map is found, and trajectory points related to the first map element in the ground truth map are extracted from the labeled trajectory data to form a subset of the trajectory of the second map element.

[0037] It should be noted that, in order to perform quality inspection on the vectorized map to be tested, elements in the ground truth map and the vectorized map that represent the same traffic sign in the actual traffic environment can be matched and detected. Specifically, the first map element in the ground truth map is marked in the trajectory data. The target trajectory that matches the first map element in the trajectory data with the second map element in the vectorized map is determined. That is, the trajectory associated with the second map element in the vectorized map is obtained. Then, the associated trajectory is detected with the second map element of the vectorized map to be tested, and the dynamic applicability of the map is evaluated.

[0038] S140: Perform spatial matching and temporal consistency checks on the trajectory subset of the second map element and the second map element to obtain the dynamic detection results of the vectorized map.

[0039] In this embodiment, spatial matching and temporal consistency checks are performed on the trajectory subset of the second map elements and the second map elements to obtain the dynamic detection result of the vectorized map. Specifically, for each target trajectory in the trajectory subset of the second map elements, the spatial matching and temporal matching of each target trajectory with the second map elements are calculated. The overlap between trajectory points in the target trajectory and the second map elements is detected to obtain the spatial matching of the second map elements. Abnormal target trajectories are extracted from the trajectory subset of the second map elements, and the abnormal time of the abnormal target trajectories and the update time of the second map elements are obtained. The time difference between the abnormal time and the update time of the abnormal target trajectories is determined to obtain the temporal matching of the second map elements. The spatial matching and temporal matching of the second map elements are then weighted and summed to obtain the dynamic detection result of the vectorized map. Taking into account both spatial constraints and temporal consistency, the dynamic applicability of the map is evaluated from both spatial and temporal dimensions.

[0040] S150: Determine the quality inspection results of the vectorized map based on the dynamic detection results.

[0041] Based on the dynamic detection results, the embodiments of this application determine the quality detection results of the vectorized map. The dynamic detection results reflect the applicability of map elements in actual traffic flow. The final quality detection results comprehensively reflect the quality detection results of the dynamic characteristics of the map, improve the accuracy of the map quality detection results, and provide a reliable decision-making basis for the autonomous driving system.

[0042] This application's embodiments introduce trajectory data of traffic participants, associate and label the trajectory data with map elements, and evaluate the dynamic applicability of the map from both spatial and temporal dimensions. By comprehensively considering spatial constraints and temporal consistency, the applicability of the map in actual traffic flow is accurately assessed, resulting in a quality inspection result that comprehensively reflects the dynamic characteristics of the map. This avoids the map being misjudged as a high-quality map due to a single static indicator evaluation, improves the accuracy of map quality inspection results, provides a reliable decision-making basis for autonomous driving systems, and thus improves the safety and reliability of autonomous driving.

[0043] Example 2 This application provides another map quality detection method, please refer to... Figure 2 This includes the following steps: S110: Acquire trajectory data of traffic participants in a preset scenario, a pre-determined ground truth map, and a vectorized map to be detected.

[0044] S120: Match the trajectory data with the first map element in the ground truth map, and mark the matching first map element in the trajectory data to obtain marked trajectory data.

[0045] S160: Match the second map element in the vectorized map with the first map element in the ground truth map to obtain the map element matching relationship; S170, determine the static detection result of the vectorized map based on the map element matching relationship; wherein, the map element matching relationship includes a one-to-one matching of the second map element and the first map element.

[0046] In this embodiment, a second map element in the vectorized map to be detected is matched with a first map element in the ground truth map to obtain a map element matching relationship. The second map element refers to a map element in the vectorized map to be detected, and the map element matching relationship is the matching relationship between the second map element in the vectorized map and the first map element in the ground truth map. The map element matching relationship indicates the matching relationship between map elements in the ground truth map and map elements in the vectorized map. The map element matching relationship includes one-to-one matching of second map elements and first map elements. For example, the map element matching relationship may include second map elements and first map elements representing the same traffic sign. The purpose of matching is to determine the correspondence between the two. For example, the map element matching relationship may include a lane line in the vectorized map corresponding to a lane line in the ground truth map.

[0047] In this embodiment, for each second map element in the vectorized map, it is matched with the first map element in the ground truth map to determine the first map element that matches the second map element, thus obtaining the map element matching relationship. Based on the map element matching relationship and preset static indicators, the static indicator value between each second map element and the first map element that matches the second map element is calculated. The preset static indicators include the spatial overlap of map elements, nearest neighbor distance, and average symmetry distance. The static indicator value of each second map element is averaged to obtain the static detection result of the vectorized map.

[0048] S130: Determine the target trajectory that matches the first map element in the labeled trajectory data with the second map element in the vectorized map, and obtain the trajectory subset of the second map element; The first map element and the second map element are elements used to represent the same traffic sign in the actual traffic environment in the ground truth map and the vectorized map, respectively.

[0049] S140: Perform spatial matching and temporal consistency checks on the trajectory subset of the second map element and the second map element to obtain the dynamic detection results of the vectorized map.

[0050] In this embodiment, steps 110 to 140 are as described in Embodiment 1, and will not be repeated here.

[0051] S180, Based on the static and dynamic detection results, determine the quality detection results of the vectorized map.

[0052] This application embodiment integrates static and dynamic detection results to determine the quality detection result of the vectorized map. The static detection result reflects the geometric accuracy of map elements, while the dynamic detection result reflects the applicability of map elements in actual traffic flow. By weighted summing of the static and dynamic detection results, the final quality detection result comprehensively reflects the quality detection results of both the static and dynamic characteristics of the map, thereby improving the accuracy of the map quality detection result and providing a reliable decision-making basis for the autonomous driving system.

[0053] The map quality detection method provided in this application introduces trajectory data of traffic participants, associates and labels trajectory data and map elements, and evaluates the dynamic applicability of the map from both spatial and temporal dimensions based on multi-dimensional static index evaluation of map elements in vectorized maps and ground truth maps. By comprehensively considering spatial constraints and temporal consistency, it accurately evaluates the applicability of the map in actual traffic flow, and obtains a quality detection result that comprehensively reflects the static and dynamic characteristics of the map. This avoids the map being misjudged as a high-quality map due to a single static index evaluation, improves the accuracy of map quality detection results, provides a reliable decision-making basis for autonomous driving systems, and thus improves the safety and reliability of autonomous driving.

[0054] To facilitate understanding of the map quality detection method provided in this application by those skilled in the art, please refer to... Figure 3 , Figure 3 This illustration shows a flowchart of a map quality detection method provided in an embodiment of this application. Specifically, it involves: acquiring traffic participant trajectory data from multiple scenarios as input; using a ground truth map as the detection benchmark; and performing multi-dimensional quality detection on the vectorized map to be detected. The multi-dimensional quality detection includes static accuracy scoring. Calculation and dynamic interactive scoring Calculation. First, a dynamic trajectory database is constructed by collecting multi-dimensional dynamic data covering scenarios such as urban intersections (high dynamic), highways (medium dynamic), and residential roads (low dynamic), including environments such as sunny days, rainy days, and nighttime, to ensure the diversity of trajectory data. The spatial coordinates of the trajectory data are then transformed to match the ground truth. Figure 1 A unified local coordinate system is used, with timestamps standardized to the millisecond level to ensure temporal and spatial correspondence between trajectories and map elements. Each trajectory is labeled with its corresponding map element, such as "trajectory". Associated lane lines "Trajectory" Related pedestrian crossings "etc.", forming a labeled dynamic trajectory library for subsequent interaction consistency analysis. Static accuracy scoring. By calculating single-element indicators, including weighted summation of indicators such as spatial overlap of map elements, nearest neighbor distance, and average symmetry distance, dynamic interactive scoring is achieved. By extracting associated trajectories, the spatial and temporal matching degrees are weighted and summed for output, and the final weight coefficients are determined to generate a dynamic interactive rating for the vectorized map. and static accuracy scoring The weighted sum is used to obtain the overall score. This serves as the quality inspection result for the vectorized map.

[0055] In some embodiments of this application, step S140, which performs spatial matching and temporal consistency detection on the trajectory subset of the second map element and the second map element to obtain the dynamic detection result of the vectorized map, may include: Sub-step 1401: For each target trajectory in the trajectory subset of the second map element, detect the overlap between the trajectory points in the target trajectory and the second map element to obtain the spatial matching degree of the second map element; Sub-step 1402: Extract abnormal target trajectories from the trajectory subset of the second map elements, obtain the abnormal time of the abnormal target trajectories and the update time of the second map elements, determine the time difference between the abnormal time and the update time of the abnormal target trajectories, and obtain the time matching degree of the second map elements. Sub-step 1403: The spatial matching degree and temporal matching degree of the second map element are weighted and summed to obtain the dynamic detection result of the vectorized map.

[0056] In this embodiment, for each target trajectory in the trajectory subset of the second map element, the spatial matching degree and temporal matching degree between each target trajectory and the second map element are calculated. Specifically, the overlap between trajectory points in the target trajectory and the second map element is detected to obtain the spatial matching degree of the second map element. Abnormal target trajectories are extracted from the trajectory subset of the second map element, and the abnormal time of the abnormal target trajectory and the update time of the second map element are obtained. The time difference between the abnormal time and the update time of the abnormal target trajectory is determined to obtain the temporal matching degree of the second map element. The spatial matching degree and temporal matching degree of the second map element are then weighted and summed to obtain the dynamic detection result of the vectorized map. The weighting weights are set and adjusted according to actual needs and are not limited here. The value range of the dynamic detection result is normalized to 0-1. The higher the dynamic detection result, the better the dynamic adaptability of the map.

[0057] As a specific implementation of this application, the "proportion of trajectory points within the map element's range" is designed for linear elements such as lane lines in the map elements, and the "overlapping area ratio between trajectory and map element" is designed for regional elements (such as pedestrian crossings) in the map elements. This directly quantifies whether map elements conform to the driving patterns of real traffic flow. For example, for pedestrian crossings, even if their static polygon interaction ratio reaches 0.9, if 90% of pedestrian trajectories do not fall within this area, the dynamic detection result will reflect that their "actual guidance has failed." For temporary lane lines such as construction detour routes, if most vehicle trajectories follow this lane line, even if their static accuracy is slightly lower, a high dynamic detection result can still be obtained, reflecting the actual value of the map. The timeliness of the map is quantified by matching the abnormal trajectory, such as a vehicle suddenly deviating from its original lane, with the map update time. For example, if road construction causes abnormal vehicle trajectories... The update time of the construction area boundary in the map Later If there is a 5-minute delay, the time matching degree will decrease, reflecting that the map has not captured scene changes in time. For static scenes such as ordinary roads, the time matching degree is always 1, which does not affect the map detection results and ensures multi-scene adaptability.

[0058] Specifically, using the second map element Trajectory subset Taking the example of a subset of trajectories, we can illustrate this further. For each target trajectory in the map, calculate its relationship with the second map element. Spatial matching degree: If the map element is a linear shape, such as a lane line, the trajectory points of the target trajectory fall on the map element. The proportion within the outline range, such as within 30cm above and below the lane line width, is considered an effective range. A higher proportion indicates a higher spatial matching degree. If the map element is a region shape, such as a pedestrian crossing, the proportion of the overlap area between the trajectory points of the target trajectory and the pedestrian crossing region to the total length of the target trajectory is calculated. A higher proportion indicates a higher spatial matching degree. For the second map element... The spatial matching degree of all target trajectories in the trajectory subset is averaged to obtain the spatial matching degree of the second map element, which is used as the spatial constraint score. , The normalized range is 0-1.

[0059] In practice, for dynamic scenarios, such as construction areas, second map elements are extracted. The abnormal target trajectories are those of vehicles and other traffic participants that deviate from their original target trajectories within a subset of the trajectories. The abnormal start time of these abnormal target trajectories is then obtained. Simultaneously acquire second map elements Such as the update time of the construction area boundary. The time difference between the abnormal time and the update time of the abnormal target trajectory is determined to obtain the time matching degree of the second map element. For example, if... This indicates that the map update occurred earlier than or simultaneously with the anomaly. The higher the time matching degree, the better. The time matching degree decreases linearly with the increase of time difference. The average time matching degree of all abnormal target trajectories is taken to obtain the time matching degree of the second map element, which is used as the time consistency score. , The normalized range is 0-1, and the score for static scenes such as ordinary roads is always 1.

[0060] In this embodiment, the spatial matching degree and temporal matching degree of the second map element are weighted and summed to obtain the dynamic detection result of the vectorized map. Specifically, the spatial constraint score corresponding to the spatial matching degree of the second map element is calculated. The time consistency score corresponding to the time matching degree of the second map element. Weighted summation is performed. If the influence of spatial constraints is greater than that of time, the spatial constraint score corresponding to the spatial matching degree has a higher weight. For example, the dynamic interaction score corresponding to the dynamic detection result of the vectorized map. .

[0061] This application comprehensively evaluates the spatial and temporal matching degree between the trajectory and map elements, taking into account both spatial constraints and temporal consistency, and assesses the dynamic applicability of the map from both spatial and temporal dimensions to ensure the accuracy of dynamic map detection.

[0062] In some embodiments of this application, S180 determines the quality inspection result of the vectorized map based on the static inspection result and the dynamic inspection result, which may specifically include the following steps. Sub-step 1801: Obtain the scene dynamic density parameters of the scene corresponding to the vectorized map; Sub-step 1802: Determine the weighting coefficients of static and dynamic detection results based on the scene dynamic density parameters; Sub-step 1803: The static detection results and dynamic detection results are weighted and summed according to the weight coefficients to obtain the quality detection results of the vectorized map.

[0063] In this embodiment, the scene dynamic density parameter of the vectorized map corresponding to the scene is obtained. The scene dynamic density parameter is set according to the real-time density of the scene and is a quantitative coefficient used to balance the proportion of static detection results and dynamic detection results. For example, if there are many traffic participants at urban intersections and the dynamics are strong, the scene dynamic density parameter will be higher. The scene dynamic density parameter is not only applicable to urban intersections, but also to different scenarios such as highways and residential roads.

[0064] In this embodiment, the weighting coefficients of static and dynamic detection results are determined based on the scene dynamic density parameter of the scene corresponding to the vectorized map. For example, the higher the scene dynamic density parameter, the higher the weighting coefficient of the dynamic detection result. The larger the value, the higher the weighting coefficient of the static detection result (1- The smaller the value of α, the lower the scene dynamic density parameter, and the smaller the weight coefficient α of the dynamic detection result, and the lower the weight coefficient α of the static detection result (1- The larger the value, the greater the sum of the weight coefficients of the static and dynamic detection results. The static and dynamic detection results are weighted and summed according to the weight coefficients to obtain the quality detection result of the vectorized map. The quality detection result of the vectorized map has a value range of 0-1, and the higher the value, the better the overall quality of the map.

[0065] It should be noted that the scene dynamic density parameter is a feature indicator describing and evaluating the "dynamic degree" of a scene. It can be qualitatively labeled as high, medium, or low, or quantitatively described as the number of traffic participants per square kilometer per hour; the weighting coefficient of the dynamic detection results It is based on this characteristic index and is used to balance the static accuracy score corresponding to the static detection results. Dynamic interactive scoring corresponding to dynamic detection results The quantification coefficients of the proportions are used to establish a one-to-one correspondence between the two through preset rules, such as high scene dynamic density corresponding to... Take 0.7, which corresponds to the scene dynamic density. A value of 0.6 corresponds to low scene dynamic density. Take 0.4, It is a specific quantitative representation of the dynamic density of the scene in the comprehensive scoring formula, ensuring that the score can adapt to the core needs of different scenes.

[0066] Specifically, the static and dynamic detection results are weighted and summed according to weighting coefficients to obtain the quality inspection result of the vectorized map. The static accuracy score corresponding to the static detection results can be obtained. Dynamic interactive scoring corresponding to dynamic detection results We perform a weighted summation to obtain the overall score. , The value range is 0-1. A higher value indicates a better overall map quality. The specific value can be calculated using the following formula:

[0067] in, This is the overall score corresponding to the quality inspection results of the vectorized map. The dynamic interactive score corresponds to the dynamic detection results. This refers to the static accuracy score corresponding to the static detection results. This refers to the weighting coefficient of the dynamic interaction score corresponding to the dynamic detection results.

[0068] The embodiments of this application adjust the weights of the static and dynamic detection results of the vectorized map according to the dynamic characteristics of the scene, so as to obtain a quality detection result that comprehensively reflects the static and dynamic characteristics of the map, thereby improving the accuracy of the map quality detection result, providing a reliable decision basis for the autonomous driving system, and thus improving the safety and reliability of autonomous driving.

[0069] In some embodiments of this application, step S120, which matches the trajectory data with a first map element in the ground truth map and marks the matching first map element in the trajectory data to obtain marked trajectory data, may include the following steps: Sub-step 1201: Convert the trajectory data to the coordinate system of the ground truth map, and perform spatiotemporal alignment between the trajectory data and the ground truth map; Sub-step 1202: Match the spatiotemporally aligned trajectory data with the first map element in the ground truth map to determine the first map element that matches the trajectory data; Sub-step 1203: Mark the first map element that matches the trajectory data in the trajectory data to obtain marked trajectory data.

[0070] In this embodiment, to annotate trajectory data based on a ground truth map for subsequent interaction consistency analysis, the trajectory data is first converted to the coordinate system of the ground truth map. The trajectory data and the ground truth map are then spatiotemporally aligned to ensure temporal and spatial correspondence between the trajectory and map elements in the ground truth map. The spatiotemporally aligned trajectory data is then matched with a first map element in the ground truth map to determine the first map element that matches the trajectory data. Based on the spatiotemporally aligned data, the trajectory is matched with the first map element. After matching, the first map element that matches the trajectory data is annotated in the trajectory data, resulting in annotated trajectory data. The matched map element labels (such as "lane lines") are then added. The labeled trajectory data is then attached to the corresponding trajectory points to form labeled trajectory data. This labeled trajectory data is stored in a dynamic trajectory library for subsequent analysis.

[0071] This application embodiment introduces trajectory data of traffic participants and associates and labels the trajectory data with map elements, which can accurately evaluate the applicability of the map in actual traffic flow and provide effective data for subsequent dynamic detection and analysis.

[0072] In some embodiments of this application, step S130 determines the target trajectory that matches the first map element in the labeled trajectory data and the second map element in the vectorized map, obtaining a subset of trajectories of the second map element, including the following steps: Sub-step 1301: For each second map element in the vectorized map, determine the first map element that matches the second map element based on the map element matching relationship; Sub-step 1302: Based on the labeled trajectory data, determine at least one target trajectory associated with the first map element; Sub-step 1303: Determine at least one target trajectory as a subset of the trajectories of the second map element.

[0073] In this embodiment, for each second map element in the vectorized map, a first map element matching the second map element is determined based on the map element matching relationship. That is, based on the map element matching relationship, the first map element corresponding to each second map element in the vectorized map and the ground truth map are determined. Based on the matching relationship, at least one target trajectory associated with the first map element is extracted from the labeled trajectory data in the dynamic trajectory library, and these target trajectories are determined as a subset of the trajectories of the second map element.

[0074] It should be noted that evaluating maps solely based on static geometric metrics may result in situations where "the map is accurate but disconnected from real traffic flow." For example, lane lines may meet static accuracy standards, but actual vehicles may deviate from those lanes. This embodiment addresses this by introducing real trajectory data of traffic participants and establishing a correlation between trajectories and map elements. This expands the evaluation criteria from "geometric fit" to "actual behavior matching." For instance, if a lane line has a static geometric metric of 0.8 (meets the standard), but 80% of vehicles deviate from that lane line, its dynamic score will significantly decrease. This prevents it from being misjudged as a high-quality map and ensures that the evaluation results directly support the reliability of autonomous driving decisions. For example, this applies to each second map element in the vectorized map, such as lane lines. Based on the matching relationship of map elements, determine the relationship with the second map element, such as lane lines. The first matching map element is the lane line. Filter out lane lines from the labeled trajectory data in the dynamic trajectory library. Related trajectories, i.e., those marked The trajectory is used to obtain the lane lines. Exclusive Trajectory Subset .

[0075] This application embodiment quickly identifies trajectory data associated with vectorized map elements by matching and mapping dynamic trajectory data, ground truth maps, and vectorized map elements. This provides a data foundation for subsequent spatial matching and temporal consistency detection, ensuring the comprehensiveness of dynamic detection.

[0076] In some embodiments of this application, the method further includes the following steps: Sub-step 01: For each second map element in the vectorized map, match it with the first map element in the ground truth map to determine the first map element that matches the second map element and obtain the map element matching relationship; Sub-step 02: Based on the map element matching relationship and the preset static index, calculate the static index value between each second map element and the first map element that matches the second map element; Sub-step 03: Average the static index values ​​of each second map element to obtain the static detection results of the vectorized map.

[0077] In this embodiment, for each second map element in the vectorized map, each second map element in the vectorized map is matched with a first map element in the ground truth map to determine the first map element that matches the second map element. The matching relationship between the second map element and the first map element is stored to obtain the map element matching relationship. The map element matching relationship can be stored as a mapping table. This embodiment does not make specific limitations on this. Based on the map element matching relationship and the preset static index, the static index value between each second map element and the first map element that matches the second map element is calculated.

[0078] In the specific implementation, the preset static indicators include the spatial overlap of map elements, nearest neighbor distance, and average symmetric distance. Spatial overlap is calculated by polygons. Intersection over Union (IoU) is used to measure the degree of spatial overlap between two polygonal regions of arbitrary shapes. It is a core metric for detecting the accuracy of vector contours of map elements such as lanes and intersections in vectorized maps. Nearest neighbor distance is determined by the Hausdorff distance between the point sets of the first and second map elements. The Hausdorff distance measures the maximum mismatch between the two point sets, i.e., the maximum nearest neighbor distance. Average Symmetric Distance (ASD) measures the average of all nearest neighbor distances between the point sets of the first and second map elements, reflecting the overall closeness of the two shapes.

[0079] In this embodiment, the static index value between each second map element and the first map element that matches the second map element is calculated by weighted summation based on the spatial overlap, nearest neighbor distance and average symmetry distance of map elements. Each static index value is normalized to the range of 0-1 to ensure the comparability between different indicators and avoid the influence of the difference in units on the map quality detection results. Then, the static index value of each second map element is averaged to obtain the static detection result of the vectorized map.

[0080] This application embodiment performs multi-dimensional static index evaluation on map elements in vectorized maps and map elements in ground truth maps, improves the static detection accuracy of vectorized maps, avoids the impact of anomalies in individual map elements on the whole, and ensures that the accuracy of the map in static aspects can support the analysis of dynamic traffic flow.

[0081] In some embodiments of this application, sub-step 02, calculating the static index value between each second map element and the first map element that matches the second map element based on the map element matching relationship and preset static indices, may include the following steps: Determine the geometric features of the first map element and the second map element; wherein, the geometric features include polygons and point sets; For each pair of second map elements that match the first map element, calculate the spatial overlap between polygons, as well as the nearest neighbor distance and average symmetric distance between point sets. The spatial overlap, nearest neighbor distance, and average symmetric distance of each second map element are weighted and summed to obtain the static index value between the second map element and the first map element that matches the second map element.

[0082] In this embodiment of the application, the static accuracy of the map is first tested and evaluated. This requires the geometric relationship between the map elements of the vectorized map and the corresponding map elements in the ground truth map. Therefore, the geometric features of the first map element and the second map element are determined. The geometric features include polygons and point sets. For each pair of second map elements and the first map element that matches the second map element, the spatial overlap between polygons and the nearest neighbor distance and average symmetry distance between point sets are calculated respectively.

[0083] In the specific implementation, spatial overlap is calculated by polygons. The process involves determining the intersection and union areas of the polygons corresponding to the first map element in each pair of second map elements, represented by a sequence of coordinate points. The intersection-union ratio is then calculated based on the ratio of the intersection area to the union area, yielding the spatial overlap between the polygons. Specifically, the spatial overlap is calculated by... Confirmed, polygon The calculation formula is as follows:

[0084] in, The ratio of the intersection area to the union area of ​​the polygons corresponding to the first map element that matches the second map element is used to represent the degree of spatial overlap.

[0085] In this embodiment, the nearest neighbor distance and average symmetric distance of the point set corresponding to the first map element that matches the second map element are specifically calculated by: taking point set A in the vectorized map and point set B in the ground truth map, first calculating the maximum nearest neighbor distance from each point in A to B. The maximum nearest neighbor distance from each point in B to A. The maximum of the two values ​​is taken, and then normalized to convert it into a positive index value. This yields the nearest neighbor distance (Hausdorff distance) between the points of the first map element that matches the second map element. This distance measures the extreme deviation between the two point sets. The average symmetric distance is calculated by first calculating the average nearest neighbor distance from A to B. The average nearest neighbor distance from B to A Then, the average of the two values ​​is taken, and the result is converted into a positive index value through normalized distance. Specifically, it can be calculated using the following formula:

[0086] in, The average symmetric distance between the points of the first map element that match each pair of second map elements is the same as the distance between the points of the first map element that match the second map element. Let A be the average nearest neighbor distance from B. Let be the average nearest neighbor distance from B to A.

[0087] In this embodiment, static index weights are pre-set based on the degree of influence of map elements on dynamic interaction. The weights of the three static indices are summed to one. For example, the weight of lane line spatial overlap is 0.4, the weight of nearest neighbor distance is 0.3, and the weight of average symmetry distance is 0.3; the weight of pedestrian crossing spatial overlap is 0.5, the weight of nearest neighbor distance is 0.2, and the weight of average symmetry distance is 0.3. The static index values ​​of a single map element are weighted and summed to obtain the static index value of that map element. The arithmetic mean of the static index values ​​of all map elements is taken to obtain the static accuracy score corresponding to the static detection accuracy of the final vectorized map. .

[0088] This application embodiment improves the static detection accuracy of vectorized maps by evaluating the spatial overlap, nearest neighbor distance, and average symmetric distance of map elements in vectorized maps and ground truth maps using multi-dimensional static indicators.

[0089] This application also provides a map quality detection device 40, please refer to... Figure 4 The device includes: The data acquisition module 410 is used to acquire trajectory data of traffic participants in a preset scenario, a pre-determined ground truth map, and a vectorized map to be detected. The first matching module 420 is used to match the trajectory data with a first map element in the truth map, and mark the matched first map element in the trajectory data to obtain marked trajectory data. The second matching module 430 is used to determine the target trajectory that matches the first map element in the labeled trajectory data with the second map element in the vectorized map, and to obtain a subset of the trajectory of the second map element; wherein the first map element and the second map element are elements in the ground truth map and the vectorized map, respectively, used to represent the same traffic sign in the actual traffic environment; The first detection module 440 is used to perform spatial matching degree and temporal consistency detection on the trajectory subset of the second map element and the second map element to obtain the dynamic detection result of the vectorized map. The first result module 450 is used to determine the quality detection result of the vectorized map based on the static detection result and the dynamic detection result.

[0090] Optionally, the device further includes: The third matching module is used to match the second map element in the vectorized map with the first map element in the ground truth map to obtain the map element matching relationship; The second detection module is used to determine the static detection result of the vectorized map based on the map element matching relationship; wherein, the map element matching relationship includes a one-to-one matching of the second map element and the first map element; The second result module is used to determine the quality detection result of the vectorized map based on the static detection result and the dynamic detection result.

[0091] Optionally, the first detection module 440 includes: The detection submodule is used to detect the overlap between the trajectory points in the target trajectory and the second map element for each target trajectory in the trajectory subset of the second map element, and obtain the spatial matching degree of the second map element. The first processing submodule is used to extract abnormal target trajectories from the trajectory subset of the second map element, obtain the abnormal time of the abnormal target trajectory and the update time of the second map element, determine the time difference between the abnormal time of the abnormal target trajectory and the update time, and obtain the time matching degree of the second map element. The second processing submodule is used to perform a weighted summation of the spatial matching degree and the temporal matching degree of the second map element to obtain the dynamic detection result of the vectorized map.

[0092] Optionally, the second result module 430 includes: The acquisition submodule is used to acquire the scene dynamic density parameters of the scene corresponding to the vectorized map; The first determining submodule is used to determine the weighting coefficients of the static detection result and the dynamic detection result based on the scene dynamic density parameter; The second processing submodule is used to sum the static detection results and the dynamic detection results according to the weighting coefficients to obtain the quality detection results of the vectorized map.

[0093] Optionally, the first matching module 420 includes: The third processing submodule is used to convert the trajectory data to the coordinate system of the truth map and perform spatiotemporal alignment of the trajectory data and the truth map. The first matching submodule is used to match the spatiotemporally aligned trajectory data with the first map element in the truth map to determine the first map element that matches the trajectory data. The annotation submodule is used to annotate the trajectory data with the first map element that matches the trajectory data, so as to obtain annotated trajectory data.

[0094] Optionally, the second matching module 440 includes: The second determining submodule is used to determine, for each second map element in the vectorized map, a first map element that matches the second map element based on the map element matching relationship; The third determining submodule is used to determine at least one target trajectory associated with the first map element based on the labeled trajectory data; The fourth determination submodule is used to determine at least one of the target trajectories as a subset of the trajectories of the second map element.

[0095] Optionally, the third matching module includes: The second matching submodule is used to match each second map element in the vectorized map with a first map element in the truth map, determine the first map element that matches the second map element, and obtain the map element matching relationship. The first calculation submodule is used to calculate the static index value between each second map element and the first map element that matches the second map element, based on the map element matching relationship and the preset static index. The fourth processing submodule is used to perform mean processing on the static index values ​​of each of the second map elements to obtain the static detection results of the vectorized map.

[0096] Optionally, the first computing submodule includes A determining unit is used to determine the geometric features of the first map element and the second map element; wherein the geometric features include polygons and point sets; The calculation unit is used to calculate the spatial overlap between the polygons, the nearest neighbor distance and the average symmetric distance between the point sets for each pair of first map elements that match the second map element. The processing unit is used to perform a weighted summation of the spatial overlap, nearest neighbor distance, and average symmetry distance of each second map element to obtain a static index value between the second map element and the first map element that matches the second map element.

[0097] The map quality detection device provided in this application acquires trajectory data of traffic participants in a preset scenario, a pre-determined ground truth map, and a vectorized map to be detected. It matches the trajectory data with a first map element in the ground truth map to obtain labeled trajectory data. It then determines the target trajectory that matches the first map element in the labeled trajectory data with a second map element in the vectorized map, thus obtaining a subset of the trajectory of the second map element. The first map element and the second map element are elements in the ground truth map and the vectorized map, respectively, used to represent the same traffic sign in the actual traffic environment. The device performs spatial matching degree and temporal consistency detection on the subset of the trajectory of the second map element and the second map element to obtain the dynamic detection result of the vectorized map. Based on the dynamic detection result, it determines the quality detection result of the vectorized map. This application's embodiments introduce trajectory data of traffic participants, associate and label the trajectory data with map elements, and evaluate the dynamic applicability of the map from both spatial and temporal dimensions. By comprehensively considering spatial constraints and temporal consistency, the applicability of the map in actual traffic flow is accurately assessed, resulting in a quality inspection result that comprehensively reflects the dynamic characteristics of the map. This avoids the map being misjudged as a high-quality map due to a single static indicator evaluation, improves the accuracy of map quality inspection results, provides a reliable decision-making basis for autonomous driving systems, and thus improves the safety and reliability of autonomous driving.

[0098] This application also provides an electronic device 50, please refer to... Figure 5 It includes a processor 510 and a memory 520, wherein the memory 510 is used to store computer programs; the processor 520 is used to execute the programs stored in the memory 510 to implement the map quality detection method described in any embodiment of this application.

[0099] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the map quality detection method described in any embodiment of this application.

[0100] This application also provides a vehicle for implementing the map quality detection method described in any of the above embodiments.

[0101] In this application, "multiple" refers to two or more.

[0102] In this application, unless otherwise expressly defined, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0103] The terms “first,” “second,” “third,” “fourth,” etc., in this application (if present) are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0104] In this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, in this application, the character "" generally indicates that the preceding and following related objects have an "or" relationship.

[0105] Unless otherwise specified, all steps in this application may be performed sequentially or randomly. For example, if the method includes steps A and B, it means that the method may include steps A and B performed sequentially, or it may include steps B and A performed sequentially. For example, if the method may also include step C, it means that step C may be added to the method in any order. For example, the method may include steps A, B, and C, or it may include steps A, C, and B, or it may include steps C, A, and B, etc.

[0106] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A map quality inspection method, characterized in that, The method includes: Acquire trajectory data of traffic participants in a preset scenario, a pre-determined ground truth map, and a vectorized map to be detected; The trajectory data is matched with a first map element in the truth map, and the matched first map element is marked in the trajectory data to obtain marked trajectory data; Determine the target trajectory that matches the first map element in the labeled trajectory data with the second map element in the vectorized map, and obtain the trajectory subset of the second map element; wherein, the first map element and the second map element are respectively elements in the ground truth map and the vectorized map used to represent the same traffic sign in the actual traffic environment; The spatial matching degree and temporal consistency of the trajectory subset of the second map element are detected with the second map element to obtain the dynamic detection result of the vectorized map; Based on the dynamic detection results, the quality detection result of the vectorized map is determined.

2. The method according to claim 1, characterized in that, The method further includes: The second map element in the vectorized map is matched with the first map element in the ground truth map to obtain the map element matching relationship; Based on the map element matching relationship, the static detection result of the vectorized map is determined; wherein, the map element matching relationship includes a one-to-one matching of the second map element and the first map element; Based on the static detection results and the dynamic detection results, the quality detection result of the vectorized map is determined.

3. The method according to claim 1 or 2, characterized in that, The step of performing spatial matching and temporal consistency detection on the trajectory subset of the second map element and the second map element to obtain the dynamic detection result of the vectorized map includes: For each target trajectory in the trajectory subset of the second map element, the overlap between the trajectory points in the target trajectory and the second map element is detected to obtain the spatial matching degree of the second map element; Extract abnormal target trajectories from the trajectory subset of the second map element, obtain the abnormal time of the abnormal target trajectory and the update time of the second map element, determine the time difference between the abnormal time of the abnormal target trajectory and the update time, and obtain the time matching degree of the second map element. The spatial matching degree and the temporal matching degree of the second map element are weighted and summed to obtain the dynamic detection result of the vectorized map.

4. The method according to claim 1 or 2, characterized in that, Determining the quality inspection result of the vectorized map based on the static inspection result and the dynamic inspection result includes: Obtain the scene dynamic density parameters corresponding to the vectorized map; Based on the scene dynamic density parameter, determine the weight coefficients of the static detection result and the dynamic detection result; The static detection results and the dynamic detection results are weighted and summed according to the weighting coefficients to obtain the quality detection results of the vectorized map.

5. The method according to claim 1 or 2, characterized in that, The step of matching the trajectory data with a first map element in the ground truth map and marking the matching first map element in the trajectory data to obtain marked trajectory data includes: The trajectory data is converted to the coordinate system of the truth map, and the trajectory data and the truth map are spatiotemporally aligned. The spatiotemporally aligned trajectory data is matched with the first map element in the ground truth map to determine the first map element that matches the trajectory data. The first map element that matches the trajectory data is marked in the trajectory data to obtain marked trajectory data.

6. The method according to claim 2, characterized in that, The step of determining the target trajectory that matches the first map element in the labeled trajectory data with the second map element in the vectorized map, and obtaining the trajectory subset of the second map element, includes: For each second map element in the vectorized map, a first map element that matches the second map element is determined based on the map element matching relationship; Based on the labeled trajectory data, determine at least one target trajectory associated with the first map element; At least one of the target trajectories is determined as a subset of the trajectories of the second map element.

7. The method according to claim 2, characterized in that, The method includes: For each second map element in the vectorized map, it is matched with a first map element in the ground truth map to determine the first map element that matches the second map element, thus obtaining the map element matching relationship; Based on the map element matching relationship and the preset static index, calculate the static index value between each second map element and the first map element that matches the second map element; The static index values ​​of each second map element are averaged to obtain the static detection results of the vectorized map.

8. The method according to claim 7, characterized in that, The step of calculating the static index value between each second map element and the first map element that matches the second map element, based on the map element matching relationship and preset static indices, includes... Determine the geometric features of the first map element and the second map element; wherein, the geometric features include polygons and point sets; For each pair of first map elements that match the second map element, calculate the spatial overlap between the polygons, as well as the nearest neighbor distance and average symmetric distance between the point sets. The spatial overlap, nearest neighbor distance, and average symmetry distance of each second map element are weighted and summed to obtain the static index value between the second map element and the first map element that matches the second map element.

9. An electronic device, characterized in that, Including processor and memory, among which Memory, used to store computer programs; A processor is used to execute a program stored in memory to implement the map quality detection method according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the map quality detection method according to any one of claims 1-8.

Images