Map element testing method and device and electronic equipment
By integrating point cloud and image data of the test scene, identifying and matching map elements, the problem of low accuracy in map availability evaluation in the prior art is solved, and more efficient map testing and evaluation is achieved.
Patent Information
- Application Number
- CN202411485268.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-23
- Publication Date
- 2025-03-07
AI Technical Summary
The prior art is difficult to comprehensively measure the availability of maps, and can only represent the deviation of maps in spatial location, and the accuracy is not high.
By obtaining point cloud and image data of the test scene, fusion and identifying target element data, matching real elements and map elements based on similarity, and evaluating map availability.
Accurate evaluation of map availability is achieved, the accuracy of map testing is improved, and the matching relationship between map elements can be determined more effectively.
Smart Images
Figure CN119579495A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present disclosure relate to the field of map technology, and in particular to a map element testing method, device, and electronic device. Background Art
[0002] Vehicles have become an indispensable part of people's daily lives, providing a convenient and efficient way of travel. Intelligent driving technology can help passengers control the vehicle, and intelligent driving technology can be combined with maps to achieve advanced driving assistance functions, so maps need to be tested.
[0003] At present, the accuracy of the map is tested by comparing the coordinate data collected on the spot with the coordinate data of the map to determine the specific accuracy of the map.
[0004] However, the test results of testing map accuracy can only indicate the deviation of the map in spatial position and cannot further measure the usability of the map. Therefore, the accuracy of this test scheme is not high. Summary of the invention
[0005] In view of the above problems, the embodiments of the present disclosure are proposed to provide a map element testing method, device and electronic device that overcome the above problems or at least partially solve the above problems.
[0006] In a first aspect, the present disclosure discloses a method for testing map elements, the method comprising:
[0007] Acquire test data corresponding to a test scene; the test scene corresponds to a map to be tested, and the test scene includes a plurality of target elements for road indication; the test data includes a point cloud and an image collected in the test scene;
[0008] Fusing the point cloud and the image, and identifying a first element data set in the fusion result; the first element data set includes first element data corresponding to each of the plurality of target elements;
[0009] Acquire a second element data set corresponding to the test scene from the map to be tested; the second element data set includes second element data corresponding to at least one target element among the multiple target elements;
[0010] Based on the similarity between each first element data and each second element data, matching is performed to obtain a matching result to complete the test; the matching result includes at least one set of first element data and second element data having a matching relationship.
[0011] In a second aspect, the present disclosure discloses a map element testing device, the device comprising:
[0012] A test scene module, used to obtain test data corresponding to a test scene; the test scene corresponds to a map to be tested, the test scene includes a plurality of target elements for road indication; the test data includes a point cloud and an image collected in the test scene;
[0013] A first element module, used for fusing the point cloud and the image, and identifying a first element data set in the fusion result; the first element data set includes first element data corresponding to each of the plurality of target elements;
[0014] A second element module is used to obtain a second element data set corresponding to the test scenario from the map to be tested; the second element data set includes second element data corresponding to at least one target element among the multiple target elements;
[0015] The element matching module is used to match each first element data with each second element data based on the similarity between them, and obtain a matching result to complete the test; the matching result includes at least one group of first element data and second element data having a matching relationship.
[0016] In a third aspect, an embodiment of the present disclosure further discloses an electronic device, including a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the program or instructions are executed by the processor, the steps of the map element testing method described in the first aspect are implemented.
[0017] In a fourth aspect, an embodiment of the present disclosure further discloses a computer-readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the map element testing method described in the first aspect are implemented.
[0018] In the disclosed embodiment, test data corresponding to the test scene can be obtained, the test data including a point cloud and an image collected in the test scene, the point cloud and the image are fused, a first element data set corresponding to multiple target elements of the test scene is identified in the fusion result, a second element data set corresponding to at least one target element among the multiple target elements of the test scene is obtained from a map, and matching is performed based on the similarity between each first element data and each second element data to obtain a matching result including at least one set of first element data and second element data having a matching relationship, and real first element data can be obtained based on the point cloud and the image collected in the test scene, and the first element data and the second element data of the map are matched for similarity, and the matching result can accurately represent the similarities and differences between the real elements and the map elements, and then the usability of the map can be determined, thereby ultimately improving the accuracy of map testing. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Figure 1 is a step diagram of a map element testing method provided by an embodiment of the present disclosure;
[0020] Figure 2 is a step diagram of another map element testing method provided by an embodiment of the present disclosure;
[0021] Figure 3 is a system schematic diagram of a map element test provided by an embodiment of the present disclosure;
[0022] Figure 4 is a block diagram of a map element testing device provided by an embodiment of the present disclosure;
[0023] Figure 5 It is a block diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION
[0024] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.
[0025] With the technological evolution of autonomous driving, the automatic assisted navigation driving (NOA, Navigate on Autopilot) function has become a focus. The NOA function can be based on different maps such as high-precision maps or advanced assisted driving maps (referred to as high-assisted maps). Among them, compared with high-precision maps, high-assisted maps have lower coverage and data freshness (indicating data timeliness and accuracy), so high-assisted maps can be a lightweight high-precision (HD lite, High Definition lite) map. Due to its low cost, high-assisted maps are gradually used to implement the NOA function, so high-assisted maps can be tested.
[0026] Figure 1 is a step diagram of a map element testing method provided by an embodiment of the present disclosure, the method comprising:
[0027] Step 101, obtaining test data corresponding to a test scene; the test scene corresponds to a map to be tested, and the test scene includes a plurality of target elements for road indication; the test data includes a point cloud and an image collected in the test scene.
[0028] In the disclosed embodiment, test data can be collected by equipping the test vehicle with a multi-line laser radar, a forward-looking monocular camera, a navigation system, a wheel speed encoder and other devices, and the efficiency and quality of data collection are improved by integrating sensors. By controlling the test vehicle to drive in the test scene, test data including point clouds and images can be obtained. The point cloud can be collected by the multi-line laser radar, and the image can be collected by the forward-looking monocular camera.
[0029] The map to be tested may include a larger area such as a city or province. The test scene may be a smaller area such as an area of a specific area, a street, etc. The test scene may also correspond to the entire map to be tested, which is not limited here.
[0030] In addition, the test scenarios can be based on the definition and description data of the map, divided according to road grades (expressways, urban expressways, urban ordinary roads), as well as static scenes such as service areas, double-sided mountains, tunnel entrances and exits, urban canyons, and under viaducts, to obtain different test scenarios.
[0031] The test scene includes multiple target elements for road signs, which can be one or more elements such as traffic signs, ground signs, lane lines, etc. Target elements such as traffic signs and ground signs can be classified and coded, and the target elements of each test scene can be summarized based on the coding of the target elements. Different test scenes can be selected according to test requirements and focus.
[0032] Step 102: fuse the point cloud and the image, and identify a first element data set in the fusion result; the first element data set includes first element data corresponding to each of the multiple target elements.
[0033] In the disclosed embodiment, since both the point cloud and the image are data collected based on the same test scene, only the data type and data source are different, the point cloud and the image can be fused. Since the test vehicle is equipped with a navigation system, different point clouds can have different location information, and different images can also have different location information, and the point cloud and the image can be fused based on the location information. In addition, the point cloud is a point cloud including objects in the test scene, and the image also includes the same objects in the test scene, so the point cloud and the image can be associated and fused based on the similarity of the objects collected by each, and the fusion method is not specifically limited here.
[0034] The fusion result based on the fusion of point cloud and image will not be overly dependent on a specific type of test data, and will be more stable. Therefore, it can provide more accurate and rich recognition data, and then the first element data set can be identified in the fusion result. The first element data set includes the first element data corresponding to multiple target elements in the test scene, that is, all target elements. For example, if there are 100 target elements in the test scene, the first element data set includes 100 first element data. Among them, each first element data is used to represent a target element, which can be data used to represent the shape of the target element, data on the size, data on the position, data on the element content (such as text and numbers on a sign, etc.), and so on.
[0035] Step 103: Acquire a second element data set corresponding to the test scenario from the map to be tested; the second element data set includes second element data corresponding to at least one target element among the multiple target elements.
[0036] In the embodiment of the present disclosure, the map data of the map to be measured includes pre-constructed element data, i.e., a second element data set. Since the data accuracy and data integrity of the map to be measured may be insufficient, the second element data set includes the second element data corresponding to at least one target element among the multiple target elements, and usually does not include the second element data of all target elements. The data type and content of the second element data are similar to those of the first element data, and only the data source is different, which will not be repeated here.
[0037] Step 104, performing matching based on the similarity between each first element data and each second element data, and obtaining a matching result to complete the test; the matching result includes at least one set of first element data and second element data having a matching relationship.
[0038] In the embodiment of the present disclosure, for each first element data, the similarity between it and each second element data can be calculated. For example, the first element data set includes data A and B, and the first element data set includes data C, D, and F. Then, the similarity between A and C, D, and F, and the similarity between B and C, D, and F need to be calculated respectively.
[0039] The number of similarities finally obtained can be multiple, and the process of matching based on multiple similarities can be to match the element data with the highest similarity. For example, for the first first element data in the first element data set, select the second element data with the highest similarity thereto, and continue to select the second element data with the highest similarity thereto for the second first element data, thereby completing the matching. It can also be other matching processes, which are not specifically limited here, and the matching results obtained include at least one set of first element data and second element data with a matching relationship.
[0040] In summary, in the embodiments of the present disclosure, test data corresponding to the test scene can be obtained, the test data including a point cloud and an image collected in the test scene, the point cloud and the image are fused, a first element data set corresponding to multiple target elements of the test scene is identified in the fusion result, a second element data set corresponding to at least one target element among the multiple target elements of the test scene is obtained from the map, and matching is performed based on the similarity between each first element data and each second element data to obtain a matching result including at least one set of first element data and second element data having a matching relationship, and real first element data can be obtained based on the point cloud and the image collected in the test scene, and the first element data and the second element data of the map are matched for similarity, and the matching result can accurately represent the similarities and differences between the real elements and the map elements, and then the usability of the map can be determined, thereby ultimately improving the accuracy of map testing.
[0041] Figure 2 is a step diagram of another map element testing method provided by an embodiment of the present disclosure, comprising:
[0042] Step 201, obtaining test data corresponding to a test scene; the test scene corresponds to a map to be tested, the test scene includes a plurality of target elements for road indication; the test data includes a point cloud and an image collected in the test scene;
[0043] Step 202, fusing the point cloud and the image, and identifying a first element data set in the fusion result; the first element data set includes first element data corresponding to each of the plurality of target elements;
[0044] Step 203, obtaining a second element data set corresponding to the test scenario from the map to be tested; the second element data set includes second element data corresponding to at least one target element among the multiple target elements;
[0045] Step 204, matching is performed based on the similarity between each first element data and each second element data to obtain a matching result to complete the test; the matching result includes at least one set of first element data and second element data having a matching relationship.
[0046] The above steps 201-204 can refer to the above Figure 1 The contents of the embodiments will not be described in detail here.
[0047] Optionally, the step 204 of performing matching based on the similarity between each first element data and each second element data to obtain a matching result to complete the test includes:
[0048] Sub-step 2041, constructing a cost matrix based on the similarity between each first element data and each second element data; the first row of the cost matrix is composed of the first element data set, and the first column of the cost matrix is composed of the second element data set; the similarity represents the matching cost between the element data;
[0049] Sub-step 2042, matching the first element data in the first element data set with the second element data in the second element data set based on the similarity in the cost matrix, and determining a total matching cost;
[0050] Sub-step 2043, when the sum of the matching costs meets the preset cost condition, obtain a matching result including multiple groups of element data; each group of element data includes first element data and second element data having a matching relationship.
[0051] In the embodiment of the present disclosure, the cost matrix means that if there are two sets A and B, where A has m elements and B has n elements, then an m×n matrix C can be constructed, where C[i,j] represents the cost of assigning the i-th element in A to the j-th element in B.
[0052] Sets A and B may be a first element data set and a second element data set, and the first row of the cost matrix may be formed based on the first element data in the first element data set, and the first column of the cost matrix may be formed based on the second element data in the second element data set. Alternatively, the first row of the cost matrix may be formed by the second element data set, and the first column of the cost matrix may be formed by the first element data set.
[0053] The cost matrix contains the similarity between each first element data and each second element data. For example, when the first element data set includes A1, A2 and A3, and the second element data set includes B1, B2 and B3, the cost matrix can be represented by the following Table 1:
[0054] A1 A2 A3 B1 a b c B2 d e f B3 g h i
[0055] In the above Table 1, a is the similarity between A1 and B1, b is the similarity between A2 and B1, and the similarity represents the matching cost between the first element data and the second element data. The other similarities are similar and will not be repeated here.
[0056] The process of allocating or combining the first element data and the second element data based on the similarity in the cost matrix can be performed using the Hungarian algorithm, the linear assignment algorithm or other algorithms. For example, considering the sparsity of the target to be matched and the semantic matching similarity, the Hungarian algorithm can be used. First, the initial zero matrix can be constructed based on the cost matrix, and the minimum value of the row is subtracted from each row, and the minimum value of the column is subtracted from each column to obtain an updated cost matrix. Then, all zeros are covered with the least number of lines, the uncovered minimum element is found, the matrix is adjusted, and the zero covering and adjustment of the matrix are repeated to find a match with a lower cost through an augmenting path. It can be understood that the specific implementation process of different algorithms is different, but the core of the matching process is to adjust the cost matrix so that multiple groups of matches can be found at a lower cost.
[0057] After matching, the corresponding matching cost can be determined based on the matched first element data and second element data, and the total matching cost can be determined based on multiple sets of matched first element data and second element data. It can be understood that matching can be performed several times, and since the matching costs between different first element data and different second element data are different, the total matching cost corresponding to each match can be different.
[0058] The preset cost condition may refer to the minimum sum of the matching costs corresponding to multiple matches, that is, if after several matches, the current matching cost sum is the minimum among the sums of several matching costs, it can be considered that the preset cost condition is met. The preset cost condition may also be a preset threshold value, without limitation.
[0059] Based on the sum of matching costs that meet the preset cost conditions, a matching result including multiple groups of element data corresponding to the sum of matching costs can be obtained, and each group of element data includes a first element data and a second element data.
[0060] In implementing the embodiments of the present disclosure, a cost matrix is constructed based on the similarity between each first element data and each second element data, and based on several similarities in the cost matrix, the first element data in the first element data set are combined with the second element data in the second element data set to obtain a matching result including multiple groups of element data. The cost matrix can be used to complete matching with lower costs, so that the first element data set and the second element data set have a higher degree of matching as a whole, thereby improving the accuracy of element data matching.
[0061] Optionally, the method further includes:
[0062] Step 205, based on at least one set of first element data and second element data having a matching relationship included in the matching result, determining target first element data that does not match any second element data;
[0063] Step 206: Determine the element omission rate of the map to be tested based on the first number of first element data in the first element data set and the second number of the target first element data.
[0064] In the disclosed embodiment, if the map data has problems with integrity and real-time performance, the second element data corresponding to the map data is less than the first element data, that is, the map data does not include some target elements actually existing in the test scene.
[0065] Since the matching results include at least one set of first element data and second element data having a matching relationship, the target first element data that is not matched with any second element data, that is, the first element data that has no corresponding matching second element data, can be determined based on the matching results. These first element data are missing from the map data.
[0066] Then, the element omission rate of the map to be tested can be determined based on the first number of the first element data in the first element data set, i.e., the total number, and the second number of the target first element data, i.e., the missing number. For example, when the first number is 100 and the second number is 15, the element omission rate of the map to be tested is 15%.
[0067] It is understandable that a corresponding test pass condition can be set for the element omission rate. For example, if the test scene is a 10-kilometer road and the element omission rate of the map to be tested is less than a preset threshold of 2%, the map to be tested passes the test.
[0068] By implementing the embodiments of the present disclosure, target first element data that are not matched with any second element data are determined based on the matching results, and the element omission rate of the map to be tested is determined based on the first number of first element data in the first element data set and the second number of target first element data. The availability of the map to be tested can be accurately determined. The element omission rate is inversely proportional to the availability. The higher the element omission rate, the lower the availability, thereby further improving the accuracy of map testing.
[0069] Optionally, the step of fusing the point cloud and the image and identifying a first element data set in the fusion result includes:
[0070] Inputting the point cloud and the image into a pre-trained multi-sensor fusion model to obtain a fusion result output by the multi-sensor fusion model;
[0071] In the fusion result, the first element data corresponding to each target element is determined to obtain a first element data set.
[0072] In the disclosed embodiment, the multi-sensor fusion model may be a convolutional neural network model BevFusion, or a Lift-Attend-Splat, DifFUSER or other model, which is not limited here.
[0073] The point cloud and image are input into the multi-sensor fusion model. The multi-sensor fusion model can fuse different types of data such as point cloud and image and output them. The output of the multi-sensor fusion model BevFusion can be a unified BEV expression.
[0074] The multi-sensor fusion model may have completed a certain scale of data training and obtained model parameters that can be used for identification and detection. Based on this, the multi-sensor fusion model identifies the fusion results, outputs the first element data corresponding to each target element, and then completes the road semantic feature extraction based on vision and laser, and finally obtains the first element data set. Among them, the first element data output by the multi-sensor fusion model recognition may include the type of target element (such as signs and logos, etc.), confidence (used to indicate the probability of belonging to the current type) and the bounding box of the target element. The bounding box may be composed of several contour points.
[0075] The first element data can be expressed by the following formula:
[0076] s t =(s1, s2, s3)
[0077] Among them, s t is the first element data, s1 is the element type, s2 is the confidence (0-100), and s3 is the bounding box.
[0078] In implementing the embodiments of the present disclosure, the point cloud and the image are input into the pre-trained multi-sensor fusion model to obtain the fusion result output by the multi-sensor fusion model, and in the fusion result, the first element data corresponding to each target element is determined to obtain the first element data set. Different types of data can be fused through the multi-sensor fusion model, and even when there are problems with the point cloud or image data, a relatively accurate fusion result can be obtained, thereby improving the reliability of the fusion result, and the fusion result can be used to identify and extract road semantic features, thereby improving the efficiency of obtaining the first element data.
[0079] Optionally, each first element data in the fusion result has a corresponding confidence level, and the confidence level indicates the accuracy of the first element data;
[0080] The method further comprises:
[0081] Determining a confidence reference value according to the confidence corresponding to each first element data;
[0082] The first element data with a confidence level lower than the confidence level reference value is deleted to obtain an updated first element data set.
[0083] In the embodiment of the present disclosure, since each first element data in the fusion result has a corresponding confidence level, the confidence level may represent the accuracy of the first element data, for example, the probability that the first element data belongs to a specific element type.
[0084] According to the confidence corresponding to each first element data, a confidence reference value is determined. The confidence reference value may be determined by averaging multiple confidences or performing other operations. The role of the confidence reference value is to screen the first element data, and delete the first element data with a confidence lower than the confidence reference value to obtain an updated first element data set. The subsequent calculation of the similarity between each first element data and each second element data may be based only on the first element data in the updated first element data set.
[0085] In implementing the embodiments of the present disclosure, a confidence reference value is determined based on the confidence corresponding to each first element data, and the first element data with a confidence lower than the confidence reference value is deleted to obtain an updated first element data set. The first element data with low accuracy can be deleted to update the first element data set, so that subsequent similarity calculations can be performed based on more accurate and fewer first element data, thereby improving the accuracy and efficiency of matching.
[0086] Optionally, the first element data is a first element frame composed of contour points, and the second element data is a second element frame composed of contour points;
[0087] The steps of matching each first element data with each second element data based on the similarity between them to obtain a matching result to complete the test include:
[0088] For each first element frame, determining an intersection-and-union ratio between the first element frame and each second element frame; the intersection-and-union ratio has a positive correlation with the similarity;
[0089] Matching is performed based on the intersection-and-union ratio between each first element frame and each second element frame, and a matching result is obtained to complete the test.
[0090] In the embodiment of the present disclosure, the first element data may be a plurality of contour points, and the contour points may constitute a first element frame corresponding to the first element data. The second element data and the second element frame are similar and will not be described in detail.
[0091] Therefore, when calculating the similarity between each first element data and each second element data, the intersection and union ratio of the first element frame and each second element frame can be determined for each first element frame. For example, for the first element frame and the second element frame, the intersection and union ratio can be calculated by the following formula:
[0092]
[0093] In the above formula, IOU represents intersection-over-union ratio, A is the first element box, B is the second element box, A∩B represents the intersection of A and B, and A∪B represents the union of A and B. There is a positive correlation between the intersection-over-union ratio and the similarity. The larger the intersection-over-union ratio, the higher the similarity.
[0094] Then, matching can be performed based on the intersection-and-union ratio between each first element frame and each second element frame to obtain a matching result to complete the test, and the specific matching process will not be described in detail.
[0095] In an embodiment of the present disclosure, for each first element frame of the first element data, the intersection-and-union ratio of the first element frame and the second element frame of each second element data is determined; there is a positive correlation between the intersection-and-union ratio and the similarity; matching is performed based on the intersection-and-union ratio between each first element frame and each second element frame, and a matching result is obtained to complete the test, and the similarity can be conveniently calculated based on the intersection-and-union ratio of the element frames, thereby improving the efficiency of element matching.
[0096] Optionally, the first element data is a group of first contour points, and the second element data is a group of second contour points;
[0097] The method further comprises:
[0098] According to each first contour point, a first contour set is constructed, and according to each second contour point, a second contour set is constructed;
[0099] constructing a spatial distance function based on the first contour set, the second contour set and a spatial position parameter; the spatial position parameter represents the relative position of the first contour set and the second contour set in space;
[0100] Iterative calculation is performed based on the spatial distance function to determine target spatial position parameters and target spatial distances corresponding to the spatial position parameters; the target spatial distances are used to represent the difference between the first contour set and the second contour set.
[0101] In the embodiment of the present disclosure, each first element data is a group of first contour points, and each second element data is a group of second contour points.
[0102] According to each first contour point q, the first contour set Q = {q1, q2, ..., q n}, and according to each second contour point p, the second contour set P = {p1, p2, ..., p n There is a spatial transformation relationship between the first contour set and the second contour set, that is, the first contour set can be close to the second contour set and overlap with the second contour set after a specific spatial transformation operation (such as rotation and translation). The spatial transformation operation can be implemented based on spatial position parameters (such as rotation parameter R and translation parameter t). The spatial position parameters limit the specific content of the spatial transformation operation. Therefore, the spatial position parameters can actually also represent the relative position of the first contour set and the second contour set in space.
[0103] Based on the first contour set, the second contour set, and the spatial position parameter, a spatial distance function is constructed. The spatial distance function may be a loss function, which may represent the distance between contour points in the Euclidean space. The spatial distance function may be represented by the following formula:
[0104]
[0105] Among them, E(R,t) represents the spatial distance, q i represents the i-th first contour point, p i The i-th second contour point, n represents the number of contour points in the first contour set or the second contour set, R is the rotation parameter, and t is the translation parameter.
[0106] Based on the spatial distance function, the Gauss-Newton algorithm can be used to perform iterative calculations to determine the target spatial position parameters and the target spatial distance corresponding to the spatial position parameters. The target spatial distance is the minimum spatial distance. The target spatial position parameters are used to limit the specific spatial transformation operations required to achieve the target spatial distance, such as rotation and translation.
[0107] The target space distance may represent the minimum error or difference between the first contour set and the second contour set. The size of the target space distance represents the accuracy of the map to be measured. The smaller the target space distance, the higher the accuracy of the map to be measured.
[0108] It is understandable that corresponding test pass conditions can be set for the error represented by the target space distance. For example, if the test scene is a 10-kilometer road and the calculated target space distance is less than 30 centimeters, the map to be tested passes the test.
[0109] In addition, if the test pass condition, such as the element missing rate, is not met, it can be considered that the map engine is faulty, and the map engine can be retested after optimization and adjustment. The map engine can be an Electronic Horizon Provider (EHP), which can obtain specific road and attribute information from the map data and output it.
[0110] Similarly, when the test pass condition is not met, such as the error shown by the target space distance, it can be considered that there is a problem with the map accuracy, and the map software development kit (SDK) can be tested for data.
[0111] A test report can also be generated based on whether the test pass conditions are met, as well as the specific values of element omission rate and error. The test report can provide testers with information on test pass status, element omission rate and error, as well as suggestions on map engine testing and data testing.
[0112] In implementing the embodiments of the present disclosure, a first contour set is constructed according to each first contour point, and a second contour set is constructed according to each second contour point. A spatial distance function is constructed based on the first contour set, the second contour set, and the spatial position parameters. An iterative calculation is performed based on the spatial distance function to determine the target spatial position parameters and the target spatial distance corresponding to the spatial position parameters. The target spatial distance can be calculated through the spatial distance function, and the error size of the map to be tested can be accurately determined, thereby improving the accuracy of map testing.
[0113] It is understandable that in addition to the error between the first contour point and the second contour point, the error between the first contour point and the target element in the actual test scene can also be calculated. For example, for the same sign, the difference between the spatial distance between the point clouds and the spatial distance between the control points on the sign can represent the error of the test data obtained in the test scene. It is calculated by the following formula:
[0114]
[0115] Among them, σ is the error, Δ represents the distance difference between the spatial distance between the point clouds and the spatial distance between the control points, and n represents the number of samples of the selected point clouds or the number of samples of the control points.
[0116] Optionally, the step of acquiring a second element data set corresponding to the test scene from the map to be tested includes:
[0117] Determine the target coordinate range corresponding to the test scene;
[0118] In the element data of the map to be measured, a second element data set of the target elements corresponding to the target coordinate range is obtained.
[0119] In the disclosed embodiment, the target coordinate range corresponding to the current test scene can be determined, and then based on the target coordinate range, a second element data set of target elements corresponding to the target coordinate range can be directly obtained from the element data of the map to be tested.
[0120] It is understandable that the target element with road indication function has corresponding element data in the map data. For example, the coordinate position of the sign is described in the map data. Then, the second element data within the target coordinate range can be determined according to the coordinates of the element data.
[0121] In addition, the second element data set corresponding to the test scene in the map to be tested can also be obtained in real time. For example, when the test vehicle is traveling in the test scene, the map engine selects road elements 2 kilometers in front and 500 meters behind the test vehicle in real time based on the high-precision positioning information released by the combined navigation of the test vehicle. It can also be further screened, for example, by screening target elements such as signs and logos. In this way, the element data of target elements such as traffic signs and ground signs can be broadcast, so that the relevant element data can be obtained in the map.
[0122] By implementing the embodiments of the present disclosure, by determining the target coordinate range corresponding to the test scene, a second element data set of the target elements corresponding to the target coordinate range is obtained in the element data of the map to be tested. The second element data can be efficiently obtained based on the target coordinate range, thereby improving the efficiency of element matching.
[0123] Figure 3 is a system schematic diagram of a map element test provided by an embodiment of the present disclosure; Figure 3 It includes a data injection system 301, a scene library center 302 and a central processor 303.
[0124] The data re-injection system 301 is used to realize the extraction of semantic features of roads based on vision and laser, the expression of features based on contour points, the output of contour points extracted based on point cloud, and the element type extracted based on image. The scene library center 302 is used to realize the selection of test scenes and write the error calculation method to the central processor 303. The central processor 303 is used to obtain the map elements (second element data) in the map, the expression of map element features, and the contour points of map elements, to realize element matching, determine the target space distance, and realize the statistics of element underreporting rate, error calculation, and test report output.
[0125] Overall, automated testing through the test method provided in this application can reduce the need for manual testing, speed up the testing process, and improve testing efficiency. The sensors of the test vehicle collect data in real time and inject it back for testing, ensuring the timeliness and accuracy of the test data. Through automated testing methods, errors in map data can be discovered and corrected in a timely manner, improving the quality of map data. Test reports for map users can be automatically output, improving the efficiency and accuracy of test report generation. Efficient testing methods help to accelerate the mass production of high-assisted maps and promote the development of autonomous driving technology. Through automated testing, the functionality and accuracy of the map engine can be evaluated more reliably to ensure the stability of autonomous driving.
[0126] In summary, in the embodiments of the present disclosure, test data corresponding to the test scene can be obtained, the test data including a point cloud and an image collected in the test scene, the point cloud and the image are fused, a first element data set corresponding to multiple target elements of the test scene is identified in the fusion result, a second element data set corresponding to at least one target element among the multiple target elements of the test scene is obtained from the map, and matching is performed based on the similarity between each first element data and each second element data to obtain a matching result including at least one set of first element data and second element data having a matching relationship, and real first element data can be obtained based on the point cloud and the image collected in the test scene, and the first element data and the second element data of the map are matched for similarity, and the matching result can accurately represent the similarities and differences between the real elements and the map elements, and then the usability of the map can be determined, thereby ultimately improving the accuracy of map testing.
[0127] Figure 4 A map element testing device 40 provided in an embodiment of the present disclosure includes:
[0128] The test scene module 401 is used to obtain test data corresponding to the test scene; the test scene corresponds to the map to be tested, and the test scene includes multiple target elements for road indication; the test data includes point clouds and images collected in the test scene;
[0129] A first element module 402 is used to fuse the point cloud and the image, and identify a first element data set in the fusion result; the first element data set includes first element data corresponding to each of the plurality of target elements;
[0130] The second element module 403 is used to obtain a second element data set corresponding to the test scenario from the map to be tested; the second element data set includes second element data corresponding to at least one target element among the multiple target elements;
[0131] The element matching module 404 is used to match each first element data with each second element data based on the similarity between them, and obtain a matching result to complete the test; the matching result includes at least one set of first element data and second element data having a matching relationship.
[0132] Optional, element matching modules, including:
[0133] A cost matrix submodule, for constructing a cost matrix based on the similarity between each first element data and each second element data; the first row of the cost matrix is composed of the first element data set, and the first column of the cost matrix is composed of the second element data set;
[0134] a matching cost submodule, configured to match the first element data in the first element data set with the second element data in the second element data set based on the similarity in the cost matrix, and determine a sum of matching costs;
[0135] The matching result submodule is used to obtain matching results including multiple groups of element data when the total matching cost meets the preset cost condition; each group of element data includes first element data and second element data having a matching relationship.
[0136] Optionally, the device further comprises:
[0137] A missing element module, configured to determine target first element data that is not matched with any second element data based on at least one set of first element data and second element data having a matching relationship included in the matching result;
[0138] The ratio determination module is used to determine the element omission rate of the map to be tested based on the first quantity of the first element data in the first element data set and the second quantity of the target first element data.
[0139] Optional, first element module, including:
[0140] A fusion model submodule, used for inputting the point cloud and the image into a pre-trained multi-sensor fusion model to obtain a fusion result output by the multi-sensor fusion model;
[0141] The model identification submodule is used to determine the first element data corresponding to each target element in the fusion result to obtain a first element data set.
[0142] Optionally, each first element data in the fusion result has a corresponding confidence level, and the confidence level indicates the accuracy of the first element data;
[0143] The device also includes:
[0144] A reference confidence module, used to determine a confidence reference value according to the confidence corresponding to each first element data;
[0145] The set updating module is used to delete the first element data whose confidence is lower than the confidence reference value to obtain an updated first element data set.
[0146] Optionally, the first element data is a first element frame composed of contour points, and the second element data is a second element frame composed of contour points;
[0147] Element matching modules, including:
[0148] An IoU determination submodule, used to determine, for each first element frame, an IoU ratio between the first element frame and each second element frame; the IoU ratio has a positive correlation with the similarity;
[0149] The intersection-and-union ratio matching submodule is used to perform matching based on the intersection-and-union ratio between each first element frame and each second element frame to obtain a matching result to complete the test.
[0150] Optionally, the first element data is a group of first contour points, and the second element data is a group of second contour points; the device further includes:
[0151] A contour set module, used to construct a first contour set according to each first contour point, and to construct a second contour set according to each second contour point;
[0152] A function construction module, configured to construct a spatial distance function based on the first contour set, the second contour set and a spatial position parameter; the spatial position parameter represents a relative position of the first contour set and the second contour set in space;
[0153] A target distance module is used to perform iterative calculation based on the spatial distance function to determine a target spatial position parameter and a target spatial distance corresponding to the spatial position parameter; the target spatial distance is used to represent the difference between the first contour set and the second contour set.
[0154] Optional, second element module, including:
[0155] A coordinate range module, used to determine the target coordinate range corresponding to the test scene;
[0156] The range element module is used to obtain a second element data set of target elements corresponding to the target coordinate range from the element data of the map to be measured.
[0157] In summary, in the embodiments of the present disclosure, test data corresponding to the test scene can be obtained, the test data including a point cloud and an image collected in the test scene, the point cloud and the image are fused, a first element data set corresponding to multiple target elements of the test scene is identified in the fusion result, a second element data set corresponding to at least one target element among the multiple target elements of the test scene is obtained from the map, and matching is performed based on the similarity between each first element data and each second element data to obtain a matching result including at least one set of first element data and second element data having a matching relationship, and real first element data can be obtained based on the point cloud and the image collected in the test scene, and the first element data and the second element data of the map are matched for similarity, and the matching result can accurately represent the similarities and differences between the real elements and the map elements, and then the usability of the map can be determined, thereby ultimately improving the accuracy of map testing.
[0158] The present application also provides an electronic device, such as Figure 5 As shown, it includes a processor 1001, a communication interface 1002, a memory 1003 and a communication bus 1004, wherein the processor 1001, the communication interface 1002 and the memory 1003 communicate with each other via the communication bus 1004. The memory 1003 is used to store computer programs.
[0159] When the processor 1001 is used to execute the program stored in the memory 1003, the following steps are implemented: obtaining test data corresponding to a test scene; the test scene corresponds to the map to be tested, and the test scene includes multiple target elements for road indication; the test data includes a point cloud and an image collected in the test scene; the point cloud and the image are fused, and a first element data set is identified in the fusion result; the first element data set includes first element data corresponding to each of the multiple target elements; obtaining a second element data set corresponding to the test scene from the map to be tested; the second element data set includes second element data corresponding to each of at least one of the multiple target elements; matching based on the similarity between each first element data and each second element data, obtaining a matching result to complete the test; the matching result includes at least one set of first element data and second element data having a matching relationship. The processor 1001 can also implement other steps in the above-mentioned map element testing method, which will not be repeated here.
[0160] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used for communication between the above electronic device and other devices.
[0161] The memory may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.
[0162] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
[0163] In another embodiment provided by the present application, a computer-readable storage medium is also provided. The computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the map element testing method described in the above embodiment.
[0164] In another embodiment provided by the present application, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute the above-mentioned map element testing method.
[0165] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.
[0166] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.
[0167] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. For the embodiments of the device, electronic device, computer-readable storage medium and computer program products containing instructions, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.
[0168] The above description is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application are included in the protection scope of the present application.
Claims
1. A map element testing method, characterized in that: The method comprises: Acquire test data corresponding to a test scene; the test scene corresponds to a map to be tested, and the test scene includes a plurality of target elements for road indication; the test data includes a point cloud and an image collected in the test scene; Fusing the point cloud and the image, and identifying a first element data set in the fusion result; the first element data set includes first element data corresponding to each of the plurality of target elements; Acquire a second element data set corresponding to the test scene from the map to be tested; the second element data set includes second element data corresponding to at least one target element among the multiple target elements; Based on the similarity between each first element data and each second element data, matching is performed to obtain a matching result to complete the test; the matching result includes at least one set of first element data and second element data having a matching relationship.
2. The method according to claim 1, characterized in that The step of matching based on the similarity between each first element data and each second element data to obtain a matching result to complete the test includes: Based on the similarity between each first element data and each second element data, a cost matrix is constructed; the first row of the cost matrix is composed of the first element data set, and the first column of the cost matrix is composed of the second element data set; the similarity represents the matching cost between the element data; Based on the similarity in the cost matrix, matching the first element data in the first element data set with the second element data in the second element data set, and determining a sum of matching costs; In the case where the sum of the matching costs meets the preset cost condition, a matching result including multiple groups of element data is obtained; each group of element data includes first element data and second element data having a matching relationship.
3. The method according to claim 1, characterized in that The method further comprises: Based on at least one set of first element data and second element data having a matching relationship included in the matching result, determining target first element data that does not match any second element data; Based on the first quantity of the first element data in the first element data set and the second quantity of the target first element data, the element omission rate of the map to be measured is determined.
4. The method according to claim 1, characterized in that The step of fusing the point cloud and the image and identifying a first element data set in the fusion result includes: Inputting the point cloud and the image into a pre-trained multi-sensor fusion model to obtain a fusion result output by the multi-sensor fusion model; In the fusion result, the first element data corresponding to each target element is determined to obtain a first element data set.
5. The method according to claim 1, characterized in that Each first element data in the fusion result has a corresponding confidence level, and the confidence level indicates the accuracy of the first element data; The method further comprises: Determining a confidence reference value according to the confidence corresponding to each first element data; The first element data with a confidence level lower than the confidence level reference value is deleted to obtain an updated first element data set.
6. The method according to claim 1, characterized in that The first element data is a first element frame composed of contour points, and the second element data is a second element frame composed of contour points; The step of matching based on the similarity between each first element data and each second element data to obtain a matching result to complete the test includes: For each first element frame, determining an intersection-and-union ratio between the first element frame and each second element frame; the intersection-and-union ratio has a positive correlation with the similarity; Matching is performed based on the intersection-and-union ratio between each first element frame and each second element frame, and a matching result is obtained to complete the test.
7. The method according to claim 1, characterized in that The first element data is a group of first contour points, and the second element data is a group of second contour points; the method further includes: According to each first contour point, a first contour set is constructed, and according to each second contour point, a second contour set is constructed; constructing a spatial distance function based on the first contour set, the second contour set and a spatial position parameter; the spatial position parameter represents the relative position of the first contour set and the second contour set in space; Iterative calculation is performed based on the spatial distance function to determine target spatial position parameters and target spatial distances corresponding to the spatial position parameters; the target spatial distances are used to represent the difference between the first contour set and the second contour set.
8. The method according to claim 1, characterized in that: The step of acquiring a second element data set corresponding to the test scene from the map to be tested includes: Determine the target coordinate range corresponding to the test scene; In the element data of the map to be measured, a second element data set of the target elements corresponding to the target coordinate range is obtained.
9. A map element testing device, characterized in that: The device comprises: A test scene module, used to obtain test data corresponding to a test scene; the test scene corresponds to a map to be tested, the test scene includes a plurality of target elements for road indication; the test data includes a point cloud and an image collected in the test scene; A first element module, used for fusing the point cloud and the image, and identifying a first element data set in the fusion result; the first element data set includes first element data corresponding to each of the plurality of target elements; A second element module is used to obtain a second element data set corresponding to the test scenario from the map to be tested; the second element data set includes second element data corresponding to at least one target element among the multiple target elements; The element matching module is used to match each first element data with each second element data based on the similarity between them, and obtain a matching result to complete the test; the matching result includes at least one group of first element data and second element data having a matching relationship.
10. An electronic device, characterized in that: include: A processor, a communication interface, a memory and a communication bus; wherein the processor, the communication interface and the memory communicate with each other via the communication bus; Memory, used to store computer programs; The processor is used to implement the steps of the map element testing method as described in any one of claims 1 to 8 when executing the program stored in the memory.