Map data detection method and device and storage medium

Abstract

The invention discloses a map data detection method and device and a storage medium, and the method comprises the steps: obtaining to-be-detected map vector data, the map vector data comprising first map data and second map data corresponding to the same target area; preprocessing the map vector data according to the detection target to obtain a target aerial view; and inputting the target aerial view into an image detection model corresponding to the detection target to obtain a target detection result, the target detection result being used for indicating a problem that the first map data and the second map data correspond to the detection target. According to the embodiment of the invention, the map vector data is converted into the target aerial view, and then the target aerial view is subjected to target detection or classification prediction by the image detection model to obtain the problem point location, so that the accuracy of the map data detection result is effectively improved.

Description

Technical Field

[0001] This disclosure generally relates to the field of map detection technology. More specifically, this disclosure relates to a method, apparatus, and storage medium for detecting map data. Background Art

[0002] In scenarios such as autonomous driving and lane-level navigation, it is often necessary to implement functions such as navigation planning based on the association relationship between HD segments in a High Definition (HD) map and SD segments in a Standard Definition (SD) map. If there is a problem with either type of data, HD data or SD data, on a route, then when the two are used in combination, due to data inconsistency, lane-level navigation will degrade to road-level navigation, which will affect the user experience.

[0003] In view of this, there is an urgent need to provide a method for detecting map data to detect possible data inconsistency problems between SD data and HD data corresponding to the same area, so as to meet the usage requirements of lane-level navigation data. Summary of the Invention

[0004] To solve at least one or more of the above-mentioned technical problems, this disclosure proposes a map data detection solution in multiple aspects.

[0005] In a first aspect, an embodiment of this disclosure proposes a method for detecting map data, the method including: obtaining map vector data to be detected, the map vector data including first map data and second map data corresponding to the same target area; preprocessing the map vector data according to a detection target to obtain a target overview map; inputting the target overview map into an image detection model corresponding to the detection target to obtain a target detection result, the target detection result being used to indicate problems of the first map data and the second map data corresponding to the detection target.

[0006] In a second aspect, an embodiment of this disclosure proposes a map data detection apparatus, the apparatus including: a data acquisition module configured to obtain map vector data to be detected, the map vector data including first map data and second map data corresponding to the same target area; a data preprocessing module configured to preprocess the map vector data according to a detection target to obtain a target overview map; a data problem detection module configured to input the target overview map into an image detection model corresponding to the detection target to obtain a target detection result, the target detection result being used to indicate problems of the first map data and the second map data corresponding to the detection target.

[0007] In a third aspect, the present disclosure provides a computer-readable storage medium including computer program instructions for performing a map data detection method by a computer. When the computer program instructions are executed by a processor, the method described in the first aspect is implemented.

[0008] Through the map data detection method, device, and storage medium provided as above, the map data detection method proposed in the embodiments of the present disclosure first obtains map vector data to be detected, where the map vector data includes first map data and second map data corresponding to the same target area. Then, according to the detection target, the map vector data is preprocessed to obtain a target bird's-eye view. Next, the target bird's-eye view is input into an image detection model corresponding to the detection target to obtain a target detection result, which is used to indicate problems of the first map data and the second map data corresponding to the detection target. Compared with the related art, in the embodiments of the present disclosure, the map vector data is preprocessed to convert the map vector data into a target bird's-eye view, and then the image detection model is used to perform target detection or classification prediction on the target bird's-eye view, which can automatically detect problems such as unconnected roads and road dead ends in the map vector data, effectively improving the accuracy of the map data detection result and the efficiency of automatic detection.

[0009] Further, inverse coordinate transformation processing is performed on the target detection result according to the coordinate transformation file to output problem points at the road level corresponding to the target detection result, which can be directly used for business production, further improving the efficiency of automatic detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features, and advantages of the exemplary embodiments of the present disclosure will become readily understood. In the drawings, several embodiments of the present disclosure are shown in an exemplary rather than restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts, where:

[0011] Figure 1 A schematic diagram of an application scenario of the map data detection method according to the embodiments of the present disclosure is shown;

[0012] Figure 2 An exemplary flowchart of a map data detection method 200 according to some embodiments of the present disclosure is shown;

[0013] Figure 3 An exemplary flowchart of a map data detection method 300 according to some other embodiments of the present disclosure is shown;

[0014] Figure 4 A schematic diagram of map sheet segmentation processing according to some embodiments of the present disclosure is shown;

[0015] Figure 5A schematic diagram of the map sheet splitting process showing another embodiment of the present disclosure;

[0016] Figure 6 An exemplary flowchart of the map data detection method 600 showing still some other embodiments of the present disclosure;

[0017] Figure 7 An exemplary flowchart of the map data detection method 700 showing yet some other embodiments of the present disclosure;

[0018] Figure 8 A schematic diagram of the position area of the target detection frame for road dead end detection showing some embodiments of the present disclosure;

[0019] Figure 9 A schematic diagram of the classification prediction result for road disconnection detection showing some embodiments of the present disclosure;

[0020] Figure 10 A schematic diagram of the classification prediction result corresponding to road disconnection showing yet some other embodiments of the present disclosure;

[0021] Figure 11 A schematic block diagram of the structure of the map data detection device 1100 showing some embodiments of the present disclosure;

[0022] Figure 12 A schematic block diagram of the electronic device 1200 showing an embodiment of the present disclosure. Detailed implementation manners

[0023] Next, the technical solutions in the embodiments of the present disclosure will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present disclosure. Obviously, the described embodiments are some embodiments of the present disclosure, rather than all embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present disclosure.

[0024] It should be understood that the terms "including" and "comprising" used in the specification and claims of the present disclosure indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0025] It should also be understood that the terms used in this disclosure specification are for the purpose of describing specific embodiments only and are not intended to limit the disclosure. As used in this disclosure specification and the claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms. It should be further understood that the term "and / or" used in this disclosure specification and the claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0026] The following will describe in detail the specific embodiments of the disclosure with reference to the accompanying drawings.

[0027] Figure 1 An exemplary scenario to which the embodiments of the disclosure can be applied is shown.

[0028] Please refer to Figure 1 , Figure 1 which shows a schematic diagram of the implementation environment of the map data detection method according to the embodiments of the disclosure. As Figure 1 shown, the implementation environment may include one or more terminal devices 101 and one or more servers 102, and the number of terminal devices and servers will not be limited here. The terminal device 101 can be network-connected to the server 102 so that the terminal device 101 can perform data interaction with the server 102 through the network connection.

[0029] As Figure 1 shown, the terminal device 101 can be: a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart watch, a vehicle terminal, a car computer, etc. The server 102 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network, content delivery network), and big data and artificial intelligence platforms.

[0030] An electronic map application is installed on the terminal device 101. The server 102 provides data services for the electronic map application of the terminal device 101. Before providing the map data to the terminal device, it is necessary to detect relevant issues of the map data. In scenarios such as autonomous driving and lane-level navigation, it is necessary to implement functions such as navigation planning based on the association relationship between HD segments and SD segments. If there is a problem of mismatch between HD data and SD data on a route, lane-level navigation will degrade to road-level navigation, thus affecting the user experience.

[0031] In the related art, a method for detecting vector algorithms for the data layer can detect mismatches between SD data and HD data. For example, by analyzing the road connection relationships and topological structures in SD data and HD data, topological differences between the two can be found, such as incorrect road connections and road disconnections. However, in the face of complex road situations, it is difficult to effectively solve the difference problem between SD map data and HD map data using a single parameter setting or processing strategy. In addition, even in the best case, this method cannot achieve a very high accuracy level.

[0032] In the related art, a method of using a large model to analyze map data requires a large amount of computing resources for training and inference. When the amount of vector map data is huge, it means higher hardware investment, which obviously increases the cost. Moreover, when using a large model to process high-precision map data, taking the entire high-precision map data as input, it is difficult to accurately identify specific problem areas or features. This method cannot locate problems to the road level, and the map data detected by this method cannot be directly put into production for use.

[0033] Therefore, the present disclosure proposes a map data detection method to solve the above problems, realizing fully automated detection of map data and directly outputting problem points at the road level, effectively improving the detection efficiency of map data problems.

[0034] Technical term explanations

[0035] The path restoration rate refers to the consistency of path expressions between the standard-precision map data and the high-precision map data in the lane-level navigation scenario. The path restoration rate is a matching index used to evaluate the standard-precision map data and the high-precision map data. The high-precision map data here includes, but is not limited to, the following high-precision map data and high-quality map data.

[0036] High-precision (HD) map data refers to a map with high precision and refined definition, and its precision needs to reach the decimeter level to distinguish each lane.

[0037] High-quality (HQ) map data is a map with a precision between HD map data and SD map data.

[0038] Standard-precision (SD) map data refers to map data with a navigation precision at the meter level.

[0039] Connectivity refers to whether it is passable between multiple roads associated in a route. For example, the road network data represented by SD map data shows a passable state. However, the road surfaces in the HD map data topologically associated with it show an impassable state, that is, the road surfaces in the HD map data topologically associated with the road network of the SD map data are not connected, which is the problem of road disconnection. In the HD map data, the unconnected road surfaces cannot be used for lane-level navigation.

[0040] The problem of a dead end means that there is a road network in the SD map data, but a certain road in the HD map data is not fully covered. A dead end represents a problem of road completeness and will directly affect the exit of lane-level navigation.

[0041] Map taking refers to converting map vector data into pixel picture data according to a preset scale and synchronously generating a coordinate transformation file.

[0042] Object detection is to detect objects in an image based on computer vision technology, find the target regions of interest, and output the position boxes of the target regions in the image.

[0043] Classification prediction is to predict the category of an image based on computer vision technology.

[0044] The following combines the attached Figure 2-11 Elaborate in detail the map data detection method of the embodiments of the present disclosure.

[0045] Figure 2 An exemplary flowchart of the map data detection method 200 of some embodiments of the present disclosure is shown.

[0046] As Figure 2 shown, in step S201, obtain the map vector data to be detected, and the map vector data includes first map data and second map data corresponding to the same target region.

[0047] In the above step, the accuracy of the first map data is lower than that of the second map data. The first map data may be SD map data. The second map data may be HQ map data or HD map data. In the SD map data, a road is usually represented as a line. The first map data includes but is not limited to SD map data. In the first map data, a road can be divided into multiple segments, and each segment is an SD road or link. The second map data includes but is not limited to HD map data and HQ map data. In the second map data, a road is usually represented as a planar shape (i.e., a polygon), which can be called a road surface, and a road can be divided into multiple segments, and each segment is represented as a road surface.

[0048] In step S202, preprocess the map vector data according to the detection target to obtain a target bird's-eye view.

[0049] In the above steps, the target overhead view is a raster image obtained by coordinate transformation of the local data of the first map data and the second map data for the same target area. For example, for the same target road, the local data of the first map data refers to the linear data corresponding to the target road in the first map data. The local data of the second map data refers to the polygon data corresponding to the target road in the second map data.

[0050] Map vector data is a data format that represents geographical information using geometric shapes (such as points, lines, and polygons). Each shape has corresponding coordinates and attribute information, which can be used to describe the spatial location and characteristics of geographical objects. There are some problems in directly detecting road disconnection in map vector data. For example, map vector data is usually represented by geometric objects such as points, lines, and polygons. Directly detecting road disconnection requires dealing with a large number of geometric calculations, such as judging the intersection of line segments, which is computationally intensive and complex. Map vector data also contains topological relationships, and directly processing requires maintaining a complex topological structure, increasing the difficulty of algorithm implementation.

[0051] By performing coordinate transformation on map vector data, the map vector data can be represented using pixel coordinates, which can simplify calculations, improve efficiency, and reduce complexity. For example, converting the geographical coordinates of map vector data into pixel coordinates, where pixel coordinates are discrete integers. When calculating connectivity, only need to judge whether the pixels are adjacent, avoiding complex geometric operations. After obtaining the target overhead view, the pixel image can be detected for matching problems through an image detection model, which can improve the efficiency of problem detection. In addition, the accuracy of the pixel coordinates after coordinate transformation is relatively low, reducing the influence of floating-point errors and making the detection results of matching problems more stable.

[0052] In the above steps, the detection target refers to identifying the problem that the second map data does not exactly match the first map data in terms of topological connectivity and adjacency. The detection targets include road dead-end detection or road disconnection detection. For example, in the first map data, road A is represented by a line shape, and if the polygon (road surface) associated with road A in the second map data is not completely covered, it is called a road dead-end. Another example is that in the first map data, it is possible to connect from road A to road B, indicating that road A and road B are passable. If there is no connectable road surface between the road surface associated with road A and the road surface associated with road B in the second map data, it means that the road surface associated with road A and the road surface associated with road B are not connected, which is called road disconnection. In map vector data, topological connectivity refers to the connection relationship between geographical space objects. For example, intersections in a road network, etc. In map vector data, these connection relationships are usually represented by nodes (point features) and edges (line features). Adjacency refers to the adjacency relationship between geographical space objects, that is, the boundary relationship between adjacent regions. In map vector data, these adjacency relationships are usually represented by the shared edges of polygon features.

[0053] Suppose the map vector data is map data within a range of 10Km * 10Km in size. When the detection target is road dead-end detection, the map vector data can be sliced into multiple sub-map sheets. For each sub-map sheet, a transformation relationship between local rectangular coordinates and pixel coordinates is established according to a pre-set pixel scale. The vector slice data corresponding to each sub-map sheet is taken as a target aerial view, and a coordinate conversion file corresponding to the target aerial view is generated according to the transformation relationship between local rectangular coordinates and pixel coordinates. The pixel scale refers to the distance length represented by each pixel for the map vector data. For example, 1 pixel represents 0.1m. Suppose the map sheet is 10Km * 10Km and is sliced into 50 * 50 slice ranges according to 200m * 200m. If 1 pixel represents 0.1m, then the size of 1 vector slice data can be approximately 2000 * 2000 in terms of image size.

[0054] When the detection target is the detection of disconnected roads, the center position of the map sheet can be used as the reference origin to establish a local rectangular coordinate system, and the map vector data can be converted from the global coordinate system to the local rectangular coordinate system. In the first map data, one or more connected candidate pairs are obtained based on the road network nodes and the road network information used to describe complex traffic intersections. According to the topological connection relationship of the connected candidate pairs, the road surface data corresponding to the connected candidate pairs is obtained from the second map data, and then the connected roads are searched according to the road surface data to obtain one or more candidate connected roads corresponding to each connected candidate pair. Then, a target bird's-eye view is taken based on the vector slice data corresponding to the candidate connected roads, and a coordinate conversion file corresponding to the target bird's-eye view is generated based on the transformation relationship between the local rectangular coordinates and the pixel coordinates.

[0055] In step S203, the target bird's-eye view image is input into an image detection model corresponding to the detection target to obtain a target detection result, which is used to indicate a problem between the first map data and the second map data corresponding to the detection target.

[0056] In the above steps, when the detection target is road break detection, the target detection result is a target detection frame corresponding to the road break scene. When the detection target is road disconnection detection scene, the target detection result is a classification prediction result corresponding to road disconnection detection.

[0057] In the above steps, when the detection target is road break detection, the image detection model corresponding to the road break detection is a target detection model based on a convolutional neural network architecture. When the detection target is road disconnection detection, the image detection model corresponding to the road disconnection detection is a classification prediction model based on a convolutional neural network architecture. The image detection model can be built based on the Yolov8 network architecture. Compared with the large model detection solution, the use of a model based on a convolutional neural network architecture can significantly improve the accuracy of the detection results and prediction results.

[0058] The generated bird's-eye view of the target is used as an annotated training sample to train the image detection model, and an image detection model corresponding to the detection target is obtained. After the bird's-eye view of the target is processed using the image detection model corresponding to the detection target, the bird's-eye view of the target corresponding to the target detection result is used as a new training sample and returned to the process of training the image detection model corresponding to the detection target, so that the image detection model corresponding to the detection target can be iteratively upgraded.

[0059] In some embodiments, under road dead-end detection, object detection is performed on the target aerial view to detect the dead-end problem in each target aerial view, and the pixel positions corresponding to the object detection boxes in the target aerial view are output. A mechanism is established. For the feedback point data, the target aerial view is fed back to the training dataset as a new increment to train the image detection model for iterative upgrade, thereby improving the generalization ability of the model.

[0060] In some embodiments, under road disconnection detection, classification prediction is performed on the target aerial view, that is, the road connectivity in each target aerial view is classified and predicted, and the corresponding classification prediction results of the target aerial view are output. A mechanism is established. For the feedback point data, multiple target aerial views that have undergone classification prediction are traced back, and the multiple target aerial views are fed back to the training dataset as a new increment to train the image detection model for iterative upgrade, thereby improving the generalization ability of the model.

[0061] Through the above processing, the generalization ability of the image detection model corresponding to the detection target can be effectively improved.

[0062] Compared with the related art, the map data detection method proposed in the embodiments of the present disclosure preprocesses the map vector data to convert the map vector data into a target aerial view, and then uses the image detection model to perform object detection or classification prediction on the target aerial view, which can automatically detect the roads with disconnection problems and dead-end problems in the map vector data, effectively improving the accuracy and detection efficiency of the road detection results.

[0063] Figure 3 An exemplary flowchart of a map data detection method 300 according to other embodiments of the present disclosure is shown. Figure 3 It is shown that steps S301 - S302 are a specific implementation of the foregoing step S202, so the relevant features described above can be similarly applied here. Figure 2 Description of the relevant features can be similarly applied herein.

[0064] As Figure 3 shown, in step S301, vector tile data is determined in the map vector data according to the detection target.

[0065] In step S302, coordinate conversion and image drawing are performed on the vector tile data according to a preset pixel scale ratio to obtain a target aerial view. The pixel scale ratio refers to the distance length represented by each pixel for the map vector data.

[0066] In the above steps, when the detection target is road dead-end detection, the map vector data is subjected to map sheet splitting processing according to the detection target, including: performing map sheet splitting processing on the map vector data according to preset splitting parameters to obtain multiple sub-map sheets, as Figure 4As shown, frame 401 represents the map vector data with a map size of 10Km * 10Km, and frame 402 represents a sub-map of 200m * 200m. The preset splitting parameters include but are not limited to the parameters for realizing the splitting, such as the size of each sub-map.

[0067] Through the above splitting process, the amount of data processed at one time can be reduced, the processing speed of target detection can be accelerated, the computational complexity of the model can be reduced, and the accuracy of target detection can be improved.

[0068] When the detection target is road connectivity detection, the map vector data is subjected to map splitting processing according to the detection target, including splitting the map vector data according to the candidate connectivity paths to obtain multiple sub-maps corresponding to the candidate connectivity paths. The candidate connectivity paths are obtained by searching in the second map data according to the connectivity candidates, and the connectivity candidates are obtained according to the road network nodes of the first map data and the road network information for describing complex traffic intersections. As Figure 5 As shown, frame 501 represents the map data after converting the map vector data with a map size of 10Km * 10Km into a local rectangular coordinate system. Frames 5021, 5022, and 5023 respectively represent the sub-maps corresponding to the candidate connectivity paths. Among them, frames 5021, 5022, and 5023 correspond to the same connectivity candidate pair.

[0069] By splitting the map according to the candidate connectivity paths as described above, for the same connectivity candidate pair, multiple candidate connectivity paths can be obtained, and then a sub-map corresponding to each candidate connectivity path can be obtained. This can reduce the amount of data processing, and classifying and predicting the sub-maps corresponding to the candidate connectivity paths can effectively improve the accuracy of the prediction results.

[0070] In the above steps, coordinate conversion and image drawing for the vector slice data corresponding to each sub-map can include, in the local rectangular coordinate system, converting the vector slice data corresponding to each sub-map from the local rectangular coordinate system to the pixel coordinate system according to the preset pixel scale and the range of the vector slice data to obtain the pixel data corresponding to each sub-map; and drawing an aerial view according to the pixel data corresponding to each sub-map to obtain the target aerial view corresponding to each sub-map.

[0071] When the detection target is road dead-end detection, after the map vector data is sliced into multiple sub-map sheets according to the preset slicing parameters, the method may further include establishing a local rectangular coordinate system with the center of each sub-map sheet as the reference origin; transforming the vector slice data corresponding to each sub-map sheet from the geographic coordinate system to the local rectangular coordinate system. Then, in the local rectangular coordinate system, coordinate transformation and image drawing are performed on the vector slice data corresponding to each sub-map sheet to obtain a target aerial view corresponding to each sub-map sheet. As Figure 4 shown, the image within the dashed circle is the target aerial view corresponding to each sub-map sheet obtained by performing the above coordinate transformation and image drawing processing on the vector slice data within the frame 402.

[0072] When the detection target is road disconnection detection, the map vector data can be transformed from the geographic coordinate system to the local rectangular coordinate system. Then, in the local rectangular coordinate system, the map vector data is sliced according to the candidate connected paths determined by the connected candidate pairs to obtain multiple sub-map sheets corresponding to the candidate connected paths. Then, coordinate transformation and image drawing are performed on the vector slice data corresponding to each sub-map sheet to obtain multiple target aerial views corresponding to the candidate connected paths. As Figure 5 shown, taking the frame 501 as the reference, on the left side of the frame 501 is the target aerial view corresponding to the candidate connected path obtained by performing the above coordinate transformation and image drawing processing on the vector slice data within the frame 5021, where the candidate connected path is {source road surface, connected road surface 1, target road surface} that conforms to the traveling direction, that is, the first candidate connected path, and its connection state is connected, indicating that this connected path is passable. Below the frame 501 is the target aerial view corresponding to the candidate connected path obtained by performing the above coordinate transformation and image drawing processing on the vector slice data within the frame 5023, where the candidate connected path is {source road surface, connected road surface 2, target road surface} that does not conform to the traveling direction, that is, the second candidate connected path, and its connection state is disconnected, indicating that this connected path is impassable. On the right side of the frame 501 is the target aerial view corresponding to the candidate connected path obtained by performing the above coordinate transformation and image drawing processing on the vector slice data within the frame 5022, where the candidate connected path is {source road surface, connected road surface 3, target road surface} that does not conform to the traveling direction, that is, the second candidate connected path, and its connection state is disconnected, indicating that this connected path is impassable.

[0073] Converting the map vector data into a target aerial view through the mapping relationship between the local rectangular coordinate and the pixel coordinate as described above can simplify the calculation and maintain the spatial consistency between the pixel image and the map data, which helps to improve the detection efficiency of map data problems.

[0074] In some embodiments, the map vector data is preprocessed according to the detection target to obtain a target overhead view, and it may further include generating a coordinate transformation file corresponding to the target overhead view according to the mapping relationship of coordinate transformation. After obtaining the target detection result, the map data detection method further includes: performing inverse coordinate transformation on the target detection result according to the coordinate transformation file to output the problem points at the road level corresponding to the target detection result.

[0075] In some embodiments, preprocessing the map vector data according to the detection target may include: determining vector slice data in the map vector data according to the detection target; converting the vector slice data into a target overhead view according to a preset pixel scale, and synchronously generating a coordinate transformation file.

[0076] The road position information corresponding to the target detection result in the global coordinate system includes, but is not limited to, road surface markings. For example, when the target detection result is a detection frame of a dead-end road, the output is the point position data and road marking data of the dead-end road corresponding to the target detection frame in the second map data, etc. The point position data of the road usually refers to the coordinate point information describing the geometric shape of the road. These data contain the position information of elements such as the center line of the road, lane lines, intersections, crosswalks, traffic signs, and markings.

[0077] In some embodiments, when the detection target is road dead-end detection, the target detection result is a target detection frame corresponding to the road dead-end. Performing inverse coordinate transformation on the target detection result according to the coordinate transformation file to output the problem points at the road level corresponding to the target detection result includes: determining the geographical location range corresponding to the target detection frame according to the coordinate transformation file; performing topological calculation on the first map data and the second map data of the corresponding same target area within the geographical location range to obtain the geographical coordinates corresponding to the problem road; directly outputting the geographical coordinates corresponding to the problem road as the dead-end points.

[0078] Through the coordinate transformation of the target detection frame, the geographical location range of the target detection frame in the second map data is obtained, and then the problem points can be determined by geometric intersection of the geographical location range and the map data, which can quickly locate the problem points, improve the processing efficiency, and enhance the user experience.

[0079] In some embodiments, when the target detection is a road disconnection detection, the target detection result includes a first classification prediction result and a second classification prediction result, the first classification prediction result indicating that the source road surface and the target road surface contained in the target bird's-eye view are connected, and the second classification prediction result indicating that the source road surface and the target road surface contained in the target bird's-eye view are disconnected. Before performing inverse coordinate transformation processing on the target detection result according to the coordinate conversion file to output the road-level problem point corresponding to the target detection result, the method also includes: performing classification statistics on the first classification prediction result and the second classification prediction result to obtain a classification statistical result; when the classification statistical result indicates that the first classification prediction result does not exist, performing inverse coordinate transformation processing on the target detection result according to the coordinate conversion file, which includes: converting the pixel coordinates corresponding to the source road surface and the target road surface contained in the target bird's-eye view into geographic coordinates according to the coordinate conversion file; and directly outputting the geographic coordinates corresponding to the source road surface and the target road surface as the disconnected problem point.

[0080] By counting the classification prediction results, when the statistical result of the first classification prediction result is 0, the problem point is output, which is beneficial to improving the accuracy of problem point detection.

[0081] In other embodiments, preprocessing the map vector data according to the detection target to obtain the target bird's-eye view may also include converting the map vector data from a global coordinate system to a local rectangular coordinate system; aligning the first map data as reference data and the second map data as data to be detected to ensure consistency in spatial position; generating non-overlapping areas based on a first layer where the first map data is located and a second layer where the second map data is located, and converting the non-overlapping areas into the target bird's-eye view according to a preset pixel scale.

[0082] The non-overlapping areas are obtained by the above-mentioned vector comparison method, and a bird's-eye view of the target is generated based on the non-overlapping areas of the vector data.

[0083] The following takes the detection target of road break detection as an example. Figure 6 The map data detection method proposed in this disclosure is further described. It can be understood that Figure 6 Steps S6021-S6025 are a specific implementation of the aforementioned step S202, so the aforementioned Figure 2 The related features described can similarly apply here. Figure 6 An exemplary flow chart of a map data detection method 600 according to some further embodiments of the present disclosure is shown.

[0084] like Figure 6 As shown, in step S601, map vector data to be detected is obtained.

[0085] In the above steps, the map vector data includes first map data and second map data corresponding to the same target area, and the accuracy of the first map data is lower than that of the second map data.

[0086] In step S6021, it is determined that the detection target is road break detection.

[0087] In step S6022, the map vector data is segmented to obtain vector slice data corresponding to a plurality of sub-map sheets.

[0088] In step S6023, for each vector slice data, the vector slice data is converted from the global coordinate system to the local rectangular coordinate system.

[0089] In step S6024, for each vector slice data, coordinate transformation and image rendering are performed on the vector slice data according to a preset pixel scale to obtain a target bird's-eye view.

[0090] In step S6025, a coordinate conversion file corresponding to the target bird's-eye view is generated.

[0091] In the above steps, according to the geometric range of the map vector data, SD road network data and HQ data are obtained from the mother database; the map vector data is segmented according to fixed length and width thresholds to generate N*M sub-map spatial segmentation ranges. For example, a range of 10Km*10Km is segmented according to 200m*200m to obtain 50*50 vector slice data, where N is 50 and M is 50. According to the resolution of 1 pixel representing 0.1m, each vector slice data can generate a target bird's-eye view of about 2000*2000 pixels. According to the conversion relationship between geographic coordinates and pixel coordinates, a coordinate conversion file expressed by the target bird's-eye view is generated.

[0092] In the above steps, coordinate conversion and image drawing are performed for the vector slice data corresponding to each sub-map, including establishing a local rectangular coordinate system with the center of each sub-map as the reference origin; transforming the vector slice data corresponding to each sub-map from the global coordinate system to the local rectangular coordinate system; calculating the pixel coordinate system and the size of the image to be generated according to the preset pixel scale and the range of the vector slice data; converting the vector slice data corresponding to each sub-map from the local rectangular coordinate system to the pixel coordinate system to obtain the pixel data corresponding to each sub-map; drawing a bird's-eye view according to the pixel data corresponding to each sub-map to obtain the target bird's-eye view corresponding to each sub-map, and synchronously generating a coordinate conversion file. For example, the coordinate conversion file is generated according to the mapping relationship of the coordinate conversion.

[0093] like Figure 4As shown, the map vector data is divided into multiple vector tile data. According to the conversion relationship between local rectangular coordinates and pixel coordinates, the vector tile data is mapped to obtain a target aerial view corresponding to the vector tile data and a coordinate conversion file corresponding to the target aerial view.

[0094] In step S603, the target aerial view is input into an image detection model corresponding to the detection target to obtain a target detection result.

[0095] In the above steps, the target detection result is used to indicate the problem of the corresponding detection target between the first map data and the second map data. When detecting road breaks, the target detection result is the target detection box in the target aerial view, and the target detection box is used to indicate that there is a problem of the corresponding detection target in the image within the target detection box.

[0096] In some embodiments, multiple target aerial views can be input into a pre-trained image detection model corresponding to road break detection for target detection to obtain one or more target detection boxes. The area corresponding to the target detection box is the target area with road break problems. One target detection box may be detected in one target aerial view, or multiple target detection boxes may be detected. The target detection box can include four coordinate values corresponding in the target aerial view, the upper left pixel coordinate, the lower left pixel coordinate, the upper right pixel coordinate, and the lower right pixel coordinate.

[0097] If the polygonal area corresponding to the target road within the target detection box does not completely cover the linear area of the target road, it indicates that there is a road mismatch problem between the second map data associated with the target aerial view and the first map data, that is, there is a road break.

[0098] In step S6041, the geographical location range corresponding to the target detection result is determined according to the coordinate conversion file.

[0099] In step S6042, topological calculations are performed on the matching first map data and second map data within the geographical location range to obtain the geographical coordinates corresponding to the problem road, and the geographical coordinates corresponding to the problem road are directly output as the break point positions.

[0100] In the above steps, after obtaining the target detection box, according to the coordinate conversion file generated by mapping, the true geographical coordinates of the target detection box in the global coordinate system can be calculated inversely. The geometric intersection of the true geographical geometric coordinate range corresponding to the target detection box and the SD / HQ roads is performed to obtain which SD and HQ roads are covered by the target detection box. These SD roads and HQ roads are used as the output results, and the output results include, but are not limited to, road-level break point positions and road data information. Such as Figure 8As shown, the road surface within the target detection frame may include but is not limited to point data and road surface data (not shown in the figure).

[0101] The above-mentioned direct output of road-level problem points is directly put into use after processing the problem points, which effectively shortens the processing time and improves the detection efficiency.

[0102] Compared with the map data detection method proposed in the related art, the technical solution provided by the present disclosure segments the map vector data under road break detection, and then converts the segmentation result into a target bird's-eye view, and then uses the image detection model to perform target detection on the target bird's-eye view to obtain the target detection result. The segmentation process can reduce the amount of data required for a single processing, speed up the processing speed of target detection, reduce the computational complexity of the model, and improve the accuracy of target detection, thereby improving the efficiency of map data detection.

[0103] The following example takes the detection target of disconnected roads as an example. Figure 7 The map data detection method provided by the present disclosure is further described. Figure 7 An exemplary flow chart of a map data detection method 700 according to yet other embodiments of the present disclosure is shown. Figure 7 Steps S7021-S7027 are a specific implementation of the aforementioned step S202, so the aforementioned Figure 2 The related features described can similarly apply here.

[0104] like Figure 7 As shown, in step S701, map vector data to be detected is obtained.

[0105] In the above steps, the map vector data includes first map data and second map data corresponding to the same target area, and the accuracy of the first map data is lower than that of the second map data.

[0106] In step S7021, it is determined that the detection target is a road disconnection detection.

[0107] In step S7022, the map vector data is transformed from the global coordinate system to the local rectangular coordinate system.

[0108] In step S7023, in a local rectangular coordinate system, connected candidate pairs are generated based on the road network nodes of the first map data and the road network information used to describe the complex traffic intersection.

[0109] In the above steps, each connectivity candidate pair includes at least a source road and a target road.

[0110] In the above steps, at least one connected candidate pair is generated according to the road network nodes of the first map data and the road network information for describing complex traffic intersections, which may include: obtaining the node information of the road network nodes from the first map data, where the node information includes a first node corresponding to the source road and a second node corresponding to the target road, and the road network information between the source road and the target road; obtaining the road network information for describing complex traffic intersections from the first map data, where the road network information includes other nodes for connecting the first node and the second node, and the traffic information from the source road identifier, passing through the roads corresponding to the other nodes, to the target road identifier; obtaining at least one connected candidate pair according to the node information of the road network nodes and the road network information for describing complex traffic intersections.

[0111] In step S7024, a set of road surfaces corresponding to the connected candidate pair is determined in the second map data.

[0112] In the above steps, the set of road surfaces includes a source road surface corresponding to the source road, a target road surface corresponding to the target road, and at least one connected road surface, where the connected road surface is a road surface having a road association relationship with the source road surface and the target road surface.

[0113] In some embodiments, determining the set of road surfaces corresponding to the connected candidate pair in the second map data may further include determining a source road surface associated with the source road and a target road surface associated with the target road in the second map data; creating a virtual line between the tail determined along the traveling direction of the source road surface and the starting part determined along the traveling direction of the target road surface; creating a buffer space according to the virtual line, and searching for a road surface having a road association relationship with the source road surface within the buffer space as the connected road surface.

[0114] In some embodiments, a reference virtual line may also be generated using a deep learning model according to the traveling direction between the source road surface and the target road surface. The virtual line created between the tail in the traveling direction of the source road surface and the starting part in the traveling direction of the target road surface is defined as the first virtual line. The higher the fitting degree of the first virtual line and the reference virtual line, the higher the accuracy of the connected road surface found within the buffer area based on the first virtual line.

[0115] By creating a virtual line, creating a buffer area according to the virtual line, and searching for a connected road surface within the buffer area, the detection area can be accurately located, an automated process can be realized, the need for manual adjustment can be reduced, and the consistency of the target detection result can be ensured.

[0116] In step S7025, the set of road surfaces is searched according to the road connectivity and legality conditions to obtain a first candidate path and a second candidate path, and the first candidate path and the second candidate path are respectively determined as vector tile data.

[0117] In the above steps, a candidate connected path refers to starting from the source road surface, along the traveling direction, sequentially searching for road surfaces adjacent to the source road surface, and when the adjacent road surface is connected to the source road surface, continuing to search for the associated road surface along the traveling direction until the target road surface is found, obtaining a connected path composed of road surfaces along the traveling direction, and taking this connected path as the first candidate connected path. During the process of searching for the connected path, the road surface adjacent to the source road may connect the source road surface and the target road surface, indicating that a connected connected path can be constructed; the road surface adjacent to the source road may also not connect the source road surface and the target road surface, indicating that a non-connected connected path can be constructed. This non-connected connected path indicates that there may be a non-connection problem between the source road surface and the target road surface in the second map data, and it needs to be output for data correction to improve the accuracy of the second map data.

[0118] In some embodiments, searching for a connected path in the set of road surfaces may further include traversing in the set of road surfaces along the traveling direction of the source road surface to find a connected road surface having a road association relationship with the source road surface; according to the traffic legality between the source road surface and the target road surface and the road surface connection judgment condition, taking the source road surface as the starting point, along the traveling direction, judging the connectivity of two adjacent road surfaces; when the judgment result indicates that the two adjacent road surfaces are connected, updating the starting point to the connected road surface adjacent to the source road surface, returning to the step of judging the connectivity of two adjacent road surfaces along the traveling direction until the road surface adjacent to the starting point is the target road surface, establishing a target connected path between the source road surface and the target road surface, and taking the target connected path as the first candidate connected path; or, when the judgment result indicates that the two adjacent road surfaces are not connected, updating the road surface adjacent to the starting point, returning to the step of judging the connectivity of two adjacent road surfaces along the traveling direction until all road surfaces adjacent to the starting point are traversed; when there is no connected path between the source road surface and the target road surface, at least taking the source road surface and the target road surface as the second candidate connected path.

[0119] In some embodiments, it may be to take the source road surface, the connected road surface associated with the source road surface, and the target road surface as the second candidate connected path, where the source road surface and the connected road surface associated with the source road are partially connected, but the source road surface and the target road surface are not connected. As Figure 9 shown, the second candidate connected path may be {source road surface, connected road surface 3, target road surface}, or {source road surface, connected road surface 2, target road surface}. As Figure 10 shown, the second candidate connected path may also be {source road surface 11, target road surface 12}.

[0120] In some embodiments, the legality of passage between the source road surface and the target road surface refers to the travel direction that complies with traffic rules. For example, the left-turn travel direction that complies with traffic rules at a certain intersection complies with traffic rules. The judgment conditions for road surface connectivity include, but are not limited to: judging whether two adjacent road surfaces intersect; judging whether the elevation difference between two adjacent road surfaces meets the threshold; judging the intersection area of two adjacent road surfaces; judging whether the travel directions of two adjacent road surfaces are the same; judging whether the two lane lines associated with each road surface are smoothly connected, and the distance between the two lane lines is greater than the distance of a single lane.

[0121] Judge whether two adjacent road surfaces intersect. If they do not intersect, it means that passage is not possible. Update the road surface adjacent to the source road surface and return to this step for a new round of judgment.

[0122] Judge whether the elevation difference between two adjacent road surfaces meets the threshold. If the elevation difference is large, it means that the two adjacent road surfaces may be elevated roads or upper and lower layer roads and passage is not possible. Update the road surface adjacent to the source road surface and return to this step for a new round of judgment.

[0123] Judge the intersection area of two adjacent road surfaces. If the intersection area is too large, it means that the two adjacent road surfaces are not passable. Update the road surface adjacent to the source road surface and return to this step for a new round of judgment. Because the passable states of two adjacent road surfaces are smoothly connected and the two adjacent road surfaces are seamlessly connected.

[0124] Judge whether the directions of two adjacent road surfaces are the same. Different directions mean that the two adjacent road surfaces are not passable. Update the road surface adjacent to the source road surface and return to this step for a new round of judgment.

[0125] Judge whether the travel directions of two adjacent road surfaces are from the starting position to the ending position. If they are not the same travel direction, it means that the two adjacent road surfaces are not passable. Update the road surface adjacent to the source road surface and return to this step for a new round of judgment. By judging whether it is the travel direction from the starting position to the ending position, it is possible to prevent the connected route from going backwards.

[0126] Judge whether the two lane lines associated with each road surface are smoothly connected. Judge whether the lane lines associated with one road surface are smoothly connected to more than two (one lane is expressed by two lane lines on the left and right), and the distance between the two lane lines meets the distance standard of a single lane. If not, update the road surface adjacent to the source road surface and return to this step for a new round of judgment.

[0127] After the above judgment, two adjacent road surfaces that meet all the above judgment conditions are connected and passable. If the connection and passage judgment conditions cannot be continuously met during the search process, record this connected path as the second candidate connected path. Then, continuously search for candidate connected paths starting from the source road surface among the remaining candidate road surfaces. Through these judgments, multiple candidate connected paths can be obtained.

[0128] In step S7026, perform coordinate conversion and image drawing on the vector tile data according to the pre-set pixel scale to obtain the target bird's-eye view.

[0129] In some embodiments, the coordinate conversion and image drawing for the vector tile data corresponding to each candidate connected path may include: in the local rectangular coordinate system, according to the pre-set pixel scale and the range of the vector tile data, convert the vector tile data corresponding to each sub-map sheet from the local rectangular coordinate system to the pixel coordinate system to obtain the pixel data corresponding to each sub-map sheet; perform bird's-eye view drawing based on the pixel data corresponding to each sub-map sheet to obtain the target bird's-eye view corresponding to each sub-map sheet.

[0130] In step S7027, generate a coordinate conversion file corresponding to the target bird's-eye view.

[0131] Through the above picture-taking method, a coordinate conversion file corresponding to the target bird's-eye view can be generated according to the mapping relationship of the coordinate conversion. The picture-taking method can simplify the calculation and maintain the spatial consistency between the pixel image and the map data.

[0132] In step S703, input the target bird's-eye view into the image detection model corresponding to the detection target to obtain the target detection result.

[0133] In the above steps, the target detection result is used to indicate the problem of the detection target corresponding to the first map data and the second map data. When the detection target is road disconnection detection, the target detection result is the classification prediction result corresponding to the road disconnection detection. The target detection result includes the first classification prediction result and the second classification prediction result. Among them, the first classification prediction result indicates that the source road surface and the target road surface included in the target bird's-eye view are connected, and the second classification prediction result indicates that the source road surface and the target road surface included in the target bird's-eye view are not connected.

[0134] In step S7041, perform classification statistics on the target detection result to obtain the classification statistics result.

[0135] In the above steps, such as Figure 9As shown, taking the center of the figure as a reference, the classification prediction result corresponding to the left side of the figure center is the first classification prediction result, indicating that the candidate connected path {source road surface, connected road surface 1, target road surface} is connected in the traveling direction. The classification prediction result corresponding to the right side of the figure center is the second classification prediction result, indicating that the candidate connected path {source road surface, connected road surface 3, target road surface} is not connected in the traveling direction. The classification prediction result corresponding to the lower part of the figure center is the second classification prediction result, indicating that the candidate connected path {source road surface, connected road surface 3, target road surface} is not connected in the traveling direction.

[0136] In step S7042, when the classification statistical result indicates that there is no first classification prediction result, the pixel coordinates corresponding to the source road surface and the target road surface included in the target aerial view are converted into geographic coordinates according to the coordinate conversion file, and the geographic coordinates corresponding to the source road surface and the target road surface are directly output as the unconnected problem points.

[0137] In the above steps, when the number of the first classification prediction results is the first value, the point data and road surface data of the source road surface and the target road surface are output, where the first value is 0. The number of the first classification prediction results being the first value indicates that the source road surface and the target road surface associated with the connection candidate pair in the second map data are not connected. When the number of the first classification prediction results is greater than or equal to the second value, it indicates that the source road surface and the target road surface associated with the connection candidate pair in the second map data are connected, where the second value can be 1.

[0138] By outputting the problem points when the statistical quantity of the classification prediction results is 0 as described above, the accuracy of problem point detection can be effectively improved. As Figure 10 shown, Figure 10 The classification prediction result on the left is the second classification prediction result, that is, it is not connected between the source road surface 11 and the target road surface 12, and the point data and road surface data of the source road surface, and the point data and road surface data of the target road surface are output.

[0139] In the disclosed embodiment, the map vector data is used to determine the candidate connected path associated with the candidate connection pair in the second map data according to the candidate connection pair of the first map data, and then the regions corresponding to the candidate connected path are respectively obtained from the map vector data as sub-map sheets, and then the sub-map sheets are converted into the corresponding target aerial views, and then the image detection model is used to classify and predict the target aerial views, find out the unconnected candidate connected paths as problem points, and finally directly output the road-level point data and road data of the problem points, which effectively improves the efficiency of automatic detection and further improves the accuracy of the detection result.

[0140] Based on the same inventive concept, some embodiments of the present disclosure describe a map data detection device. The implementation scheme provided by the device to solve the problem is similar to the implementation scheme recorded in the above method. For the specific definition of the map data detection device, please refer to the definition of the map data detection method above, which will not be repeated here.

[0141] Figure 11 FIG. 1 is a schematic block diagram showing the structure of a map data detection device 1100 according to some embodiments of the present disclosure. Figure 11 As shown, the map data detection device 1100 includes:

[0142] The data acquisition module 1101 is configured to acquire map vector data to be detected, where the map vector data includes first map data and second map data corresponding to the same target area.

[0143] The data preprocessing module 1102 is configured to preprocess the map vector data according to the detected target to obtain a bird's-eye view of the target.

[0144] The data problem detection module 1103 is configured to input the target bird's-eye view image into an image detection model corresponding to the detection target to obtain a target detection result, where the target detection result is used to indicate a problem with the detection target corresponding to the first map data and the second map data.

[0145] In some embodiments, the data preprocessing module 1102 includes: a vector slice data determination unit, configured to determine the vector slice data in the map vector data according to the detected target; a picture taking unit, configured to perform coordinate conversion and image drawing on the vector slice data according to a preset pixel scale to obtain a bird's-eye view of the target, where the pixel scale refers to the distance length of each pixel used to represent the map vector data.

[0146] In some embodiments, the image capturing unit is further configured to convert the local rectangular coordinates corresponding to the vector slice data into pixel coordinates according to the pixel scale; and draw the image according to the pixel coordinates to obtain a bird's-eye view of the target.

[0147] In some embodiments, when the detection target is road end detection, the vector slice data determination unit is configured to perform map sheet segmentation processing on the map vector data to obtain vector slice data corresponding to multiple sub-map sheets one by one; for each vector slice data, the vector slice data is converted from the global coordinate system to the local rectangular coordinate system.

[0148] In some embodiments, when the detection target is road disconnection detection, the vector tile data determination unit is configured to convert the map vector data from the global coordinate system into a local rectangular coordinate system; in the local rectangular coordinate system, based on the road network nodes and complex traffic intersection information of the first map data, generate connected candidate pairs; determine the road surface set corresponding to the connected candidate pairs in the second map data; search the road surface set according to the road connectivity and legality conditions to obtain a first candidate path and a second candidate path, and respectively determine the first candidate path and the second candidate path as vector tile data.

[0149] In some embodiments, the vector tile data unit is further configured to determine a source road surface and a target road surface corresponding to the connected candidate pairs, and create a virtual line between the two; construct a buffer based on the virtual line; find the connected road surfaces associated with the source road surface in the buffer to determine the road surface set.

[0150] In some embodiments, the vector tile data unit is further configured to start from the source road surface, search for connected road surfaces along the traveling direction, check the connectivity between adjacent road surfaces according to the road connectivity and legality conditions, and find the target road surface through iterative advancement to obtain a first candidate path, where the first candidate path indicates that the source road surface and the target road surface are connected; or, if the target road surface is not found through iterative advancement, record the source road surface and the target road surface as the second candidate path, where the second candidate path indicates that the source road surface and the target road surface are not connected.

[0151] In some embodiments, the data preprocessing module 1102 is configured to generate a coordinate conversion file corresponding to the target bird's-eye view.

[0152] The device further includes: a detection result output module, which is configured to, after obtaining the target detection result, perform inverse coordinate transformation on the target detection result according to the coordinate conversion file to output the problem points at the road level corresponding to the target detection result.

[0153] In the embodiments of the present disclosure, by preprocessing the map vector data according to the detection target, establishing the transformation relationship between the local rectangular coordinates and pixel coordinates according to the preset pixel scale, taking the map vector data as the target bird's-eye view and the coordinate conversion file corresponding to the target bird's-eye view, then using the image detection model to perform target detection or classification prediction on the target bird's-eye view, automatically detecting the problem points of road disconnection and dead-end problem points, and finally outputting the problem points at the road level, which effectively improves the efficiency of automatic detection and significantly improves the accuracy of the detection result.

[0154] The technical solutions provided in this disclosure are evaluated for accuracy. The accurate detection index for broken ends reaches 91.6%, and the accurate detection index for disconnection reaches 95.23%, both of which are higher than those of related detection algorithms (the related detection algorithms are about 85%).

[0155] After introducing the above embodiments, next, reference is made to Figure 12 Describe related products of the map data detection method according to the exemplary embodiments of this disclosure.

[0156] Figure 12 A schematic block diagram of the electronic device 1200 according to an embodiment of this disclosure is shown. As Figure 12 shown, the electronic device 1200 may include a processor 1201 and a memory 1202. The memory 1202 stores computer program instructions for performing map data detection. When the computer program instructions are run by the processor 1201, the electronic device 1200 is caused to execute the method described in conjunction with the foregoing Figures 2 to 7 For example, in some embodiments, the electronic device 1200 is used to implement the map data detection method. For details, reference may be made to the foregoing embodiments, which will not be elaborated herein.

[0157] Those skilled in the art know that the embodiments of this disclosure can be implemented as a system, a method, or a computer program product. Therefore, this disclosure can be specifically implemented in the following forms: completely hardware, completely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as "circuit", "module", "unit", or "system". In addition, in some embodiments, this disclosure can also be implemented in the form of a computer program product in one or more computer-readable media, which contain computer-readable program code.

[0158] Any combination of one or more computer-readable media can be adopted. The computer-readable media can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (non-exhaustive examples) of the computer-readable storage medium can include, for example: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, the computer-readable storage medium can be any tangible medium that contains or stores a program, which can be used by or in combination with an instruction execution system, apparatus, or device.

[0159] The program code contained on a computer-readable medium can be transmitted with any appropriate medium, including but not limited to wireless, wire, optical fiber cable, RF, etc., or any suitable combination of the above.

[0160] It should be understood that each block of the flowchart and / or block diagram, and the combination of blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine. These computer program instructions, when executed by a computer or other programmable data processing device, produce a device for implementing the functions / operations specified in the blocks of the flowchart and / or block diagram.

[0161] Although multiple embodiments of the present disclosure have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many changes, alterations, and alternative forms may be contemplated by those skilled in the art without departing from the spirit and scope of the present disclosure. It should be understood that various alternatives to the embodiments of the present disclosure described herein may be employed in practicing the present disclosure. The appended claims are intended to define the scope of the present disclosure and thus cover equivalents or alternatives within the scope of these claims.

Claims

1. A method for detecting map data, characterized in that, The method includes: Acquire map vector data to be detected, wherein the map vector data includes first map data and second map data corresponding to the same target area; Preprocessing the map vector data according to the detection target to obtain a bird's-eye view of the target; The target bird's-eye view image is input into an image detection model corresponding to the detection target to obtain a target detection result, wherein the target detection result is used to indicate a problem in which the first map data and the second map data correspond to the detection target.

2. The method according to claim 1, wherein Preprocessing the map vector data according to the detection target includes: Determining vector slice data in the map vector data according to the detection target; The vector slice data is subjected to coordinate transformation and image rendering according to a preset pixel scale to obtain the target bird's-eye view map, wherein the pixel scale refers to the distance length used by each pixel to represent the map vector data.

3. The method according to claim 2, wherein The vector slice data is subjected to coordinate transformation and image drawing according to a preset pixel scale, including: Converting the local rectangular coordinates corresponding to the vector slice data into pixel coordinates according to the pixel scale; An image is drawn according to the pixel coordinates to obtain a bird's-eye view of the target.

4. The method according to claim 2, characterized in that When the detection target is road break detection, determining vector slice data in the map vector data according to the detection target includes: Performing a map-sheet segmentation process on the map vector data to obtain vector slice data corresponding to a plurality of sub-map sheets one by one; For each of the vector slice data, the vector slice data is transformed from a global coordinate system to a local rectangular coordinate system.

5. The method according to claim 2, wherein When the detection target is a road disconnection detection, determining vector slice data in the map vector data according to the detection target includes: transforming the map vector data from a global coordinate system into a local rectangular coordinate system; In the local rectangular coordinate system, generating connected candidate pairs according to the road network nodes and complex traffic intersection information of the first map data; determining a road surface set corresponding to the connected candidate pair in the second map data; The road surface set is searched according to road connectivity and legality conditions to obtain a first candidate path and a second candidate path, and the first candidate path and the second candidate path are respectively determined to be vector slice data.

6. The method according to claim 5, wherein Determining a road surface set corresponding to the connected candidate pair in the second map data includes: Determine a source road surface and a target road surface corresponding to the connected candidate pair, and create a virtual line between the two; constructing a buffer zone based on the virtual line; Connected road surfaces associated with the source road surface are searched in the buffer to determine the road surface set.

7. The method according to claim 5, characterized in that, The searching of the road surface set according to the road connectivity and legality conditions comprises: Starting from the source road surface, searching for connected road surfaces along the travel direction, checking the connectivity between adjacent road surfaces according to road connectivity and legality conditions, finding the target road surface through iterative advancement, and obtaining a first candidate path, wherein the first candidate path indicates that the source road surface is connected to the target road surface; or, If the target road surface is not found by iterative advancement, record the source road surface and the target road surface as the second candidate path, where the second candidate path indicates that there is no connection between the source road surface and the target road surface.

8. The method according to claim 1, wherein The preprocessing of the map vector data according to the detection target further includes: Generating a coordinate conversion file corresponding to the target bird's-eye view; After obtaining the target detection result, the method further includes: Performing inverse coordinate transformation on the target detection result according to the coordinate conversion file to output the problem points at the road level corresponding to the target detection result.

9. A map data detection device, characterized in that, The device includes: A data acquisition module configured to acquire map vector data to be detected, where the map vector data includes first map data and second map data corresponding to the same target area; A data preprocessing module configured to preprocess the map vector data according to the detection target to obtain a target bird's-eye view; A data problem detection module configured to input the target bird's-eye view into an image detection model corresponding to the detection target to obtain a target detection result, where the target detection result is used to indicate the problems of the first map data and the second map data corresponding to the detection target.

10. A computer-readable storage medium, characterized in that, Including computer program instructions for implementing a map data detection method by a computer, which, when executed by a processor, enable the implementation of the method according to any one of claims 1-8.

Images