Method for generating and updating high-precision map of automatic driving scene library

Abstract

The invention discloses a method for generating and updating a high-precision map of an automatic driving scene library, and belongs to the technical field of automatic driving, and the method comprises the following steps: S1, carrying out the element feature recognition of an image obtained through frame extraction, extracting the road network element features in a target detection frame, and reserving an effective target detection frame; s2, constructing a road network element connection relation; s3, the missing road network elements and the connection relation are compensated; s4, the road network elements are semantized, element attributes are configured, and high-precision map data are output. According to the method, the problems of low element identification precision, imperfect connection relation construction, poor missing compensation robustness and the like in the existing map are solved, frame extraction processing and element feature identification are performed by using the high-altitude image video, more comprehensive and accurate road network information can be obtained, and the precision of the high-precision map is improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a method for generating and updating a high-precision map of an autonomous driving scenario library. Background Art

[0002] With the rapid development of autonomous driving technology, high-precision maps, as a key component of the autonomous driving system, play a vital role in the safety and reliability of autonomous driving due to their accuracy and real-time performance.

[0003] High-precision maps are one of the core supporting technologies of autonomous driving systems. They contain a wealth of road information, such as lane lines, traffic signs, intersections, etc., and can provide accurate positioning and navigation support for autonomous driving vehicles. They provide key data for vehicle positioning, path planning, and decision-making control by accurately describing road geometry, traffic signs, lane attributes, and other information. Especially in the testing phase of autonomous driving, a high-precision map format with clear road feature elements is required to support efficient testing and optimization of autonomous driving algorithms, so that intelligent driving vehicles can meet safety requirements when on the road and choose the most appropriate driving strategy based on actual road conditions to ensure efficient and safe autonomous driving effects.

[0004] However, there are still many technical problems with the existing high-precision map generation technology. The traditional method of identifying small targets (such as traffic markings and signs) is easily limited by image resolution, resulting in insufficient recognition accuracy of road feature elements, and easy omission or misdetection. The existing map coordinate system conversion error is large. When converting image data collected by drones or vehicle-mounted equipment into the actual map coordinate system, the equipment posture and the influence of the earth's curvature are not fully considered, resulting in spatial mapping deviation. The connection relationship between road network elements in many map systems is not well constructed. The correlation judgment of elements such as lanes, intersections, and traffic markings relies on simple rules, and lacks dynamic analysis of traffic sign semantics and lane direction consistency. Existing technologies do not provide a reliable compensation method for missing data, or the compensation effect is poor and inconsistent with the actual road network topology, affecting the reliability of the map.

[0005] The above problems lead to the insufficient accuracy of the generated high-precision maps in complex scenarios, making it difficult to meet the high safety requirements of autonomous vehicles. Therefore, there is an urgent need for a method that can efficiently generate and dynamically update high-precision maps to improve element recognition accuracy, optimize connection relationship construction logic, and realize intelligent missing compensation and exception handling. Summary of the invention

[0006] The present invention overcomes the problems of low element recognition accuracy, imperfect connection relationship construction, poor robustness of missing compensation in existing maps, and provides a high-precision map generation and update method for an autonomous driving scenario library. It uses high-altitude image video for frame extraction and element feature recognition, combines a variety of data processing technologies, and integrates multi-source data to obtain more comprehensive and accurate road network information, improve the accuracy of high-precision maps, compensate for missing road network elements and connection relationships, and ensure that map data is consistent with actual road conditions.

[0007] In order to achieve the above object, the present invention adopts the following scheme: The method for generating and updating a high-precision map of an autonomous driving scenario library comprises the following steps: S1: Obtain high-altitude video and perform frame extraction, set a target detection frame for the image obtained by frame extraction, and extract the characteristics of each road network element in the target detection frame, including element attributes and relevance; S2: Constructing the connection relationship of road network elements according to the element attributes and relevance of each road network element, including the following processes: Determine the lane-to-lane connection based on the relative position, direction consistency, and traffic markings of the two lanes; Determine the lane-intersection connection relationship based on the lane direction and the direction of the intersection traffic sign; The binding relationship between traffic markings and lanes is established based on the minimum distance matching principle between traffic markings and lanes and the functional rules of traffic markings; Determine the mapping relationship between traffic signs and lanes based on the spatial proximity and matching of sign semantic rules with lane attributes; S3: Check the integrity of the connection relationship of the road network elements and compensate for the missing road network elements and connection relationships, including: Compensate for missing lane elements based on adjacent lane information, intersection connectivity, and traffic markings; Compensate for missing traffic marking elements based on identified traffic markings and their continuity principles and setting rules; Re-analyze the fuzzy lane connection relationship, repair the wrong lane connection relationship and compensate for the missing lane connection relationship; for the newly identified lane, supplement the connection relationship between the new lane and the identified lane; S4: semanticize the compensated road network elements and connection relationships and configure element and relationship attributes, format the geometric expressions of element and relationship attributes in XML and output them as high-precision map data.

[0008] Preferably, in step S1, after setting the target detection frame for the image obtained by frame extraction, a step of converting the image coordinate system into the actual map coordinate system is also included, which includes the following specific steps: In the image coordinate system, take half of the image width and height pixel values ​​to get the coordinates of the image center, and translate the vertex coordinates of the target detection box to the coordinate system with the image center as the origin; Construct the rotation matrix R(θ) based on the device yaw angle θ: The coordinates after translation (x p ,y p ) to perform a rotation transformation: The rotated coordinates (x r ,y r ) is superimposed on the actual geographic coordinates of the device to complete the spatial position mapping, and the result is corrected by Gauss-Krüger projection to eliminate the earth curvature error.

[0009] Preferably, step S1 further includes enhancing recognition of a small target detection frame whose area is smaller than a preset threshold, which specifically includes the following steps: For target areas whose area is smaller than the preset pixel threshold, the bicubic interpolation algorithm is used to double the magnification: Among them, (u, v) is the pixel coordinate of the image after interpolation and amplification, I zoom (u,v) is the pixel value of the enlarged image, I(x+i,y+j) is the pixel value of the original image; W(t) is the weight function of bicubic interpolation, and its expression is: Where a is the bicubic interpolation smoothing coefficient, corresponding to hour, ; corresponds to When ; Adaptively enhance the contrast of the enlarged area, divide the enlarged area into 8×8 sub-blocks, perform histogram cropping on the sub-blocks and limit the histogram grayscale increment to be less than or equal to 3; The non-local mean denoising algorithm is used to denoise the sub-blocks: the search window size is set to 21×21, the similar block size is set to 7×7, and the pixel weight is calculated by comparing similar blocks within the search window; The element features of the enhanced image area are re-identified. When the intersection-over-intersection ratio of the new and old target detection frames exceeds 0.6, the confidence of the new and old target detection frames is calculated. If the confidence of the new target detection frame is at least 0.2 higher than that of the old target detection frame, the element list is updated with the element features of the new target detection frame.

[0010] Preferably, in step S1, the overlap of the road network elements is further detected, and the overlapping road network elements are hierarchically filtered to retain the valid target detection frame, wherein the hierarchical filtering of the overlapping road network elements includes the following specific steps: Set the intersection-and-merge ratio thresholds for traffic sign elements, lane line elements, and intersection elements respectively. When the intersection-and-merge ratio of the target detection frame exceeds the corresponding intersection-and-merge ratio threshold, perform target detection frame conflict analysis: The conflict graph model is constructed with the target detection boxes as nodes and the overlapping conflict relationships between target detection boxes as edges. The node weights are calculated as follows: Among them, w i is the node weight of the i-th node, A i is the area of ​​the i-th target detection box, α is the category priority coefficient, and β is the area influence coefficient; P r (C i ) is the category priority value, which represents the node category C of the i-th node i The importance priority; The conflict graph model is traversed using the branch and bound algorithm. During the traversal process, the current optimal solution of the node is solved and updated until the upper bound of the subproblem is less than the current optimal value, and the branch is terminated to obtain the maximum weighted independent set. In the maximum weighted independent set, if there is a cross-category conflict between nodes, the target detection box with a large category priority value in the cross-category conflict is retained; if there is a same-category conflict between nodes, the node weight w is retained. i In the first 10% of target detection frames; delete the target detection frames that are not selected for saving, and complete the hierarchical filtering of overlapping road network elements.

[0011] Preferably, in step S2, the lane-to-lane connection relationship is determined in the following manner: Extract the endpoints of the lane centerline to generate a direction vector, and calculate the direction angle between the two lanes through the vector dot product. If the direction angle is less than a predetermined threshold and the minimum centerline distance is less than a distance threshold, the directions are determined to be consistent. A sliding window detection is performed on the gap area between adjacent lanes, and the frequencies of occurrence of white dotted lines and yellow solid lines are counted. A connection relationship is forcibly established between adjacent lanes where the frequency of occurrence of white dotted lines is greater than 0.8, and a connection relationship is prohibited between adjacent lanes where the frequency of occurrence of yellow solid lines is greater than 0.6.

[0012] Preferably, in step S1, the element attributes include lane type, lane direction, and lane speed limit information, and the identification and setting of the lane speed limit information attributes include the following steps: Calculate the radius of curvature R of the lane centerline, and calculate the speed limit V of the curve based on the radius of curvature of the lane centerline c : Where μ is the tire-road friction coefficient, g is the acceleration due to gravity, and θ c is the transverse slope angle; If a speed limit sign value is detected, the lane speed limit information attribute is set to the smaller value between the curve speed limit value and the speed limit sign value; the uphill speed limit is reduced by 2km / h for every 1% slope gradient on steep slopes; and speed transition is performed using linear interpolation within 50m before and after the speed limit change interval of the lane.

[0013] Preferably, in step S3, lane elements are compensated using the following method: Set three control points P0, P1, and P2 in the identified part of the missing lane and calculate the directional derivative α of the adjacent lane: Among them, (x start ,y start ) is the starting point coordinate of the adjacent lane, (x end ,y end ) are the coordinates of the end point of the adjacent lane; Calculate the newly added control point P3: in, is the direction vector, ; k is the expansion coefficient, ; Extend the missing lane based on the cubic Bezier curve and construct the parametric equation: Set the maximum curvature radius limit of the curve. If the calculated curvature radius exceeds the limit, adjust the position of P3 to , ; At the intersection, a uniform B-spline curve is generated with the number of connected lanes as the order, and the node vectors of the curve are distributed in equal proportion; the curvature continuity of the interrupted lane dividing line is repaired according to the parabola equation.

[0014] Preferably, in step S3, the method of compensating the traffic marking elements includes the following steps: Compensate for missing dashed line segments in traffic marking elements: Fourier transform is performed on the solid line segments before and after the dashed line, the spectrum features are extracted and the phase consistency is calculated. If the phase consistency is greater than 0.85, a new dashed line is generated in the missing segment according to the standard dashed-to-solid ratio; To compensate for missing flow lines: The reference radius R is obtained based on the curvature radius R of the lane centerline b :If R≤50m, then the corresponding Rb =R; otherwise R b =50+0.2(R-50); the involute equation in the polar coordinate system is superimposed on the sine wave coefficient to generate a natural transition trajectory, and the polar coordinate system is converted to the Cartesian coordinate system, the steering angle is discretely sampled, and a continuous guide line is generated for compensation.

[0015] Preferably, in step S2, after determining the connection relationship between the lane and the intersection, the intersection is also marked with a driving path priority, including the following steps: Assign unique serial numbers to each entrance lane and exit lane of the intersection, and generate all legal driving paths based on traffic rules; Count the historical driving times of each driving path, normalize the driving times of the same direction path, and generate the efficiency score E ij : Among them, C ij is the number of vehicles traveling from the i-th import lane to the j-th exit lane, C max is the maximum number of trips on the same direction path; For each driving path, calculate the conflict area A between it and the driving paths that can be driven at the same time c and the number of intersections N c , the conflict area and the number of intersections are combined into a conflict index to generate a safety score S ij : Where σ is the exponential adjustment coefficient; The comprehensive priority score of each driving path is obtained by weighting the efficiency score and the safety score according to the weight coefficient: Among them, F ij is the comprehensive priority score of the driving path from the i-th import lane to the j-th exit lane, and γ is the normalized weight coefficient.

[0016] Preferably, in step S3, data anomaly processing is also performed in the following manner: When the number of compensations for an area exceeds 3 times or the final confidence of the area is lower than 0.4, the boundary of the area is extracted and a spiral search path is generated to control the drone to retake the photo. The search path radius is expanded from 10 meters to 50 meters in steps of 5 meters, and the search angle is increased by 30 degrees. The newly collected data is fused with the original data through Kalman filtering. The state equation of Kalman filter fusion adopts a uniform velocity model to construct the observation equation, and the observation matrix is ​​used to directly map the observation position coordinates.

[0017] The present invention has at least the following beneficial effects: (1) By using high-altitude image videos for frame extraction and element feature recognition, combined with a variety of data processing technologies, more detailed information can be obtained compared with traditional acquisition methods. By fusion of multi-source data, more comprehensive and accurate road network information can be obtained, thereby improving the accuracy of high-precision maps; (2) Compensation for missing road network elements and connection relationships, as well as reshooting and fusion during data anomaly processing, ensure that map data is consistent with actual road conditions, and the dynamic update mechanism can timely reflect changes in the road environment; (3) Reasonable road network element connection relationship construction and attribute setting provide a reliable basis for path planning and decision-making of autonomous driving vehicles, provide accurate lane connection relationships, speed limit information, etc., and ensure safe and efficient driving of vehicles; (4) Enhanced recognition of small target detection frames and hierarchical filtering of overlapping road network elements improve the quality and reliability of data and reduce the impact of erroneous information on maps; (5) Due to the use of efficient image acquisition and processing algorithms, the characteristics of various elements in the road can be efficiently and accurately identified, and abnormal data can be compensated, making this method highly adaptable and extensible and applicable to different road scenes and autonomous driving systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A principle block diagram of the method for generating and updating a high-precision map of an autonomous driving scenario library provided by the present invention. DETAILED DESCRIPTION

[0019] The present invention is further described in detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.

[0020] like Figure 1 As shown, the method for generating and updating a high-precision map of an autonomous driving scenario library provided by the present invention comprises the following steps: S1: Acquire high-altitude video and perform frame extraction, perform element feature recognition on the extracted image, and extract the road network element features within the target detection frame; convert the image coordinate system into the actual map coordinate system; enhance the recognition of small target detection frames with an area smaller than a preset threshold, and perform hierarchical filtering on overlapping road network elements to retain valid target detection frames.

[0021] High-altitude video frame extraction processing uses drones or satellites to obtain continuous high-altitude bird's-eye view images, and uses fixed time interval sampling or key frame extraction algorithms (such as scene change detection) to extract video frames to balance data volume and processing efficiency. Element feature recognition and extraction uses convolutional neural networks (such as road network element recognition operator models trained based on YOLO5) to detect road markings, traffic signs, lane boundaries and other elements, generate rectangular target detection frames with category labels, and extract feature vectors such as texture, color, and geometric shape. The result data output by the parsing operator contains the bounding box information of the detected road network elements, that is, the coordinates of the upper left and lower right corners of the bounding box, as well as the corresponding category labels, motor vehicle lanes, intersections, white dotted lines and other information, and is serialized into a list. Each element in the list contains the coordinates and category information of the corresponding road network element.

[0022] The two-dimensional pixel coordinate system of the camera image (the origin is in the upper left corner of the image) is converted to the coordinate system of the map through coordinate system conversion. At the same time, the lens distortion correction and geo-registration parameters are considered to establish the mapping relationship between the pixel point and the earth's ellipsoid. The coordinates of the shooting device in the actual map and the shooting angle are obtained. For the vertex coordinates of each detection frame, first subtract the coordinates of the image center from its coordinates in the image coordinate system to complete the translation operation and move the coordinate origin to the center of the image; then, according to the angle of the shooting device, the trigonometric function calculation is used to rotate; finally, the rotated coordinates are added to the coordinates of the shooting device in the actual map to realize the conversion from the image coordinate system to the actual map coordinate system, thereby obtaining the position of the road network element in the actual map.

[0023] Small target enhancement recognition strategy For small-sized targets such as signboards caused by long-distance shooting, super-resolution reconstruction technology is used to improve the recognition rate. Local feature extraction can also be enhanced by combining the attention mechanism. Traverse the recognition result list and calculate the area of ​​each detection frame. The calculation method is to subtract the upper left corner from the lower right corner of the detection frame, and then multiply the lower right corner from the upper left corner. For detection frames with an area smaller than the preset threshold, these detection frames may correspond to small target road network elements, such as small traffic signs. Extract the image area corresponding to these small targets, use the bicubic interpolation algorithm to enlarge the area, increase the number of pixels, and make the small targets clearer; then enhance the enlarged area again, such as applying histogram equalization again to improve the contrast. After the processing is completed, the enhanced image area is re-input into the operator for recognition, and the recognition result list is updated to improve the recognition accuracy of small targets. For any two detection frames in the recognition result list, their intersection area is calculated. When calculating, first take the larger value of the minimum value of the horizontal coordinates of the two detection frames and the smaller value of the maximum value of the horizontal coordinates, and subtract the two to get the horizontal width of the intersection area; similarly, take the larger value of the minimum value of the vertical coordinates and the smaller value of the maximum value of the vertical coordinates, and subtract the two to get the vertical height of the intersection area. Multiply the horizontal width and the vertical height to get the intersection area. The union area is calculated by adding the areas of the two detection frames and then subtracting the intersection area. Then calculate the intersection-union ratio, that is, the intersection area divided by the union area. If the intersection-union ratio is greater than the set threshold, it means that the two detection frames may overlap. At this time, according to the priority of the element category (such as traffic signs have a higher priority than lane lines) and the size of the area, the correct detection results are retained, and duplicate or incorrect detection frames are removed to ensure the accuracy of the recognition results. The hierarchical filtering mechanism sets conflict resolution rules according to the semantic priority of traffic elements (such as traffic signs> lane lines> intersections). When the detection frames overlap, according to the results of weighted calculation, high-priority elements are given priority to ensure the integrity of key information.

[0024] S2: Construct the connection relationship of road network elements according to the attributes and associations of each road network element, including the following processes: The connection between lanes is determined based on the relative position, direction consistency, and traffic markings of the two lanes; the topological connection is established by calculating the continuity of the extension direction of the lane centerline (small angle between direction vectors) and physical connectivity (small spacing), combined with the lane line type (such as dashed lines allow lane changes). For any two identified lanes, obtain the coordinate point sequence of their centerlines. Calculate the distance of each point pair between the two centerlines by summing the square of the difference between the horizontal coordinates of the two points and the square of the difference between the vertical coordinates, and then taking the square root. Find the minimum distance among all point pair distances. If the minimum distance is less than the preset distance threshold, it is determined that the two lanes may be connected, because the short distance means that they are close in spatial position and may be connected. Calculate the vertical coordinate difference between the starting point and the end point of the lane centerline and divide it by the horizontal coordinate difference to obtain the slope of the lane centerline, thereby determining the lane direction. For two lanes that may be connected, use the inverse tangent function to calculate the angle between their directions based on their slopes. If the angle is less than the set angle threshold, it further supports the connection between them, because lanes with similar directions are more likely to be connected. For adjacent lanes that may be connected, check the type of traffic markings between them. If the traffic markings are dotted, since dotted lines indicate that crossing is allowed, it usually means that there is a connection between the lanes; if they are solid lines, solid lines indicate that crossing is prohibited, so it is determined that there is no direct connection.

[0025] Determine the connection between the lane and the intersection based on the direction of the lane and the direction of the traffic sign at the intersection; analyze the correspondence between the direction of the lane arrow and the direction of traffic at the intersection, and establish a mapping of lane and intersection access rights. For example, a left-turn lane can only be connected to a left-turn guide arrow. Determine the center coordinates of the intersection and the endpoint coordinates of the lane centerline at one end close to the intersection. Calculate the vector pointing from the endpoint to the center of the intersection by subtracting the endpoint coordinates of the lane centerline from the center coordinates of the intersection. Determine whether the lane is entering the intersection (the vector points to the center of the intersection) or exiting the intersection (the vector points away from the center of the intersection) based on the direction of the vector, thereby determining the connection direction between the lane and the intersection. Traverse the traffic signs at the intersection, such as the arrow sign, and the direction the arrow points to is the direction of the lane at the intersection.

[0026] The binding relationship between traffic markings and lanes is established according to the minimum distance matching principle between traffic markings and lanes, and the functional rules of traffic markings; the nearest neighbor search algorithm is used to match lane lines and adjacent lanes, and spatial and functional associations are established in combination with the marking type (such as the guide line needs to be bound to a specific turning lane). The coordinates of the lane boundary and the traffic marking are obtained, and the minimum distance between the traffic marking and the lane boundary in the horizontal and vertical directions is calculated. In the horizontal direction, the absolute value of the difference between the horizontal coordinate of the traffic marking and the horizontal coordinate of the lane boundary is calculated, and the minimum value is taken; the same is true in the vertical direction. If the distance is less than the set distance threshold, it is determined that the marking and the lane are closely related in position, because the close distance indicates that they are closely related in position. According to the type and function of the traffic marking, it is matched with the driving rules of the lane. For example, the white dotted line is used to separate the same-direction driving lanes. If the white dotted line is detected between two same-direction driving lanes, it is determined that the dotted line matches the function of the two lanes; the yellow solid line is used to separate the opposite-direction driving lanes. If the yellow solid line is detected between two opposite-direction driving lanes, it is determined that the solid line matches the function of the two lanes.

[0027] According to the spatial proximity and the matching of the sign semantic rules with the lane attributes, the mapping relationship between the traffic sign and the lane is determined; the traffic sign coordinates and the lane coordinates are obtained, and the distance between the traffic sign and the lane in the horizontal and vertical directions is calculated. In the horizontal direction, the absolute value of the difference between the traffic sign abscissa and the lane abscissa is calculated; in the vertical direction, the absolute value of the difference between the ordinate is calculated. If both distances are less than the set distance threshold, it is determined that there is a positional correspondence between the traffic sign and the lane, indicating that the traffic sign is set for this lane. According to the meaning and rules of the traffic sign, it is matched with the driving rules of the lane. For example, the arrow sign indicates the driving direction of the lane. If the arrow point is consistent with the determined driving direction of the lane, it is determined that the sign matches the lane rule; the speed limit sign specifies the speed limit value of the lane. If the speed limit value of the speed limit sign is within a reasonable range with the speed limit value inferred from other factors (such as road type) of the lane, it is determined that the sign matches the lane rule. For example, the speed limit of urban main roads is usually high. If the speed limit value of the speed limit sign meets the speed limit range of urban main roads, the two are considered to match.

[0028] S3: Identify and set the attributes of each road network element, including lane type, lane direction, and lane speed limit information; scan and identify the speed limit sign value, and combine the lane line type (solid line / dashed line) to determine the lane function (bus lane, tidal lane), and integrate GPS trajectory data to count the actual lane speed. In the actual map coordinate system, according to the lane boundary coordinates, calculate the sum of the square of the horizontal coordinate difference and the square of the vertical coordinate difference of the two diagonal vertex coordinates, and then take the square root to get the lane width. Compare this width with the standard width range of different types of lanes, such as the standard width range of motor vehicle lanes is 3-3.75 meters, and the standard width range of non-motor vehicle lanes is 1.5-2.5 meters, etc., and determine the lane type by comparison. Check whether there are specific signs on the lane, such as bicycle signs on non-motor vehicle lanes and pedestrian signs on sidewalks. If a bicycle sign is detected, the lane is determined to be a non-motor vehicle lane; if a pedestrian sign is detected, the lane is determined to be a sidewalk. Analyze the width relationship between a lane and the adjacent lane. If the width of a lane is smaller than that of the adjacent lane, and the width of the lane is within the non-motorized vehicle lane width threshold, and the traffic markings between the two lanes meet the rules for separating non-motorized vehicle lanes from motor vehicle lanes, such as a solid white line in the middle, the lane is determined to be a non-motorized vehicle lane.

[0029] The lane direction setting can be based on arrow sign recognition: identify the arrow sign on the lane, and determine the direction of the lane as the driving direction of the lane. It can also be referenced based on the intersection connection relationship: if the lane is connected to the entrance road of the intersection, the lane direction points to the intersection; if it is connected to the exit road, the lane direction is away from the intersection.

[0030] The lane speed limit setting can refer to the speed limit sign recognition: find the detection frame corresponding to the speed limit sign category label from the recognition result, and obtain the speed limit value information in the detection frame. It can also be judged according to the road type: by analyzing the area where the road is located (such as the city center, suburbs, etc.), the surrounding building types (such as commercial areas, residential areas, etc.) and the connection method with other roads, the road type, such as urban main roads, secondary roads, etc., can be judged. For example, a road located in the city center, connecting important commercial areas and residential areas, with a wide road and a large traffic volume, may be determined as an urban main road. In addition, it can be further analyzed according to the characteristics of the road section: consider whether there is a curve in the road section. For a curve, the adjusted speed limit value is calculated based on the curve radius and the vehicle driving safety standard. Generally speaking, the smaller the curve radius, the lower the speed limit value. For example, according to the empirical formula, the speed limit value may be inversely proportional to the square root of the curve radius; for a steep slope, the adjusted speed limit value is calculated based on the slope and the vehicle driving safety standard. The greater the slope, the lower the speed limit value. According to experimental data, the speed limit value decreases by a certain proportion for each increase in slope.

[0031] S4: Compensate for missing network elements and connections, including: Compensate for missing lane elements based on adjacent lane information, intersection connectivity, and traffic markings; Compensation based on adjacent lane information: When lane elements are incomplete in the image, obtain the direction and length information of the adjacent lane. If the adjacent lane is a straight line with a clear direction, extend the incomplete lane in the direction of the adjacent lane. For example, if the adjacent lane is a straight line horizontally to the right, the incomplete lane also extends horizontally to the right. According to the length of the adjacent lane, reasonably estimate the length of the missing part of the incomplete lane. For example, if the length of the adjacent lane is 50 meters, and the ratio between the incomplete lane and the adjacent lane can be determined in the identified part, the length of the missing part is estimated according to this ratio.

[0032] Compensation based on intersection connection relationship: Combined with the connection relationship between the identified lanes and intersections, if a lane is known to be connected to an entrance lane of the intersection and the lane is partially missing in the image, the connection method and extension direction of the missing lane at the intersection are determined based on the structure of the intersection and the conditions of other related lanes. For example, if the intersection is a crossroads, a lane is known to enter the intersection from the east and the lane is partially missing, the correct connection method and extension direction of the missing lane at the intersection are determined based on the direction of other lanes and the design rules of the intersection.

[0033] Compensation based on traffic markings: Observe the traffic markings around the incomplete lane, such as lane dividing lines, edge lines, etc. If the lane dividing line is suddenly interrupted at the incomplete lane, according to its previous direction and the situation of the adjacent lanes, according to the continuity rules of the markings, reasonably infer the extension position of the dividing line, so as to determine the boundary of the lane. For example, if the lane dividing line was previously straight and parallel to the dividing line of the adjacent lane, then its extension position is inferred according to this parallel relationship.

[0034] Compensate for missing traffic marking elements based on identified traffic markings and their continuity principles and setting rules; Compensation based on the continuity of traffic signs: If part of the traffic marking is missing, check the identified traffic markings around it. For continuous markings such as white dotted lines, fill in the missing dotted lines based on the spacing and position of the front and rear dotted lines and the principle of line continuity. For example, if the spacing between the front and rear dotted lines is 5 meters, then fill in the missing part with dotted lines at a spacing of 5 meters.

[0035] Compensation based on traffic marking setting rules: Based on the common setting rules of traffic markings and the surrounding road network elements, the type and location of missing markings are inferred. For example, at intersections, guide lines are usually set at specific locations to guide the direction of vehicle travel. If the guide line at this location is partially missing, the missing guide line is supplemented according to the setting rules of the guide line and the surrounding lanes, traffic signs and other information. For example, guide lines are generally set at the turning point of the intersection. According to the shape of the intersection and the direction of other lanes, the correct position and shape of the guide line are determined for supplementation.

[0036] Re-analyze the ambiguous lane connection relationship, repair the wrong lane connection relationship and compensate for the missing lane connection relationship; when the connection relationship between lanes is uncertain due to missing elements or recognition errors, re-analyze the position, direction and traffic markings of adjacent lanes. Check the continuity of the lane centerline again. If the centerline is interrupted at the connection but can be reasonably connected, such as the direction and position relationship of the centerlines on both sides of the interruption can be connected by simple extension or angle adjustment, the connection relationship can be repaired; according to the guidance of traffic markings and other information, determine their correct connection relationship.

[0037] For newly identified lanes, the connection relationship between the new lanes and the identified lanes is supplemented; if a new lane element is found in some areas, its position, direction, traffic markings and other information relative to the surrounding existing lanes are analyzed. According to the lane connection relationship reasoning method, the connection method between the new and old lanes is determined to supplement the missing connection relationship. For example, if the newly discovered lane is close to the existing lane in position, by analyzing their directions and the traffic markings in the middle, it is determined whether there is a connection relationship and how to connect them.

[0038] The connection between lanes and intersections can also be compensated, including connection correction: when the connection between lanes and intersections is wrong or uncertain, the direction of lanes around the intersection, traffic signs and markings are re-analyzed. Traffic signs, markings and vehicle driving trajectories at intersections are observed to determine the correct connection between lanes and intersections and correct incorrect connection. For example, the connection direction between lanes and intersections that was originally judged to be wrong can be corrected by re-observing traffic signs and vehicle driving trajectories.

[0039] If a new lane element is found near an intersection, analyze the intersection structure, the location of the new and old lanes, and traffic signs. According to the reasoning method of the connection relationship between lanes and intersections, determine the connection method between the newly discovered lane and the intersection, and supplement the missing connection information. For example, if the newly discovered lane is in a corner of the intersection, determine the correct connection method between the lane and the intersection by analyzing the structure of the intersection and the connection of other lanes.

[0040] S5: Semantize the road network elements and configure the element attributes, format the geometric expressions of the road network elements in XML, and then output them as high-precision map data.

[0041] The semantic road network elements, attribute configuration, connection relationship topology and geometric features are encoded into XML format that conforms to the OpenDRIVE standard, and high-precision map data containing lane-level geometric expressions, speed limit attributes and topological connections is output. The categories of elements in the map include intersections, lanes, lane markings, road arrows, road facilities, road signs, etc.

[0042] The specific elements under the intersection category include: intersection, safety island, diversion island, traffic island, intersection center circle, intersection edge line, intersection guide line, and intersection grid line.

[0043] The specific elements under the lane category include: driving lane, central dividing strip, guide line, turning area, and green belt.

[0044] Specific elements under the lane marking category include: solid white line, dashed white line, solid yellow line, dashed yellow line, double solid white line, double dashed white line, double solid yellow line, double dashed yellow line, solid white dashed, solid white dashed, solid white dashed, solid yellow dashed, solid yellow dashed, and stop line.

[0045] Specific elements under the road arrow category include: straight arrow, straight left turn arrow, straight right turn arrow, left turn arrow, right turn arrow, left U-turn, left U-turn, left merge arrow, right merge arrow.

[0046] Specific elements under the road facilities category include: pedestrian crossings, cones, parking spaces, construction signs, motor vehicles, traffic bollards, crash barrels, and water barriers.

[0047] Specific elements under the road sign category include: pedestrian waiting area, grid lines, speed reduction markings, longitudinal speed reduction markings, bicycle signs, pedestrian crossing warnings, text, slow down and give way (lines), slow down and give way (arrows), stop and give way (lines), stop and give way (arrows).

[0048] This method uses high-altitude image videos for frame extraction and element feature recognition, combined with a variety of data processing technologies. Compared with traditional acquisition methods, it can obtain more detailed information. Through multi-source data fusion, it can obtain more comprehensive and accurate road network information, and improve the accuracy of high-precision maps; it compensates for missing road network elements and connection relationships to ensure that map data is consistent with actual road conditions, and the dynamic update mechanism can timely reflect changes in the road environment; reasonable road network element connection relationship construction and attribute setting provide a reliable basis for path planning and decision-making of autonomous driving vehicles, provide accurate lane connection relationships, speed limit information, etc., to ensure safe and efficient driving of vehicles; enhanced recognition of small target detection frames and hierarchical filtering of overlapping road network elements improve the quality and reliability of data and reduce the impact of erroneous information on maps; due to the use of efficient image acquisition and processing algorithms, it can efficiently and accurately identify the characteristics of various elements in the road and compensate for abnormal data, making this method highly adaptable and extensible, and can be applied to different road scenes and autonomous driving systems.

[0049] In another technical solution, in step S1, converting the image coordinate system into the actual map coordinate system includes the following specific steps: In the image coordinate system, take half of the image width and height pixel values ​​to get the coordinates of the image center, and translate the vertex coordinates of the target detection box to the coordinate system with the image center as the origin; Construct the rotation matrix R(θ) based on the device yaw angle θ: The coordinates after translation (x p ,y p ) to perform a rotation transformation: The rotated coordinates (x r ,y r ) is superimposed on the actual geographic coordinates of the device to complete the spatial position mapping, and the result is corrected by Gauss-Krüger projection to eliminate the earth curvature error.

[0050] The problem of odd and even pixel deviation that may exist in traditional solutions is eliminated through coordinate translation quantization. The angle parameters in radians are used to avoid the accumulation of angle conversion errors and improve the calculation accuracy of rotation transformation. The spatial resolution value is dynamically calculated through the geometric relationship between the flight altitude and the camera's field of view. Compared with fixed resolution parameters, it can adapt to different aerial photography scenes. On the basis of the traditional Gauss-Krüger projection linear terms, quadratic and quartic terms can be added to correct the coordinate error from the meter level to the centimeter level. By establishing a strict mathematical symbol system and physical meaning mapping, the coordinate transformation is made more accurate.

[0051] In another technical solution, in step S1, enhancing the recognition of the small target detection frame specifically includes the following steps: For target areas whose area is smaller than the preset pixel threshold, the bicubic interpolation algorithm is used to double the magnification: Among them, (u, v) is the pixel coordinate of the image after interpolation and amplification, I zoom (u,v) is the pixel value of the enlarged image, I(x+i,y+j) is the pixel value of the original image; W(t) is the weight function of bicubic interpolation, and its expression is: Where a is the bicubic interpolation smoothing coefficient, corresponding to hour, ; corresponds to When ; Adaptively enhance the contrast of the enlarged area, divide the enlarged area into 8×8 sub-blocks, perform histogram cropping on the sub-blocks and limit the histogram grayscale increment to be less than or equal to 3; The non-local mean denoising algorithm is used to denoise the sub-blocks: the search window size is set to 21×21, the similar block size is set to 7×7, and the pixel weight is calculated by comparing similar blocks within the search window; The element features of the enhanced image area are re-identified. When the intersection-over-intersection ratio of the new and old target detection frames exceeds 0.6, the confidence of the new and old target detection frames is calculated. If the confidence of the new target detection frame is at least 0.2 higher than that of the old target detection frame, the element list is updated with the element features of the new target detection frame.

[0052] Dynamic screening of small targets is achieved through the small target area threshold to avoid misjudgment caused by fixed size threshold. A smoothing coefficient is introduced into the bicubic interpolation formula to achieve a balance between edge sharpening and noise suppression. The block size in CLAHE contrast enhancement has been experimentally verified to achieve the optimal balance between enhancement effect and computational overhead. Non-local mean denoising uses a combination of a large search window and a small similarity window to improve texture retention. The fusion threshold can be set to IoU>0.6 to prevent repeated box interference; the confidence difference threshold is set to 0.2 to prevent low-quality enhancement results from covering the original correct detection.

[0053] In step S1, hierarchical filtering of overlapping road network elements includes the following specific steps: Set the intersection-and-merge ratio thresholds for traffic sign elements, lane line elements, and intersection elements respectively. When the intersection-and-merge ratio of the target detection frame exceeds the corresponding intersection-and-merge ratio threshold, perform target detection frame conflict analysis: The conflict graph model is constructed with the target detection boxes as nodes and the overlapping conflict relationships between target detection boxes as edges. The node weights are calculated as follows: Among them, w i is the node weight of the i-th node, A i is the area of ​​the i-th target detection box, α is the category priority coefficient, and β is the area influence coefficient; P r (C i ) is the category priority value, which represents the node category C of the i-th node i The importance priority; The conflict graph model is traversed using the branch and bound algorithm. During the traversal process, the current optimal solution of the node is solved and updated until the upper bound of the subproblem is less than the current optimal value, and the branch is terminated to obtain the maximum weighted independent set. In the maximum weighted independent set, if there is a cross-category conflict between nodes, the target detection box with a large category priority value in the cross-category conflict is retained; if there is a same-category conflict between nodes, the node weight w is retained. i In the first 10% of target detection frames; delete the filtered target detection frames to complete the hierarchical filtering of overlapping road network elements.

[0054] For traffic signs (high priority but prone to dense appearance), a lower intersection-and-union ratio threshold (such as 0.4) is set to avoid missing duplicate boxes; for lane lines (low priority but need to be detected continuously), a higher threshold (such as 0.6) is set to retain reasonable overlap. The cross-category conflict forced retention rule is used to solve the problem of category priority being ignored in the traditional independent set algorithm; the edge weight reflects the severity of overlap and provides gradient information for node optimization. The weight function design weakens the advantage of large target area and balances the weight competition between large and small targets; the coefficients α and β are used to achieve quantitative fusion of artificial rules (category priority) and data features (target size). The introduction of the upper bound of the cache subproblem in the branch and bound method can reduce the amount of repeated calculations by about 40%; for conflict graphs with more than 100 nodes, the greedy strategy pre-screening can be enabled.

[0055] In another technical solution, in step S2, the lane-to-lane connection relationship is determined by: Extract the endpoints of the lane centerline to generate a direction vector, and calculate the direction angle between the two lanes through the vector dot product. If the direction angle is less than a predetermined threshold and the minimum centerline distance is less than a distance threshold, the directions are determined to be consistent. A sliding window detection is performed on the gap area between adjacent lanes, and the frequencies of occurrence of white dotted lines and yellow solid lines are counted. A connection relationship is forcibly established between adjacent lanes where the frequency of occurrence of white dotted lines is greater than 0.8, and a connection relationship is prohibited between adjacent lanes where the frequency of occurrence of yellow solid lines is greater than 0.6.

[0056] By dynamically selecting endpoints and extension points, we can avoid misjudgments caused by fixed sampling points, and balance directional sensitivity and noise tolerance through interval design. The angle threshold of the directional angle is set to 15°, which corresponds to the tolerance for curvature changes of the common curve radius (50 meters) of urban roads; the distance threshold can be set to 2 meters (57% of the standard lane width of 3.5 meters) to ensure the physical connectivity of the lane; the frequency threshold of 0.8 / 0.6 in the sliding window confidence mechanism can effectively filter out occasional detection errors; the window size is set to 1m×1m to match the standard size of traffic markings (the length of the dashed line segment is 2-4 meters). Anti-interference design can also be performed for ambiguous situations outside the frequency threshold, and the multi-frame verification mechanism is activated (3 consecutive frames of detection are required to meet the standard). When the minimum lane distance is greater than 5 meters, it is directly determined that there is no connection relationship to reduce the amount of calculation.

[0057] In step S3, setting the lane speed limit information attribute includes the following steps: Calculate the radius of curvature R of the lane centerline, and calculate the speed limit V of the curve based on the radius of curvature of the lane centerline c : Where μ is the tire-road friction coefficient, g is the acceleration due to gravity, and θ c is the transverse slope angle; If a speed limit sign value is detected, the lane speed limit information attribute is set to the smaller value between the curve speed limit value and the speed limit sign value; the uphill speed limit is reduced by 2km / h for every 1% slope gradient on steep slopes; and speed transition is performed using linear interpolation within 50m before and after the speed limit change interval of the lane.

[0058] In the curve frame number, the moving least square method can be used to dynamically fit the curvature radius, which has higher accuracy than the fixed window method. By introducing the lateral slope compensation term tanθ c , solving the defect of traditional formula ignoring ultra-high design.

[0059] Furthermore, when the calculated value is not greater than the speed limit sign value, the speed limit sign value and the curve model calculated value can be weighted and summed at a ratio of 6:4 to obtain the final speed limit value, which respects traffic regulations and takes into account actual physical constraints to prevent abnormal model calculation values; Optimization is carried out on steep slopes, and a step function (attenuation of 2km / h for every 1% slope) is used instead of linear attenuation, which is more in line with the expectations of autonomous driving. A hard lower limit of 20km / h avoids non-physical speed limits on extreme slopes. A smooth transition mechanism is used to match national standards, and linear interpolation ensures the continuity of speed limit changes to avoid the autonomous driving system accidentally triggering emergency braking.

[0060] In another technical solution, in step S4, lane elements are compensated using the following method: Set three control points P0, P1, and P2 in the identified part of the missing lane and calculate the directional derivative α of the adjacent lane: Among them, (x start ,y start ) is the starting point coordinate of the adjacent lane, (x end ,y end ) are the coordinates of the end point of the adjacent lane; Calculate the newly added control point P3: in, is the direction vector, ; k is the expansion coefficient, ; Extend the missing lane based on the cubic Bezier curve and construct the parametric equation: Set the maximum curvature radius limit of the curve. If the calculated curvature radius exceeds the limit, adjust the position of P3 to , ; At the intersection, a uniform B-spline curve is generated with the number of connected lanes as the order, and the node vectors of the curve are distributed in equal proportion; the curvature continuity of the interrupted lane dividing line is repaired according to the parabola equation.

[0061] This compensation mechanism implements a directional adaptive mechanism, dynamically adjusting the extension vector through the direction angle of the adjacent lane to ensure smooth connection between the compensation lane and the existing road network; the tan(α) term is introduced in the expansion factor k to automatically shorten the extension distance in sharp bend scenarios (α is large) to avoid path oscillation. The number of lanes connected at the intersection n determines the number of splines, achieving spline order adaptation and accurate topological expression of complex intersections. The Bezier curve can be discretized with a fixed step size of Δt=0.01 to meet the requirements of balancing accuracy and real-time performance; the B-spline basis function is recursively calculated using the De Boer algorithm.

[0062] In step S4, the method of compensating the traffic marking elements includes the following steps: Compensate for missing dashed line segments in traffic marking elements: Fourier transform is performed on the solid line segments before and after the dashed line, the spectrum features are extracted and the phase consistency is calculated. If the phase consistency is greater than 0.85, a new dashed line is generated in the missing segment according to the standard dashed-to-solid ratio; To compensate for missing flow lines: The reference radius R is obtained based on the curvature radius R of the lane centerline b :If R≤50m, then the corresponding R b =R; otherwise R b=50+0.2(R-50); the involute equation in the polar coordinate system is superimposed on the sine wave coefficient to generate a natural transition trajectory, and the polar coordinate system is converted to the Cartesian coordinate system, the steering angle is discretely sampled, and a continuous guide line is generated for compensation.

[0063] The spectrum analysis uses a 256-pixel window to adapt to the standard dashed line length (6 meters @ 0.05 meters / pixel) to ensure that the spectrum resolution reaches ±0.1 Hz; the phase consistency threshold is set to 0.85 to filter out random noise interference, which can make the misjudgment rate <5%.

[0064] The original design period of the marking is inferred through the dominant frequency f to achieve dynamic period generation, solving the problem of dotted line deformation caused by zooming in and out of the image; a 2:3 virtual-to-real ratio is used to strictly follow the national standard for lane dividing lines.

[0065] In order to achieve deformation control of the guide line, a radius compression strategy is adopted (reducing sensitivity when R>50 meters) to avoid excessive distortion of the guide line in large radius curves; three-cycle sinusoidal modulation is used to generate a naturally transitioned diversion trajectory to adapt to complex scenes such as three-way intersections; the Fourier transform adopts the FFT algorithm to support real-time processing; the polar coordinate sampling interval of 0.01rad corresponds to a ground resolution of 0.1 meter, which meets the requirements of high-precision maps.

[0066] In another technical solution, in step S2, after determining the connection relationship between the lane and the intersection, the intersection is also marked with a driving path priority, including the following steps: Assign a unique serial number to each entrance lane and exit lane of the intersection, and generate all legal driving paths based on traffic rules; the entrance lane and exit lane are lanes for entering and exiting the intersection respectively, and different numbering sequences can be used respectively, and each serial number corresponds to a lane displacement, such as the entrance lanes are numbered 1 to M, and the exit lanes are numbered 1 to N, but they are distinguished by different symbols. An example of setting a legal driving path: the right-turn entrance lane can only be connected to the exit lane on the right. If there are two exit lanes on the right, there are two driving paths for turning right from an entrance lane, which is in compliance with traffic rules; for the entrance lane that can be straight and turn, there are more different options for its driving path and direction.

[0067] Count the historical driving times of each driving path, normalize the driving times of the same direction path, and generate the efficiency score E ij : Among them, C ij is the number of vehicles traveling from the i-th import lane to the j-th exit lane, C max is the maximum number of trips on the same direction path; For each driving path, calculate the conflict area A between it and the driving paths that can be driven at the same time c and the number of intersections N c , the conflict area and the number of intersections are combined into a conflict index to generate a safety score S ij : Where σ is the exponential adjustment coefficient; The comprehensive priority score of each driving path is obtained by weighting the efficiency score and the safety score according to the weight coefficient: Among them, F ij is the comprehensive priority score of the driving path from the i-th import lane to the j-th exit lane, and γ is the normalized weight coefficient.

[0068] The number of historical driving times is a historical record within a period of time. It can be obtained through the traffic system or monitoring statistics to obtain complete data of a representative cycle. The same-direction path is defined as: the import lanes are the same or adjacent and have the right to drive in the same direction (such as multiple lanes for straight driving), and the exit lanes are the same or adjacent. For example, there are three adjacent import lanes R1, R2, R3 and exit lanes L1, L2, L3 in the same direction. There are 9 straight driving paths. The flow of different lanes is usually different, which also causes the selection of lanes and driving paths to affect the efficiency of passing through the intersection. Therefore, it is necessary to calculate the efficiency score of the lane. At the same time, considering that there is a risk of collision when vehicles change lanes (such as changing from R1 lane to L2 lane), the motor vehicle lane close to the non-motor vehicle lane also has the risk of being disturbed by the irregular driving of non-motor vehicles, so it is necessary to consider the safety of the selected lane and calculate the safety score. The efficiency and safety of the driving path affect each other. For example, sometimes the number of vehicles in the more dangerous lane is small, and the efficiency is high when it can pass normally. It may also be due to operations such as meeting or avoiding vehicles, resulting in low passing efficiency. Therefore, a comprehensive score is obtained by weighting the efficiency score and the safety score, so that vehicles driving under braking can have more reference options. When choosing a driving path with a higher comprehensive score, you can better balance safety and traffic efficiency; choosing a driving path with a higher safety score but a lower efficiency score can maximize driving safety and avoid traffic accidents; choosing a situation with a higher efficiency score but a lower safety score is an important test for autonomous driving testing. The highly intelligent control method of the autonomous driving system can respond quickly to various emergencies, seeking to balance driving safety issues while driving in the most efficient way. By marking the scores of different driving paths at the intersection of the high-precision map, the autonomous driving vehicle can choose the appropriate driving path based on actual needs and its own intelligent control level, without relying on the real-time calculation and judgment of the vehicle system.

[0069] In another technical solution, in step S4, data anomaly processing is also performed in the following manner: When the number of compensations for an area exceeds 3 times or the final confidence of the area is lower than 0.4, the boundary of the area is extracted and a spiral search path is generated to control the drone to retake the photo. The search path radius is expanded from 10 meters to 50 meters in steps of 5 meters, and the search angle is increased by 30 degrees. The newly collected data is fused with the original data through Kalman filtering. The state equation of Kalman filter fusion adopts a uniform velocity model to construct the observation equation, and the observation matrix is ​​used to directly map the observation position coordinates.

[0070] Generate a spiral search path for spiral coverage optimization, with an angular step length of Δθ=π / 6 to ensure 30° heading adjustment, and achieve blind spot coverage with an 85% overlap rate; the radius is dynamically expanded (Δρ=5 meters) to adapt to the maximum search radius of 50 meters, taking into account both efficiency and integrity. Use the earth curvature correction term to map the plane spiral to a spherical track, with a longitude conversion error of <0.0001°, to achieve accurate mapping of geographic coordinates.

[0071] Using the Kalman filter dual model, the process model is used for the uniform motion assumption, which is suitable for the stable flight state of the UAV; the observation model performs direct position observation, matches the GPS / RTK positioning data, and can use adaptive noise covariance to achieve multi-source data fusion.

[0072] It should be noted that, although the above describes the various steps in a specific order, it does not mean that the various steps must be performed in the above specific order. In fact, some of these steps can be performed concurrently or even in a different order, as long as the required functions can be achieved. The number of devices and processing scales described here are used to simplify the description of the present invention, and the application, modification and variation of the present invention are obvious to those skilled in the art.

[0073] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes, and they can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.

Claims

1. A method for generating and updating a high-precision map of an autonomous driving scenario library, characterized in that: The following steps are involved: S1: Obtain high-altitude video and perform frame extraction, set a target detection frame for the image obtained by frame extraction, and extract the characteristics of each road network element in the target detection frame, including element attributes and relevance; S2: Constructing the connection relationship of road network elements according to the element attributes and relevance of each road network element, including the following processes: Determine the lane-to-lane connection based on the relative position, direction consistency, and traffic markings of the two lanes; Determine the lane-intersection connection relationship based on the lane direction and the direction of the intersection traffic sign; The binding relationship between traffic markings and lanes is established based on the minimum distance matching principle between traffic markings and lanes and the functional rules of traffic markings; Determine the mapping relationship between traffic signs and lanes based on the spatial proximity and matching of sign semantic rules with lane attributes; S3: Check the integrity of the connection relationship of the road network elements and compensate for the missing road network elements and connection relationships, including: Compensate for missing lane elements based on adjacent lane information, intersection connectivity, and traffic markings; Compensate for missing traffic marking elements based on identified traffic markings and their continuity principles and setting rules; Re-analyze the fuzzy lane connection relationship, repair the wrong lane connection relationship and compensate for the missing lane connection relationship; for the newly identified lane, supplement the connection relationship between the new lane and the identified lane; S4: semanticize the compensated road network elements and connection relationships and configure element and relationship attributes, format the geometric expressions of element and relationship attributes in XML and output them as high-precision map data.

2. The method for generating and updating a high-precision map of an autonomous driving scenario library according to claim 1, characterized in that: In step S1, after setting the target detection frame for the image obtained by frame extraction, the step of converting the image coordinate system into the actual map coordinate system is also included, which includes the following specific steps: In the image coordinate system, take half of the image width and height pixel values ​​to get the coordinates of the image center, and translate the vertex coordinates of the target detection box to the coordinate system with the image center as the origin; Construct the rotation matrix R(θ) based on the device yaw angle θ: The coordinates after translation (x p ,y p ) to perform a rotation transformation: The rotated coordinates (x r ,y r ) is superimposed on the actual geographic coordinates of the device to complete the spatial position mapping, and the result is corrected by Gauss-Krüger projection to eliminate the earth curvature error.

3. The method for generating and updating a high-precision map of an autonomous driving scenario library according to claim 1, characterized in that: Step S1 also includes enhancing the recognition of a small target detection frame whose area is smaller than a preset threshold, which specifically includes the following steps: For target areas whose area is smaller than the preset pixel threshold, the bicubic interpolation algorithm is used to double the magnification: Among them, (u, v) is the pixel coordinate of the image after interpolation and amplification, I zoom (u,v) is the pixel value of the enlarged image, I(x+i,y+j) is the pixel value of the original image; W(t) is the weight function of bicubic interpolation, and its expression is: Where a is the bicubic interpolation smoothing coefficient, corresponding to hour, ; corresponds to When ; Adaptively enhance the contrast of the enlarged area, divide the enlarged area into 8×8 sub-blocks, perform histogram cropping on the sub-blocks and limit the histogram grayscale increment to be less than or equal to 3; The non-local mean denoising algorithm is used to denoise the sub-blocks: the search window size is set to 21×21, the similar block size is set to 7×7, and the pixel weight is calculated by comparing similar blocks within the search window; The element features of the enhanced image area are re-identified. When the intersection-over-intersection ratio of the new and old target detection frames exceeds 0.6, the confidence of the new and old target detection frames is calculated. If the confidence of the new target detection frame is at least 0.2 higher than that of the old target detection frame, the element list is updated with the element features of the new target detection frame.

4. The method for generating and updating a high-precision map of an autonomous driving scenario library according to claim 1, characterized in that: In step S1, the overlap of the road network elements is also detected, and the overlapping road network elements are hierarchically filtered to retain the valid target detection frame, wherein the hierarchical filtering of the overlapping road network elements includes the following specific steps: Set the intersection-and-merge ratio thresholds for traffic sign elements, lane line elements, and intersection elements respectively. When the intersection-and-merge ratio of the target detection frame exceeds the corresponding intersection-and-merge ratio threshold, perform target detection frame conflict analysis: A conflict graph model is constructed with target detection boxes as nodes and overlapping conflict relationships between target detection boxes as edges. Calculate node weight: Among them, w i is the node weight of the i-th node, A i is the area of ​​the i-th target detection box, α is the category priority coefficient, and β is the area influence coefficient; P r (C i ) is the category priority value, which represents the node category C of the i-th node i The importance priority; The conflict graph model is traversed using the branch and bound algorithm. During the traversal process, the current optimal solution of the node is solved and updated until the upper bound of the subproblem is less than the current optimal value, and the branch is terminated to obtain the maximum weighted independent set. In the maximum weighted independent set, if there is a cross-category conflict between nodes, the target detection box with a large category priority value in the cross-category conflict is retained; if there is a same-category conflict between nodes, the node weight w is retained. i In the first 10% of target detection frames; delete the target detection frames that are not selected for saving, and complete the hierarchical filtering of overlapping road network elements.

5. The method for generating and updating a high-precision map of an autonomous driving scenario library according to claim 1, characterized in that: In step S2, the lane-to-lane connection relationship is determined in the following manner: Extract the endpoints of the lane centerline to generate a direction vector, and calculate the direction angle between the two lanes through the vector dot product. If the direction angle is less than a predetermined threshold and the minimum centerline distance is less than a distance threshold, the directions are determined to be consistent. A sliding window detection is performed on the gap area between adjacent lanes, and the frequencies of occurrence of white dotted lines and yellow solid lines are counted. A connection relationship is forcibly established between adjacent lanes where the frequency of occurrence of white dotted lines is greater than 0.8, and a connection relationship is prohibited between adjacent lanes where the frequency of occurrence of yellow solid lines is greater than 0.

6.

6. The method for generating and updating a high-precision map of an autonomous driving scenario library according to claim 1, characterized in that: In step S1, the element attributes include lane type, lane direction, and lane speed limit information. The identification and setting of the lane speed limit information attributes include the following steps: Calculate the radius of curvature R of the lane centerline, and calculate the speed limit V of the curve based on the radius of curvature of the lane centerline c : Where μ is the tire-road friction coefficient, g is the acceleration due to gravity, and θ c is the transverse slope angle; If a speed limit sign value is detected, the lane speed limit information attribute is set to the smaller value between the curve speed limit value and the speed limit sign value; the uphill speed limit is reduced by 2km / h for every 1% slope gradient on steep slopes; and speed transition is performed using linear interpolation within 50m before and after the speed limit change interval of the lane.

7. The method for generating and updating a high-precision map of an autonomous driving scenario library according to claim 1, characterized in that: In step S3, lane elements are compensated using the following method: Set three control points P0, P1, and P2 in the identified part of the missing lane and calculate the directional derivative α of the adjacent lane: Among them, (x start ,y start ) is the starting point coordinate of the adjacent lane, (x end ,y end ) are the coordinates of the end point of the adjacent lane; Calculate the newly added control point P3: in, is the direction vector, ; k is the expansion coefficient, ; Extend the missing lane based on the cubic Bezier curve and construct the parametric equation: Set the maximum curvature radius limit of the curve. If the calculated curvature radius exceeds the limit, adjust the position of P3 to , ; At the intersection, a uniform B-spline curve is generated with the number of connected lanes as the order, and the node vectors of the curve are distributed in equal proportion; the curvature continuity of the interrupted lane dividing line is repaired according to the parabola equation.

8. The method for generating and updating a high-precision map of an autonomous driving scenario library according to claim 1, characterized in that: In step S3, the method of compensating the traffic marking elements includes the following steps: Compensate for missing dashed line segments in traffic marking elements: Fourier transform is performed on the solid line segments before and after the dashed line, the spectrum features are extracted and the phase consistency is calculated. If the phase consistency is greater than 0.85, a new dashed line is generated in the missing segment according to the standard dashed-to-solid ratio; To compensate for missing flow lines: The reference radius R is obtained based on the curvature radius R of the lane centerline b :If R≤50m, then the corresponding R b =R; otherwise R b =50+0.2(R-50); the involute equation in the polar coordinate system is superimposed on the sine wave coefficient to generate a natural transition trajectory, and the polar coordinate system is converted to the Cartesian coordinate system, the steering angle is discretely sampled, and a continuous guide line is generated for compensation.

9. The method for generating and updating a high-precision map of an autonomous driving scenario library according to claim 1, characterized in that: In step S2, after determining the connection relationship between the lane and the intersection, the intersection is also marked with a driving path priority, including the following steps: Assign unique serial numbers to each entrance lane and exit lane of the intersection, and generate all legal driving paths based on traffic rules; Count the historical driving times of each driving path, normalize the driving times of the same direction path, and generate the efficiency score E ij : Among them, C ij is the number of vehicles traveling from the i-th import lane to the j-th exit lane, C max is the maximum number of trips on the same direction path; For each driving path, calculate the conflict area A between it and the driving paths that can be driven at the same time c and the number of intersections N c , the conflict area and the number of intersections are combined into a conflict index to generate a safety score S ij : Where σ is the exponential adjustment coefficient; The comprehensive priority score of each driving path is obtained by weighting the efficiency score and the safety score according to the weight coefficient: Among them, F ij is the comprehensive priority score of the driving path from the i-th import lane to the j-th exit lane, and γ is the normalized weight coefficient.

10. The method for generating and updating a high-precision map of an autonomous driving scenario library according to claim 1, characterized in that: In step S3, data anomaly processing is also performed in the following manner: When the number of compensations for an area exceeds 3 times or the final confidence of the area is lower than 0.4, the boundary of the area is extracted and a spiral search path is generated to control the drone to retake the photo. The search path radius is expanded from 10 meters to 50 meters in steps of 5 meters, and the search angle is increased by 30 degrees. The newly collected data is fused with the original data through Kalman filtering. The state equation of Kalman filter fusion adopts a uniform velocity model to construct the observation equation, and the observation matrix is ​​used to directly map the observation position coordinates.

Images