Road marking simulation design method based on original point cloud data

Abstract

The present application relates to the technical field of road marking simulation, and relates to a road marking simulation design method based on original point cloud data. The present application collects mobile laser scanning original point cloud data of a road scene, analyzes the spatial distribution characteristics of the scanning line point cloud, identifies the road surface point cloud in combination with the local area terrain roughness, compares and identifies the road surface point cloud with the constructed various road marking type marking data sets, obtains marking candidate point clouds, extracts each marking instance in the marking candidate point clouds, performs marking type identification on each marking instance based on a marking edge template library to generate semantic labels, spatially correlates the semantic labels of each marking instance with lane contour data in a lane map, determines the associated lane of each marking instance, checks the layout compliance of each marking instance, and outputs road marking simulation design and verification results, so as to ensure that the road marking simulation design result conforms to the actual traffic specification and improve the standardization level of road marking layout.

Description

Technical Field

[0001] This invention relates to the field of road marking simulation technology, and specifically to a road marking simulation design method based on raw point cloud data. Background Technology

[0002] As a component of road traffic infrastructure, the accurate identification and compliance verification of road markings' location, geometric shape, and semantic type are directly related to the reliability of road traffic safety management and lane-level navigation services. With the development of mobile laser scanning technology, digital reconstruction of road scenes based on high-density point cloud data provides a new technical approach for the automated identification and simulation design of road markings. How to fully utilize the spatial distribution characteristics and geometric semantic information in point cloud data to achieve type identification of road markings and spatial correlation verification with lane maps has become a crucial aspect of intelligent operation and maintenance of road facilities and the construction of digital transportation.

[0003] In the existing technology, the recognition and simulation design of road markings have the following technical problems: 1. The existing technology uses two-dimensional image analysis or manual screening of point clouds to identify markings, which does not make full use of the three-dimensional spatial distribution features and local terrain roughness information in the point cloud data. This results in insufficient separation accuracy between the road surface point cloud and the marking point cloud, and problems such as inaccurate identification of road surface points and incomplete removal of non-road surface interference points. This leads to an increased error rate in marking type recognition and affects the accuracy of road marking simulation design.

[0004] 2. Existing technologies rely solely on simple shape and contour comparison for road marking recognition, without constructing multi-type road marking datasets for point cloud feature comparison. This leads to issues such as confusion between dashed and solid road markings, and misjudgment of directional arrows and pedestrian crossings. Consequently, the reliability of road marking semantic label generation is low, making it difficult to meet the requirements of lane-level map construction for road marking semantic accuracy.

[0005] 3. Existing technologies do not spatially correlate road markings with lane map data after the road marking simulation design is completed. This results in the inability to automatically identify violations of marking layout, and poses hidden dangers such as markings deviating from lane boundaries or crossing the center of lanes. Consequently, the simulation design results do not comply with traffic regulations, affecting the safety assessment of road traffic organization schemes. Summary of the Invention

[0006] This invention aims to overcome the deficiencies in the existing technology and provide a road marking simulation design method based on raw point cloud data. This method enables accurate extraction of road surface points, accurate identification of marking types, accurate association of lanes, and compliance verification of layout, thereby comprehensively improving the accuracy, reliability, and practicality of road marking simulation design.

[0007] The technical solution adopted by this invention to solve its technical problem is as follows: This invention provides a road marking simulation design method based on raw point cloud data, including: S1, collecting raw point cloud data of a moving laser scan of a road scene, analyzing the spatial distribution characteristics of continuous scan line point clouds, including the elevation jump amplitude between adjacent point clouds and the point cloud density per unit length of scan line; screening road surface smoothness measurement points based on elevation jump amplitude and continuous road surface measurement points based on point cloud density to determine candidate road surface point cloud areas; performing plane fitting to calculate the local terrain roughness of the area, screening candidate road surface point clouds, and merging them to obtain the road surface point cloud.

[0008] S2. Based on historical road marking point cloud samples, after primary classification by functional category and secondary classification by geometric shape, extract the average relative elevation, distribution uniformity and plane normal vector direction of the sample point cloud to form sample feature vectors and construct a road marking dataset; compare and identify the road surface point cloud with the road marking dataset to obtain candidate road marking point clouds.

[0009] S3. Cluster the candidate point clouds of the grading lines, obtain the point cloud clusters corresponding to each grading line instance, and extract the boundary geometric parameters of the point cloud clusters.

[0010] S4. Based on the boundary geometric parameters of the point cloud clusters corresponding to each marking instance, and combined with the marking edge template library, mark type identification is performed on each marking instance, and semantic tags are generated for each marking instance.

[0011] S5. Spatial association between the semantic labels of each lane marking instance and the lane outline data in the lane map, calculate the shortest distance between each lane marking instance and all lanes, determine the associated lanes of each lane marking instance, verify the compliance of the layout of each lane marking instance, and output the simulation design and verification results of road markings.

[0012] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention collects the original point cloud data of the moving laser scanning of the road scene, analyzes the spatial distribution characteristics of the continuous scanning line point cloud, and identifies the road surface point cloud by combining the local terrain roughness, so as to achieve accurate separation of the road surface point cloud and background noise, improve the integrity and accuracy of the road surface point cloud extraction, effectively reduce the marking recognition error, and provide a high-quality data foundation for subsequent marking recognition.

[0013] (2) This invention compares and identifies road surface point clouds with a variety of road marking type datasets to obtain candidate point clouds of markings, extracts each marking instance in the candidate point clouds of markings, and generates semantic labels for each marking instance based on the marking edge template library. This achieves accurate identification and semantic annotation of marking types, avoids the problem of marking type confusion, improves the reliability of marking semantic label generation, and meets the requirements of lane-level map construction for marking semantic accuracy.

[0014] (3) This invention spatially associates the semantic tags of each lane marking instance with the lane outline data in the lane map, calculates the shortest distance between each lane marking instance and all lanes, and determines the associated lanes, thereby accurately locating the lane to which the lane marking belongs, providing a reliable spatial benchmark for compliance verification, effectively solving the problem of lane marking and lane matching deviation, and improving lane matching accuracy.

[0015] (4) Based on the standard layout rules for different types of road markings in the road design specification knowledge base, this invention verifies the compliance of the layout of each marking instance, identifies the marking instances that violate the layout rules, and outputs the simulation design and verification results of road markings, ensuring that the simulation design results of road markings conform to the actual traffic specifications and improving the standardization level of road marking layout. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the method steps of the present invention;

[0018] Figure 2 This is a schematic diagram of the road surface point cloud recognition steps in this invention;

[0019] Figure 3 This is a schematic diagram illustrating the semantic tag generation steps for each punctuation instance in this invention. Detailed Implementation

[0020] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention. Furthermore, it should be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale.

[0021] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use. Techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and apparatus should be considered part of the specification.

[0022] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0023] Please see Figure 1 As shown, this invention provides a road marking simulation design method based on raw point cloud data, including: S1, collecting raw point cloud data from a moving laser scan of a road scene, analyzing the spatial distribution characteristics of continuous scan line point clouds, including the elevation jump amplitude between adjacent point clouds and the point cloud density per unit length of scan line; filtering road surface smoothness measurement points based on elevation jump amplitude and continuous road surface measurement points based on point cloud density to determine candidate road surface point cloud regions; performing plane fitting to calculate local terrain roughness in the region, filtering candidate road surface point clouds, and merging them to obtain the road surface point cloud.

[0024] S2. Based on historical road marking point cloud samples, after primary classification by functional category and secondary classification by geometric shape, extract the average relative elevation, distribution uniformity and plane normal vector direction of the sample point cloud to form sample feature vectors and construct a road marking dataset; compare and identify the road surface point cloud with the road marking dataset to obtain candidate road marking point clouds.

[0025] S3. Cluster the candidate point clouds of the grading lines, obtain the point cloud clusters corresponding to each grading line instance, and extract the boundary geometric parameters of the point cloud clusters.

[0026] S4. Based on the boundary geometric parameters of the point cloud clusters corresponding to each marking instance, and combined with the marking edge template library, mark type identification is performed on each marking instance, and semantic tags are generated for each marking instance.

[0027] S5. Spatial association between the semantic labels of each lane marking instance and the lane outline data in the lane map, calculate the shortest distance between each lane marking instance and all lanes, determine the associated lanes of each lane marking instance, verify the compliance of the layout of each lane marking instance, and output the simulation design and verification results of road markings.

[0028] Considering that the original point cloud data from mobile laser scanning contains three-dimensional spatial distribution information, and existing technologies do not utilize the three-dimensional spatial features of the original point cloud, resulting in insufficient accuracy in separating the road surface point cloud and failing to provide reference data for road marking recognition, it is necessary to analyze the spatial distribution features of the scan lines and quantify the local terrain roughness to achieve accurate extraction of the road surface point cloud and reduce road marking recognition errors.

[0029] Based on this, such as Figure 2 As shown, the specific implementation of parsing the spatial distribution characteristics of continuous scan line point clouds and identifying road surface point clouds in this invention includes: S11, parsing the scan lines of the original point cloud data from the mobile laser scan to obtain a set of continuous scan line point clouds arranged in the scanning sequence. Each continuous scan line contains multiple spatial point clouds collected at the same time during the laser scan.

[0030] It should be noted that the mobile laser scanning raw point cloud data is collected by a laser scanning device mounted on a mobile surveying vehicle along the direction of travel of the road, and a three-dimensional point cloud dataset including the road surface, roadside facilities, vegetation and background buildings is obtained.

[0031] S12. Obtain the elevation coordinate sequence of the point cloud of each scan line, calculate the elevation difference between adjacent point clouds in the elevation coordinate sequence, and use it as the elevation jump amplitude value.

[0032] S13. Count the number of point clouds in each scan line within a unit length, calculate the point cloud density of each scan line per unit length, and construct the spatial distribution characteristics by combining the point cloud density of each scan line per unit length with the elevation jump amplitude between all adjacent point clouds.

[0033] It should be noted that the unit length is set according to the ranging accuracy of the laser scanner, for example, a unit length of 1m, and the number of point clouds is counted segment by segment along the scanning line direction.

[0034] S14. Based on spatial distribution characteristics, determine candidate road surface point cloud regions, perform least squares plane fitting on the original point cloud data within the candidate road surface point cloud regions to generate a fitting plane, calculate the distance variance between the original point cloud data and the fitting plane, and use it as the local terrain roughness.

[0035] S15. If the local terrain roughness is lower than the road surface roughness threshold, the point cloud in the candidate road surface point cloud area is identified as a candidate road surface point cloud, and all candidate road surface point clouds are merged into connected components to obtain the road surface point cloud.

[0036] Preferably, the road surface roughness threshold is set according to the typical roughness of asphalt concrete pavement and cement concrete pavement. For example, for asphalt concrete pavement, the road surface roughness threshold is set to 0.02 m²; for cement concrete pavement, the road surface roughness threshold is set to 0.01 m². When the local terrain roughness is lower than the corresponding road surface roughness threshold, it indicates that the point cloud in that area meets the characteristics of a smooth and flat road surface, and it is retained as a candidate road surface point cloud.

[0037] It should be noted that the steps for determining candidate road surface point cloud regions based on spatial distribution characteristics are as follows: First, based on the elevation jump amplitude values ​​between adjacent point clouds in the spatial distribution characteristics of each scan line point cloud, point clouds whose elevation jump amplitude values ​​are within the preset road surface elevation tolerance range are selected as road surface leveling measurement points.

[0038] Secondly, based on the point cloud density of each unit length scan line in the point cloud of each scan line, the unit length scan lines whose point cloud density is within the range of the road surface reference point cloud density are selected, and the point cloud in the selected unit length scan lines is used as continuous measurement points of the road surface.

[0039] Preferably, in a specific embodiment of the present invention, the preset road surface elevation tolerance range is set according to the allowable deviation of the road surface longitudinal profile elevation in the road design specification, for example, it is set to ±3cm, that is, when the elevation jump amplitude between adjacent point clouds is less than 3cm, the point cloud is determined to meet the road surface flatness characteristics, effectively eliminating non-road surface point clouds such as guardrails and curbs with abrupt elevation changes.

[0040] The reference point cloud density range of the road surface is calibrated based on the echo characteristics of the mobile laser device under different road surface materials. For example, the reference point cloud density range of asphalt concrete road surface is set to 800 to 1200 points per meter, which effectively eliminates interfering point clouds such as sparsely distributed vegetation and densely distributed buildings.

[0041] Finally, the spatial intersection of all road surface leveling points and continuous road surface points is calculated to obtain candidate road surface points. Spatial connectivity is constructed for all candidate road surface points, isolated discrete points are eliminated, and candidate road surface point cloud regions are formed.

[0042] This invention collects raw point cloud data from mobile laser scanning of road scenes, analyzes the spatial distribution characteristics of continuous scan line point clouds, and combines local terrain roughness to identify road surface point clouds. This achieves accurate separation of road surface point clouds from background noise, improves the completeness and accuracy of road surface point cloud extraction, effectively reduces road marking recognition errors, and provides a high-quality data foundation for subsequent road marking recognition.

[0043] Considering that accurate identification of road marking information in road point clouds requires feature comparison, current technologies lack multi-type road marking datasets for point cloud feature comparison. This leads to confusion between dashed and solid road markings, and misjudgment of directional arrows and pedestrian crossings, resulting in low reliability of road marking semantic label generation. Therefore, it is necessary to construct a road marking type dataset through multi-level classification annotation and feature vector extraction of historical samples to achieve accurate feature matching between road point clouds and road marking point clouds.

[0044] Based on this, the specific implementation of constructing a dataset of multiple road marking types and obtaining candidate point clouds of markings in this invention includes: S21, retrieving mobile laser scanning point cloud data from historical road scenes, extracting historical road marking point cloud samples from them, performing primary classification according to the marking function category, and obtaining lane dividing line marking samples, guide arrow marking samples, pedestrian crossing marking samples, and prohibition marking samples.

[0045] S22. For the road marking point cloud samples under each primary category, perform secondary classification according to the geometric shape of the markings to obtain solid line marking samples, dashed line marking samples, arrow marking samples, and text symbol marking samples.

[0046] S23. Mark the bounding boxes of the road marking point cloud samples under each secondary category to determine the spatial boundary contours of the road marking point cloud samples of each category.

[0047] S24. Extract the average relative elevation of the sample point cloud, the baseline value of the uniformity of the sample point cloud distribution, and the baseline direction of the plane normal vector of the sample point cloud from the spatial boundary contour to form the sample feature vector.

[0048] It should be noted that the average relative elevation of the sample point cloud is the average elevation difference of all point clouds within the marking sample relative to the fitting plane; the benchmark value of the uniformity of the sample point cloud distribution is the ratio of the standard deviation to the mean of the point cloud density projected onto the horizontal plane within the marking sample; and the benchmark direction of the normal vector of the sample point cloud plane is the angle between the normal vector of the local fitting plane of the marking sample and the normal vector of the road surface plane.

[0049] S25. Classify and store the sample feature vectors corresponding to various road markings according to the marking type to form a marking dataset for multiple road marking types.

[0050] S26. Extract the local geometric features of all measuring points in the road surface point cloud, including the relative elevation of the point cloud, the uniformity of the point cloud distribution in the neighborhood of the measuring point, and the plane normal vector direction of the local area where the measuring point is located, to form the feature vector of the road surface point cloud.

[0051] S27. Perform feature matching between the feature vector of the road surface point cloud and the feature vector of the corresponding sample of various road markings, and calculate the feature similarity.

[0052] S28. Filter out measurement points with feature similarity higher than the preset matching feature similarity, and integrate the filtered measurement points into a candidate point cloud for the datum line.

[0053] It should be noted that the feature similarity is calculated using cosine similarity, and the preset matching feature similarity is calibrated based on the matching accuracy of various types of caliper samples in the caliper dataset. For example, the matching feature similarity is set to 80%, and the implementer can also adjust the preset matching feature similarity themselves.

[0054] This invention compares and identifies road surface point clouds with a constructed dataset of various road marking types to obtain candidate point clouds for road markings. This effectively distinguishes the point clouds for road markings from the road surface background point clouds, avoiding initial confusion between dashed and solid road markings, and providing a reliable candidate dataset for subsequent extraction of road marking instances and type identification.

[0055] Given that the candidate point cloud for road markings contains multiple independent road marking instances, and existing technologies do not perform clustering and boundary geometric parameter extraction on the candidate point cloud, different road marking instances become stuck together and their boundary contours become blurred, affecting the accuracy of subsequent type identification. Therefore, spatial distance clustering and minimum bounding rectangle fitting are needed to achieve accurate separation of road marking instances and extraction of boundary parameters.

[0056] Based on this, the specific implementation of extracting boundary geometric parameters in this invention includes: S31, based on the spatial position of each candidate measurement point in the candidate point cloud of the grading line, obtaining the spatial distance between each candidate measurement point, filtering adjacent candidate measurement points whose spatial distance is less than a set spatial distance threshold, merging connected components, and forming a point cloud cluster corresponding to each grading line instance.

[0057] Preferably, the set spatial distance threshold is determined based on the smallest geometric unit size of the marking. For example, for a standard lane dividing line, the length of the dashed line segment is usually 2m and the interval is 4m. The set spatial distance threshold is set to 0.2m to ensure that the point cloud within the same dashed line segment is merged into the same instance, while avoiding the adhesion of adjacent dashed line segments.

[0058] S32. Perform minimum bounding rectangle fitting on the point cloud clusters corresponding to each caliper instance to determine the projection bounding rectangle of the point cloud clusters on the horizontal plane, and extract the edge lines of the projection bounding rectangles.

[0059] S33. Based on the distances from all candidate measurement points in the point cloud cluster to each edge line in each calibration instance, determine the effective contour boundary points of the point cloud cluster, connect all effective contour boundary points in the point cloud cluster, and generate the contour boundary of the point cloud cluster.

[0060] S34. Based on the contour boundary of the point cloud cluster, obtain the contour length, contour width, and aspect ratio of the contour boundary, and use them as the boundary geometric parameters of the point cloud cluster.

[0061] Preferably, in a specific embodiment of the present invention, the method for determining the effective contour boundary points of the point cloud cluster is as follows: calculate the vertical distance from each candidate measurement point in the point cloud cluster to each side of the projection circumscribed rectangle, and select measurement points whose vertical distance is less than the contour boundary distance threshold and are located on the same side of the outermost layer as effective contour boundary points; wherein the contour boundary distance threshold is set according to the allowable deviation of the calibrated line width, for example, it is set to 20% of the calibrated line width.

[0062] This invention achieves accurate separation and geometric morphology quantification of different line marking instances by clustering and grouping candidate point clouds, obtaining the point cloud clusters corresponding to each line marking instance and extracting boundary geometric parameters, thus avoiding type identification errors caused by line marking instances sticking together.

[0063] Considering that accurate identification of road marking types requires comparison based on standardized geometric features, while existing technologies rely only on simple shape contour comparison, resulting in misjudgment of directional arrows and pedestrian crossings, and low recognition rate of text symbol-type road markings, the reliability of road marking semantic label generation is insufficient. Therefore, it is necessary to achieve accurate identification of road marking types and generation of semantic labels by comparing the similarity of boundary geometric parameters and determining thresholds.

[0064] Based on this, such as Figure 3As shown, the specific implementation of generating semantic tags for each road marking instance in this invention includes: S41, retrieving the edge contours of each standard road marking from the road marking edge template library, and obtaining the boundary geometric parameters of the edge contours of each standard road marking.

[0065] It should be noted that the road marking edge template library pre-stores the standard edge contours and corresponding boundary geometric parameter reference values ​​of various types of road markings as specified by national standards, including the standard contour length, standard contour width, and standard aspect ratio of road markings such as lane dividing lines, guide arrows, pedestrian crossings, and prohibition markings.

[0066] S42. Compare the similarity between the boundary geometric parameters of the point cloud clusters corresponding to each road marking instance and the boundary geometric parameters of the edge contours of each standard road marking, and select the standard road marking with the highest similarity in boundary geometric parameters as the candidate road marking type.

[0067] S43. If the geometric similarity of the highest boundary is higher than the set similarity threshold for marking, the candidate marking type is determined as the marking type of the marking instance, and a semantic label for the corresponding marking instance is generated. Otherwise, the marking type of the marking instance is recorded as an unknown type marking, and the manual review process is triggered.

[0068] Preferably, the similarity comparison is calculated using the cosine similarity of the angle between the normalized boundary geometric parameter vectors. The set similarity threshold for marking determination is set to 0.85, but the implementer can also adjust the marking determination similarity themselves; for markings of unknown type, a manual review process is automatically triggered, whereby the type determination and label supplementation are completed manually to avoid omissions or misidentifications.

[0069] This invention compares and identifies road surface point clouds with a constructed dataset of various road marking types to obtain candidate point clouds for road markings. It then extracts each marking instance from the candidate point clouds and generates semantic labels for each marking instance based on a marking edge template library. This achieves accurate discrimination and semantic annotation of marking types, avoids confusion of marking types, improves the reliability of marking semantic label generation, and meets the requirements of lane-level map construction for marking semantic accuracy.

[0070] Considering that the ultimate goal of road marking simulation design is to ensure that the markings comply with traffic regulations, existing technologies, after completing the road marking simulation design, do not spatially correlate the markings with lane map data. This results in the inability to automatically identify violations in marking placement, leading to safety hazards such as markings deviating from lane boundaries or crossing lane centers. Therefore, it is necessary to achieve precise positioning of the lane to which the markings belong and to verify compliance through spatial correlation between the markings and lane contour data.

[0071] Based on this, the specific implementation of determining the associated lanes of each lane marking instance in this invention includes: S51, obtaining lane map data that matches the road scene, and extracting the contour data of all lanes in the lane map data, the contour data including lane boundary lines and lane center lines.

[0072] It should be noted that the lane map data is a pre-constructed high-precision lane-level map data, which includes the spatial coordinate sequence of the left boundary line, right boundary line and lane center line of each lane. The coordinate system is unified with the original point cloud data of the mobile laser scanning to the same geospatial coordinate system.

[0073] S52. Based on the outline boundary of the point cloud cluster corresponding to each lane marking instance, determine the geometric center position of the outline boundary, transform the geometric center position to the spatial coordinate system where the lane map is located, and determine the geometric center coordinates of each lane marking instance.

[0074] S53. Calculate the shortest distance between the geometric center coordinates of each lane marking instance and all lane boundary lines and the shortest distance between all lane center lines, and determine the associated lanes of each lane marking instance based on the shortest distance.

[0075] It should be noted that the method of determining the associated lanes of each lane marking instance based on the shortest distance is as follows: the geometric center coordinates of each lane marking instance are superimposed with the shortest distance of all lane boundary lines and the shortest distance of all lane center lines to obtain the sum of the shortest distances. The lane corresponding to the minimum value of the sum of the shortest distances of each lane marking instance is selected as the associated lane of each lane marking instance.

[0076] Preferably, if the geometric center coordinates of a lane marking instance are located between the left and right boundary lines of a lane, and the shortest distance to the lane center line of that lane is less than the shortest distance to the lane center line of the adjacent lane, then the associated lane of that lane is determined first to ensure the spatial consistency of lane association.

[0077] This invention spatially correlates the semantic tags of each lane marking instance with lane outline data in the lane map, calculates the shortest distance between each lane marking instance and all lanes, and determines the associated lanes, thereby accurately locating the lane to which the lane marking belongs. This provides a reliable spatial benchmark for compliance verification, effectively solves the problem of lane marking and lane matching deviation, and improves lane matching accuracy.

[0078] In a specific embodiment of the present invention, the steps of verifying the compliance of the layout of each marking instance and outputting the simulation design and verification results of road markings are as follows: S54. Based on the road design specification knowledge base, extract the standard layout rules specified for different marking types, wherein the standard layout rules include the compliant distance range between the lane, the marking and the lane boundary line and the center line.

[0079] It should be noted that the road design specification knowledge base is constructed based on the national road traffic sign and marking setting specifications, and stores the legal placement location of various marking types, the minimum safe distance from the lane boundary line, and the allowable offset distance from the lane center line.

[0080] S55. Based on the marking type in the semantic tags of each marking instance, retrieve the corresponding standard layout rules, and combine the shortest distance between each marking instance and the associated lane boundary line and center line.

[0081] S56. If the associated lane is compliant, and the shortest distance between the lane marking instance and the boundary line and center line of the associated lane are all within the corresponding compliant distance range, then the lane marking instance is deemed to be compliant; otherwise, the lane marking instance is deemed to be non-compliant.

[0082] S57. Summarize the compliance verification results of all road marking instances, synchronize the corresponding semantic tags and associated lanes, and output the road marking simulation design and verification results.

[0083] This invention verifies the compliance of the layout of each road marking instance based on the standard layout rules for different types of road markings in the road design specification knowledge base, identifies instances of illegal layout of road markings, and outputs road marking simulation design and verification results to ensure that the road marking simulation design results conform to actual traffic regulations and improve the standardization level of road marking layout.

[0084] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0085] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0086] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0087] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0088] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A road marking simulation design method based on raw point cloud data, characterized in that, include: The system collects raw point cloud data from mobile laser scanning of road scenes, analyzes the spatial distribution characteristics of continuous scan line point clouds, including the elevation jump amplitude between adjacent point clouds and the point cloud density per unit length of scan line; and selects road surface smoothness measurement points based on elevation jump amplitude and continuous road surface measurement points based on point cloud density to determine candidate road surface point cloud regions. The local terrain roughness is calculated by plane fitting in the region, and candidate road surface point clouds are selected and merged to obtain the road surface point cloud. Based on historical road marking point cloud samples, after primary classification by functional category and secondary classification by geometric shape, the average relative elevation, distribution uniformity, and plane normal vector direction of the sample point clouds are extracted to form sample feature vectors, and a road marking dataset is constructed. The road surface point cloud is compared and identified with the road marking dataset to obtain candidate point clouds for road markings. Cluster the candidate point clouds of the marking lines, obtain the point cloud clusters corresponding to each marking line instance, and extract the boundary geometric parameters of the point cloud clusters; Based on the boundary geometric parameters of the point cloud clusters corresponding to each marking instance, and combined with the marking edge template library, the marking type of each marking instance is identified, and semantic tags for each marking instance are generated. The semantic labels of each road marking instance are spatially associated with the lane outline data in the lane map. The shortest distance between each road marking instance and all lanes is calculated, the associated lanes of each road marking instance are determined, the compliance of the layout of each road marking instance is verified, and the simulation design and verification results of road markings are output.

2. The road marking simulation design method based on raw point cloud data according to claim 1, characterized in that, The analytical method for the spatial distribution characteristics is as follows: Scan line parsing is performed on the raw point cloud data from mobile laser scanning to obtain a set of continuous scan line point clouds arranged in the scanning sequence; Obtain the elevation coordinate sequence of the point cloud of each scan line, calculate the elevation difference between adjacent point clouds in the elevation coordinate sequence, and use it as the elevation jump amplitude value. The number of point clouds per unit length for each scan line is counted, the point cloud density of each unit length scan line is calculated, and the point cloud density of each unit length scan line is combined with the elevation jump amplitude between all adjacent point clouds to form a spatial distribution feature.

3. The road marking simulation design method based on raw point cloud data according to claim 2, characterized in that, The method for identifying the road surface point cloud is as follows: Candidate road surface point cloud regions are determined based on spatial distribution characteristics. The original point cloud data within the candidate road surface point cloud regions are fitted with a plane to generate a fitting plane. The distance variance between the original point cloud data and the fitting plane is calculated and used as the local terrain roughness. If the local terrain roughness is lower than the road surface roughness threshold, the point cloud in the candidate road surface point cloud area is identified as a candidate road surface point cloud, and all candidate road surface point clouds are merged into a connected component to obtain the road surface point cloud.

4. The road marking simulation design method based on raw point cloud data according to claim 3, characterized in that, The steps for determining candidate road surface point cloud regions based on spatial distribution characteristics are as follows: Based on the spatial distribution characteristics of the point cloud of each scan line, the elevation jump amplitude between adjacent point clouds is selected, and the point cloud whose elevation jump amplitude is within the preset road surface elevation tolerance range is used as the road surface flatness measurement point. Based on the point cloud density of each unit length scan line in the point cloud of each scan line, the unit length scan lines whose point cloud density is within the range of the road surface reference point cloud density are selected, and the point cloud in the selected unit length scan lines is used as continuous measurement points of the road surface. All road surface leveling measurement points and continuous road surface measurement points are subjected to spatial location intersection calculation to obtain road surface candidate measurement points. Spatial connectivity domains are constructed for all road surface candidate measurement points to form candidate road surface point cloud regions.

5. The road marking simulation design method based on raw point cloud data according to claim 1, characterized in that, The method for constructing the road marking dataset for various road marking types is as follows: Retrieve mobile laser scanning point cloud data from historical road scenes, extract historical road marking point cloud samples from them, and classify them into primary categories according to the functional categories of the markings to obtain lane dividing line marking samples, guide arrow marking samples, pedestrian crossing marking samples, and prohibition marking samples. For each primary category of road marking point cloud samples, secondary classification is performed according to the geometric shape of the markings to obtain solid line marking samples, dashed line marking samples, arrow marking samples, and text symbol marking samples. Boundary boxes were annotated for road marking point cloud samples under each secondary category to determine the spatial boundary contours of various types of road marking point cloud samples. Extract the average relative elevation of the sample point cloud, the baseline value of the uniformity of the sample point cloud distribution, and the baseline direction of the plane normal vector of the sample point cloud from the spatial boundary contour to form the sample feature vector. The sample feature vectors corresponding to various road markings are categorized and stored according to the marking type, forming a marking dataset for multiple road marking types.

6. The road marking simulation design method based on raw point cloud data according to claim 5, characterized in that, The method for obtaining the candidate point cloud of the gradation line is as follows: Local geometric features of all measuring points in the road surface point cloud are extracted, including the relative elevation of the point cloud, the uniformity of the point cloud distribution in the neighborhood of the measuring point, and the plane normal vector direction of the local area where the measuring point is located, to form the feature vector of the road surface point cloud. The feature vectors of the road surface point cloud are matched with the feature vectors of corresponding samples of various road markings, and the feature similarity is calculated. Filter out measurement points whose feature similarity is higher than the preset matching feature similarity, and integrate the filtered measurement points into a candidate point cloud for the caliper.

7. The road marking simulation design method based on raw point cloud data according to claim 1, characterized in that, The method for extracting the boundary geometric parameters of the point cloud cluster is as follows: Based on the spatial location of each candidate measurement point in the candidate point cloud of the paving line, the spatial distance between each candidate measurement point is obtained, and adjacent candidate measurement points with a spatial distance less than a set spatial distance threshold are filtered and connected components are merged to form the point cloud cluster corresponding to each paving line instance. For each gradation instance, the minimum bounding rectangle is fitted to the point cloud cluster to determine the projection bounding rectangle of the point cloud cluster on the horizontal plane, and the edge of the projection bounding rectangle is extracted. Based on the distances from all candidate measurement points in the point cloud cluster to each edge line in each calibration instance, the effective contour boundary points of the point cloud cluster are determined, and all effective contour boundary points in the point cloud cluster are connected to generate the contour boundary of the point cloud cluster. Based on the contour boundary of the point cloud cluster, the contour length, contour width, and aspect ratio of the contour boundary are obtained and used as the boundary geometric parameters of the point cloud cluster.

8. The road marking simulation design method based on raw point cloud data according to claim 7, characterized in that, The method for generating semantic tags for each punctuation instance is as follows: Retrieve the edge contours of each standard road marking from the road marking edge template library, and obtain the boundary geometric parameters of each standard road marking edge contour; The similarity between the boundary geometric parameters of the point cloud clusters corresponding to each road marking instance and the boundary geometric parameters of the edge contours of each standard road marking is compared, and the standard road marking with the highest similarity in boundary geometric parameters is selected as the candidate road marking type. If the geometric parameter similarity of the highest boundary is higher than the set similarity threshold for marking, then the candidate marking type is determined as the marking type of the marking instance, and a semantic label for the corresponding marking instance is generated.

9. The road marking simulation design method based on raw point cloud data according to claim 8, characterized in that, The method for determining the associated lanes of each lane marking instance is as follows: Obtain lane map data that matches the road scene, and extract the contour data of all lanes in the lane map data, including lane boundary lines and lane center lines. Based on the outline boundary of the point cloud cluster corresponding to each lane marking instance, the geometric center position of the outline boundary is determined, and the geometric center position is transformed into the spatial coordinate system where the lane map is located to determine the geometric center coordinates of each lane marking instance. Calculate the shortest distance between the geometric center coordinates of each lane marking instance and all lane boundary lines and the shortest distance between all lane center lines, and determine the associated lanes of each lane marking instance based on the shortest distance.

10. The road marking simulation design method based on raw point cloud data according to claim 9, characterized in that, The steps for verifying the compliance of the layout of each road marking instance and outputting the road marking simulation design and verification results are as follows: Based on the road design specification knowledge base, the standard layout rules specified for different types of road markings are extracted. The standard layout rules include the compliant distance range between the lane, the road marking and the lane boundary line and the center line. Based on the marking type in the semantic tags of each marking instance, the corresponding standard layout rules are retrieved, and the shortest distance between each marking instance and the associated lane boundary line and center line is combined. If the associated lane is compliant, and the shortest distance between the lane marking instance and the boundary line and center line of the associated lane are all within the corresponding compliant distance range, then the lane marking instance is deemed to be compliant; otherwise, the lane marking instance is deemed to be non-compliant. Summarize the compliance verification results of all road marking instances, synchronize the corresponding semantic tags and associated lanes, and output the road marking simulation design and verification results.

Images