A crowdsourcing map evaluation method and device based on lane lines

Abstract

The present invention belongs to the technical field of high-precision map quality evaluation for autonomous driving. The present invention provides a crowdsourced map evaluation method and device based on lane lines, wherein the method includes: obtaining a sequence of lane line points based on crowdsourced data, using a custom distance calculation method between lane lines, hierarchically clustering the lane lines collected multiple times to obtain multiple clusters; for each cluster, performing polynomial fitting on the lane lines within the cluster, calculating the distance from each lane line to the fitted lane line, and mapping the distance to between 0 and 1 using a custom function to obtain the cluster-internal confidence; and calculating the global confidence of the map based on the confidence within each cluster and the total length of the lane lines. In the absence of a true map value, the present invention performs relevant calculations and processing on the lane lines to obtain the local and global confidence of the crowdsourced map, thereby respectively evaluating the local and global quality of the crowdsourced map.

Description

Technical Field

[0001] The present invention belongs to the technical field of high-precision map quality evaluation for autonomous driving, and in particular relates to a crowdsourcing map evaluation method and device based on lane lines. Background Art

[0002] Crowdsourced maps are generated by traffic information uploaded and shared during driving. By efficiently uploading and sharing traffic information (traffic jams, event information) in real time, timely and accurate map traffic information is formed and then pushed to all other users with application needs.

[0003] In the field of autonomous driving, low-cost crowdsourcing collection vehicles are widely deployed to collect high-fresh data at a high frequency. The equipment cost is low, the collection accuracy is poor, and the amount of data collected is huge. When mapping, due to uncertain factors such as collected data and algorithms, the quality and availability of the map need to be evaluated. The evaluation of crowdsourcing maps can also serve as important reference data for downstream autonomous driving positioning and navigation.

[0004] Currently, most methods in the industry use confidence levels to evaluate crowdsourced maps by comparing them with the ground truth. Higher confidence levels indicate higher quality, while lower confidence levels indicate lower quality. However, in reality, the process of obtaining ground truth maps is complex, and ground truth maps are often unavailable immediately, making crowdsourced map evaluation more difficult.

[0005] In view of this, overcoming the defects of the prior art is an urgent problem to be solved in this technical field. Summary of the Invention

[0006] The present invention provides a solution to the technical problem that existing methods cannot effectively evaluate the mapping quality of crowdsourced maps without map truth.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0008] In a first aspect, the present invention provides a crowdsourced map evaluation method based on lane lines, comprising:

[0009] Based on crowdsourced data, a lane line point sequence is obtained. A custom distance calculation method is used to calculate the distance between lane lines. Hierarchical clustering is performed on the lane lines collected multiple times to obtain multiple clusters.

[0010] For each cluster, a polynomial fit is performed on the lane lines within the cluster, and the distance from each lane line to the fitted lane line is calculated. The distance is then mapped to a range between 0 and 1 using a custom function to obtain the cluster's internal confidence.

[0011] The global confidence of the map is calculated by combining the confidence within each cluster and the total length of the lane lines.

[0012] Preferably, the method for calculating the distance between lane lines includes:

[0013] For any two lane lines L1 and L2, fit lane line L1 and lane line L2 to obtain lane line L3;

[0014] Based on the shape point a on the lane line L1, find the shape point b on the lane line L3 so that the distance between the shape point a and the shape point b is the shortest.

[0015] Preferably, the distance between the shape point a and the shape point b is calculated as the average value of the projection distance from the shape point a to the shape point b and the projection distance from the shape point b to the shape point a.

[0016] Preferably, the lane lines collected multiple times are hierarchically clustered to obtain multiple clusters, including:

[0017] Collect each lane line multiple times to obtain clusters of lane lines, and initialize the lane lines in each cluster into a class;

[0018] According to the distance calculation method between lane lines, the distance matrix between each cluster is calculated to obtain the two clusters with the closest distance;

[0019] A distance threshold between lane lines is preset. When the inter-cluster distance between the two closest clusters is less than the distance threshold, the two classes corresponding to the two clusters with the closest inter-cluster distance are aggregated one by one.

[0020] Preferably, the calculating of the confidence within each cluster includes:

[0021] Preset the ratio threshold of the number of lane lines in each cluster to the number of aggregation times;

[0022] If the ratio of the number of lane lines in each cluster to the number of aggregations is less than the preset threshold of the ratio of the number of lane lines in each cluster to the number of aggregations, the cluster is regarded as noise data with a confidence level of 0.

[0023] Preferably, the preset threshold value of the ratio of the number of lane lines in each cluster to the number of aggregation times is 0.2.

[0024] Preferably, the polynomial fitting method for the lane lines within the cluster is a cubic polynomial.

[0025] Preferably, the lane lines obtained by crowdsourcing data collection are obtained through SLAM optimization.

[0026] Preferably, the interpolation value between the shape point sequences is 0.5 meters.

[0027] In a second aspect, the present invention provides a crowdsourced map evaluation device based on lane lines, comprising:

[0028] at least one processor; and,

[0029] A memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the processor to perform the lane line-based crowdsourcing map evaluation method as described in the first aspect.

[0030] In view of the deficiencies in the prior art, the present invention can achieve the following beneficial effects:

[0031] In the absence of true map values, the present invention obtains the local and global confidence of the crowdsourced map by performing relevant calculations and processing on lane lines. Based on the local and global confidence of the crowdsourced map, the local and global quality of the crowdsourced map can be evaluated respectively. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0033] Figure 1 This is a flowchart of a lane-based crowdsourcing map evaluation method provided in Implementation 1;

[0034] Figure 2 This is a calculation process diagram of a lane-based crowdsourcing map evaluation method provided by Implementation 1;

[0035] Figure 3 This is an example diagram of an application of a lane-based crowdsourcing map evaluation method provided in Implementation 1;

[0036] Figure 4 This is a structural diagram of a lane-line-based crowdsourcing map evaluation device provided in Implementation 2. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, the technical features in the various embodiments or single embodiments provided by the present invention can be arbitrarily combined with each other to form a feasible technical solution. This combination is not restricted by the sequence of steps and / or structural composition mode, but must be based on the ability of ordinary technicians in this field to implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0038] Example 1:

[0039] In order to solve the technical problem that existing methods cannot effectively evaluate the quality of crowdsourced maps without map truth, this embodiment 1 provides a crowdsourced map evaluation method based on lane lines, such as Figure 1 Shown, including:

[0040] S10, based on the crowdsourced data, a lane line point sequence is obtained, and a custom distance calculation method between lane lines is used to perform hierarchical clustering on the lane lines collected multiple times to obtain multiple clusters.

[0041] In practice, the crowdsourced data is used to obtain lane lines, which are then optimized using SLAM (Simultaneous Localization and Mapping). SLAM (Simultaneous Localization and Mapping) is primarily used to solve the problem of positioning and mapping vehicles in unknown environments. In this step, the crowdsourced data is optimized using SLAM to create a certain degree of clustering in the lane lines.

[0042] For each lane line, there is a time sequence label P = {N1, N2, N3...N n} is composed of a sequence of shape points. In actual application, preferably, the interpolation value between the shape point sequences is 0.5 meters, that is, the distance between each shape point of each lane line is 0.5 meters; wherein (x, y, z) represents the coordinates of the shape point, and the vector (i, j, k) represents the direction of the shape point.

[0043] As one implementation, the method for calculating the distance between lane lines includes:

[0044] For any two lane lines L1 and lane line L2, lane lines L1 and lane line L2 are fitted to obtain lane line L3. After obtaining lane line L3, further, based on the shape point a on lane line L1, shape point b on lane line L3 is found so that the distance between shape point a and shape point b is the shortest.

[0045] In a specific implementation, the distance between the shape point a and the shape point b is calculated as: the average value of the projection distance from the shape point a to the shape point b and the projection distance from the shape point b to the shape point a.

[0046] like Figure 2 The figure shows a calculation process diagram of a crowdsourcing map evaluation method based on lane lines.

[0047] For ease of understanding, the calculation process is described with lane line L1 and lane line L2 as straight lines. In actual road conditions, lane line L1 and lane line L2 can also be curved.

[0048] Assume that the coordinates of the shape point a on lane line L1 are p1 = (x1, y1, z1), the direction is v1 = (i1, j1, k1), the coordinates of the shape point b on lane line L3 are p2 = (x2, y2, z2), the direction is v2 = (i2, j2, k2), and the lane line L1 = {N 11 ,N 12 ,N 13 ...N 1m}, lane line L2 = {N 21 ,N 22 ,N 23 ...N 2n}, the lane line L1 and lane line L2 are fitted by a cubic polynomial, and the lane line L3 is obtained as follows: 31 ,N 32 ,N 33 ...N 3k};

[0049] When calculating, the direction from point a to point b is p1p2=p2-p1.

[0050]

[0051] Among them, α and β are the angles with the direction from shape point a to shape point b respectively.

[0052] The distance between the shape point a and the shape point b is calculated by the spatial distance function, and the calculation method is:

[0053] spatial_dist=(|p1p2|*sinα+|p1p2|*sinβ) / 2

[0054] By calculating the angle function, the angular distance between shape point a and shape point b can be obtained as:

[0055]

[0056] In summary, the shortest distance between shape point a and shape point b is:

[0057] dist_ab=spatial_dist+2.5*angular_dist

[0058] Assume that the coordinates of the shape point c on lane line L2 are p3 = (x3, y3, z3) and the direction is v2 = (i3, j3, k3). Similarly, we can calculate and find the shape point on lane line L3 that is closest to the shape point c on lane line L2.

[0059] Finally, the distance between lane line L1 and lane line L2 is:

[0060]

[0061] Among them, d 13 d 23 are the distances between lane line L1, lane line L2 and lane line L3 respectively.

[0062] The above-mentioned method for calculating the distance between lane lines is based on lane lines L1 and L2, and fits lane line L3 for calculating the distance transition between lane lines L1 and L2. As a more convenient alternative, the method for calculating the distance between lane lines can directly project the shape point a directly onto lane line L2, and then directly use the projected point as the shape point closest to the shape point.

[0063] In this step, the lane lines collected multiple times are hierarchically clustered to obtain multiple clusters, including:

[0064] S11: collect data for each lane line multiple times to obtain clusters of lane lines, and initialize the lane lines in each cluster into a class.

[0065] During the application process, the number of times each lane line is collected can be determined based on actual conditions. The more times each lane line is collected, the more lane lines there are in the lane line cluster.

[0066] Assume that the lane line after SLAM optimization is L = {L1, L2, L3...L n}, using agglomerative clustering to classify each lane line, assuming that the lane line in the cluster is initialized to a class C = {C1, C2, C3...C n}.

[0067] S12: Calculate the distance matrix between clusters according to the distance calculation method between lane lines, and obtain the two clusters with the closest distance.

[0068] The method for calculating the distance between lane lines is the same as described in S10.

[0069] For any two classes C a ={L a,1 ,L a,2 ,L a,3 ...L a,m} and C b ={L b,1 ,L b,2 ,L b,3 ...L b,n}, according to the distance calculation method between lane lines, C can be calculated a and C b The distance matrix D m×n , where D represents the lane line L a,i and lane line L b,j The distance, C a and C b The distance between clusters is d_ ab =min(D m×n ).

[0070] S13, a distance threshold between lane lines is preset. When the inter-cluster distance between the two closest clusters is less than the distance threshold, the two classes corresponding to the two clusters with the closest inter-cluster distance are aggregated one by one.

[0071] Set the distance threshold d between the preset lane lines tol In actual application, the distance threshold d between lane lines tol = 0.9 m, when d_ ab ≤d tol When C a and C b Polymerization was performed to obtain C new ={C a ,C b}.

[0072] Repeat the above aggregation process until all the inter-cluster distances d_ ab Both are greater than d tol .

[0073] S20, for each cluster, perform polynomial fitting on the lane lines within the cluster, calculate the distance between each lane line and the fitted lane line, and map the distance to between 0 and 1 through a custom function to obtain the cluster internal confidence.

[0074] The lane lines within the cluster are subjected to polynomial fitting, including but not limited to quadratic polynomial fitting, cubic polynomial fitting, and quartic polynomial fitting. In combination with the actual application process, the fitting method is cubic polynomial fitting. The distance between each lane line and the fitted lane line is calculated as described in S10.

[0075] The in-cluster confidence, that is, the local confidence of the crowdsourced map, represents the local quality of the crowdsourced map to a certain extent and is used to evaluate the local quality of the crowdsourced map.

[0076] S30: Calculate the global confidence of the map using the confidence within each cluster and the total length of the lane lines.

[0077] In this step, the confidence level within each cluster is calculated, including:

[0078] A threshold value for the ratio of the number of lane lines in each cluster to the number of aggregations is preset. In actual application, the threshold value for the ratio of the number of lane lines in each cluster to the number of aggregations is preset to 0.2. If the ratio of the number of lane lines in each cluster to the number of aggregations is less than the threshold value, the cluster is treated as noise data with a confidence level of 0.

[0079] Suppose that after N aggregations, the number of lane lines in the cluster is M, and the lane line L = {L1, L2, L3...L M Specifically, if M / N < 0.2, the cluster is considered as noise data. If M / N ≥ 0.2, the lane lines in the cluster are fitted with a cubic polynomial to obtain the lane line L new It is understandable that when the number of lane lines in a cluster is constant (i.e., M remains unchanged), the greater the number of fitting times (i.e., the larger N is), the lower the calculated confidence within the cluster. In specific implementation, the calculation method for the confidence within the cluster is:

[0080]

[0081] Among them, d i From the i-th lane line in L to L new distance; t is a constant parameter with a reference value of 3.

[0082] like Figure 3 The figure shown is an example of the application of the crowdsourcing map evaluation method based on lane lines.

[0083] Assume that through a series of calculations, the cluster internal confidence of cluster 1, cluster 2 and cluster 3 are 0.97, 0, and 0.6 respectively, that is, the local map quality status corresponding to cluster 1, cluster 2 and cluster 3 are 0.97, 0, and 0.6 respectively.

[0084] In order to evaluate the global quality of the crowdsourced map, after calculating the in-cluster (local) confidence, the local confidence calculation result of the map is set to C = {C1, C2, C3...C M}, the sum of lane lengths in each cluster is l = {l1,l2,l3...l M}, the length of all lane lines can be calculated as Then, the global confidence calculation formula of the crowdsourcing map can be expressed as The global confidence of the crowdsourced map is used to evaluate the global quality of the crowdsourced map.

[0085] This embodiment 1 provides a crowdsourced map evaluation method based on lane lines. Lane lines are classified through distance calculation. The inter-class distance is calculated and mapped to a range from 0 to 1 to represent the local confidence of the lane lines within the cluster. This can be used to evaluate the local quality of the crowdsourced map.

[0086] Furthermore, the global confidence is calculated by the local confidence and the lane line length within the cluster, which can be used to evaluate the global quality of the crowdsourced map.

[0087] Example 2:

[0088] Based on the same general technical solution as Example 1, Figure 4 2 is a schematic structural diagram of a lane-based crowdsourcing map evaluation device provided in Example 2, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the processor to perform the lane-based crowdsourcing map evaluation method as described in Example 1.

[0089] In summary, the present invention provides a crowdsourced map evaluation method and device based on lane lines. In the absence of a true map value, the local and global confidence levels of the crowdsourced map are obtained by performing relevant calculations and processing on the lane lines. Based on the local and global confidence levels of the crowdsourced map, the local and global quality of the crowdsourced map can be evaluated respectively.

[0090] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0091] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, electronic devices, or computer software program products. Thus, the present invention may take the form of a fully hardware embodiment, a fully software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0092] The present invention is described with reference to flowcharts and / or block diagrams of methods, systems, electronic devices, or computer software program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A system that specifies the functions of a box or boxes.

[0093] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction system that is implemented in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0094] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0095] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

Claims

1. A crowdsourcing map evaluation method based on lane lines, characterized by: include: Based on crowdsourced data, a lane line point sequence is obtained. A custom distance calculation method is used to calculate the distance between lane lines. Hierarchical clustering is performed on the lane lines collected multiple times to obtain multiple clusters. For each cluster, a polynomial fit is performed on the lane lines within the cluster, and the distance from each lane line to the fitted lane line is calculated. The distance is then mapped to a range between 0 and 1 using a custom function to obtain the cluster's internal confidence. The global confidence of the map is calculated by combining the confidence within each cluster and the total length of the lane lines.

2. The crowdsourcing map evaluation method based on lane lines according to claim 1 is characterized in that: The method for calculating the distance between lane lines includes: For any two lane lines L1 and L2, fit lane line L1 and lane line L2 to obtain lane line L3; Based on the shape point a on the lane line L1, find the shape point b on the lane line L3 so that the distance between the shape point a and the shape point b is the shortest.

3. The crowdsourcing map evaluation method based on lane lines according to claim 2, characterized in that: The distance between the shape point a and the shape point b is calculated as the average value of the projection distance from the shape point a to the shape point b and the projection distance from the shape point b to the shape point a.

4. The crowdsourcing map evaluation method based on lane lines according to claim 3 is characterized in that: The lane lines collected multiple times are hierarchically clustered to obtain multiple clusters, including: Collect each lane line multiple times to obtain clusters of lane lines, and initialize the lane lines in each cluster into a class; According to the distance calculation method between lane lines, the distance matrix between each cluster is calculated to obtain the two clusters with the closest distance; A distance threshold between lane lines is preset. When the inter-cluster distance between the two closest clusters is less than the distance threshold, the two classes corresponding to the two clusters with the closest inter-cluster distance are aggregated one by one.

5. The crowdsourcing map evaluation method based on lane lines according to claim 4 is characterized in that: The calculation of the confidence within each cluster includes: Preset the ratio threshold of the number of lane lines in each cluster to the number of aggregation times; If the ratio of the number of lane lines in each cluster to the number of aggregations is less than the preset threshold of the ratio of the number of lane lines in each cluster to the number of aggregations, the cluster is regarded as noise data with a confidence level of 0.

6. The crowdsourcing map evaluation method based on lane lines according to claim 5 is characterized in that: The preset threshold value of the ratio of the number of lane lines in each cluster to the number of aggregation times is 0.

2.

7. The crowdsourcing map evaluation method based on lane lines according to claim 1 is characterized in that: The polynomial fitting method for the lane lines within the cluster is a cubic polynomial.

8. The crowdsourcing map evaluation method based on lane lines according to claim 1, characterized in that: The lane lines obtained by crowdsourcing data collection are obtained after SLAM optimization.

9. The crowdsourcing map evaluation method based on lane lines according to claim 1, characterized in that: The interpolation value between the shape point sequences is 0.5 meters.

10. A crowdsourcing map evaluation device based on lane lines, characterized in that: include: at least one processor; as well as, A memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the processor to perform the lane line-based crowdsourcing map evaluation method according to any one of claims 1 to 9.

Images