Road marking extraction method and system for vehicle-mounted laser point cloud with insufficient reflection intensity

Abstract

The application discloses a road marking line extraction method and system for a vehicle-mounted laser point cloud with insufficient reflection intensity, and the method is based on a vehicle-mounted laser point cloud, uses a Kd-tree data structure to select neighborhood point clouds of any road surface point cloud, performs K-means++ clustering on the reflection intensity of the point cloud, calculates the average distance of the neighborhood points according to the candidate point cloud set of the marking line obtained through clustering, removes noise points by adopting a three-standard-deviation judgment standard, and realizes accurate extraction of the road marking line. The application reduces the influence caused by long-distance collection of a laser radar and wear of the marking line, solves the problem of complete extraction of the marking line from the vehicle-mounted laser radar point cloud with low reflection intensity, and has the advantages of good adaptability and high accuracy.

Description

Technical Field

[0001] This invention belongs to the field of vehicle-mounted laser point cloud data processing, and specifically relates to a method and system for extracting road markings with insufficient reflection intensity from vehicle-mounted laser point clouds. Background Technology

[0002] Road markings are symbols painted on the road surface, consisting of lines, arrows, and text, used to convey traffic information such as guidance, warnings, and restrictions to pedestrians and drivers, and are used to regulate and guide traffic. Unlike traditional image acquisition, vehicle-mounted LiDAR systems can scan the surrounding environment while driving, acquiring high-precision 3D laser point clouds of the road. Road marking points are contained within road surface points and are extracted primarily based on the difference in reflection intensity values ​​between road markings and other surrounding road surface points. Current technology generally extracts markings by determining a threshold for point cloud reflection intensity. However, under the same acquisition equipment parameters, the reflection intensity values ​​of the point cloud are affected by the distance between the laser emitter and the target point. If the vehicle-mounted LiDAR system is traveling in the rightmost lane, the laser emitter is closer to the right lane, resulting in higher reflection intensity values ​​for each point cloud point in the right lane compared to the left, and a more significant difference between the markings and surrounding road surface points. Conversely, the markings in the leftmost lane, being farther from the laser emitter, have relatively lower intensity values, showing little difference compared to, or even lower than, the non-marked road surface in the rightmost lane, which is closer to the laser emitter. Therefore, marking points cannot be extracted by directly setting a threshold. On the other hand, as time goes by, the reflectivity of road markings will gradually decrease, and the intensity difference between the marking points and the surrounding non-marking points will gradually decrease, making it impossible to manually select a suitable, single threshold. Summary of the Invention

[0003] The technical problem to be solved by this invention is to propose a method and system for extracting road markings with insufficient reflection intensity from vehicle-mounted laser point clouds. The method uses a Kd-tree data structure to select the neighboring point clouds of any road surface point cloud, performs K-means++ clustering on the reflection intensity of the point clouds to obtain a set of candidate point clouds for road markings, calculates the average distance of the neighboring points, and removes noise points to achieve accurate extraction of road markings. This solves the problem of completely extracting road markings from vehicle-mounted laser radar point clouds with low reflection intensity.

[0004] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0005] A method for extracting road markings with insufficient reflection intensity from vehicle-mounted laser point clouds includes the following steps:

[0006] S1. Construct a data index for the vehicle-mounted laser point cloud based on the distance between the driving trajectory points and the corresponding trajectory points.

[0007] S2. Perform elevation filtering based on the data index in step S1 to obtain near-ground point clouds.

[0008] S3. Use principal component analysis to extract point clouds that represent the plane, and perform DBSCAN clustering based on the point cloud density to obtain the road surface point cloud.

[0009] S4. Perform K-means++ clustering on the reflection intensity of the neighborhood points selected based on the road surface point cloud and its Kd-tree data structure to obtain a cloud set of candidate points for road markings.

[0010] S5. Based on the cluster of candidate points for road markings obtained by clustering, calculate the average distance from any point in the cluster to its neighboring points, use three times the standard deviation as a criterion to remove noise points, and extract the road markings.

[0011] Furthermore, in step S1, the specific steps for constructing the vehicle-mounted laser point cloud data index are as follows:

[0012] S101. For any point i in the point cloud data, use the Kd-tree data structure to search for the nearest driving trajectory point T from the driving trajectory points. i and the corresponding shortest distance d min .

[0013] Where the x-coordinate of point i is greater than that of the driving trajectory point T i x-coordinate, then d min A positive sign indicates that the point is to the right of the driving trajectory; otherwise, d min Taking a negative sign indicates that the point is on the left side of the driving trajectory.

[0014] S102, Represent the vehicle-mounted laser point cloud data index as (T) i ,d min ).

[0015] S103. Repeat steps S101 and S102 until the data index construction of all points in the vehicle-mounted laser point cloud is completed.

[0016] Furthermore, in step S2, the specific content of obtaining the near-ground point cloud is as follows: The elevation of the i-th point in the point cloud data is set to h. i Its nearest driving trajectory point T i The elevation is h t Dynamic elevation threshold h N If h i ≤h t +h N If the point is a near-surface point cloud, then that point is a near-surface point cloud.

[0017] Furthermore, in step S3, the specific steps for obtaining the road surface point cloud are as follows:

[0018] S301. Using the Kd-tree data structure, search for k points in the neighborhood of each point in the near-ground point cloud data. Standardize the data of these k points and perform principal component analysis to obtain eigenvalues ​​λ1, λ2, and λ3. Define the minimum eigenvalue as λ3 and its corresponding eigenvector as σ3, which represents the normal vector of the plane.

[0019] Let the coordinates of the eigenvector σ3 be... Z-axis direction The angle between the eigenvector σ3 and the Z-axis is θ, and the specific calculation formula is as follows:

[0020]

[0021]

[0022] S302, make the included angle θ≤θ z The road surface points are separated to obtain a planar point cloud; where θ z The included angle threshold is typically set between 0 and 0.05.

[0023] S303. Based on the planar point cloud, use a Kd-tree data structure to search for the l nearest neighbor data points and their corresponding nearest neighbor distances. Set the average value of the l nearest neighbor distances to D. i Then the formula for calculating the radius r is:

[0024]

[0025] S304. For each point in the planar point cloud, use a Kd-tree data structure to search for the number of points P in its radius r neighborhood. i Then the formula for calculating the threshold minpt is:

[0026]

[0027] S305. Use the DBSCAN algorithm to cluster the planar point cloud to obtain the road surface point cloud.

[0028] Furthermore, in step S4, the specific steps for obtaining the set of candidate points for the grading line are as follows:

[0029] S401. Use the Kd-tree data structure to select n points in the neighborhood of any point i in the road surface point cloud.

[0030] S402. Using K-means++ clustering, divide point i and its n neighbors into two classes. Compare the strength values ​​of the cluster centers of the two classes, and denote the cluster center of the class with the smaller strength value as I. A The number is N A The cluster center of the class with the larger intensity value is denoted as I.B The number is N B .

[0031] S403. Determine the cluster centers of each point in the road surface point cloud:

[0032] (1) If I B -I A G Then point i will be assigned to the non-standard line point set;

[0033] (2) If I B -I A ≥I G And N A ≤N G Then point i will be assigned to the non-standard line point set;

[0034] (3) If I B -I A ≥I G And N A >N G If point i is in category A, then point i is assigned to the candidate point cloud of the datum line; if point i is in category B, then point i is assigned to the non-datum line point cloud.

[0035] Among them, I G N represents the threshold value for the difference in reflection intensity. G This represents the threshold for the number of clusters.

[0036] S404. Based on the judgment results, a cluster of candidate points for the grading line is generated.

[0037] Furthermore, in step S5, the specific steps for extracting road markings are as follows:

[0038] S501, any point S in the candidate point cloud of the caliper line. i Calculate S for each of the m points in its neighborhood. i The average distance d from these m points i Let D be the set.

[0039] S502. Solve for the mean μ and standard deviation σ of the data in set D.

[0040] S503, based on the average distance d i The numerical value is used to determine the following for the candidate point cloud set of the datum line:

[0041] (1) When point S i Average neighborhood distance d i If a point is a noise point, it is discarded. Noise points satisfy the following formula:

[0042] d i ≥μ+3σ. ​

[0043] (2) When point S i Average neighborhood distance d i If it is a marking point, then extract it. Marking points satisfy the following formula:

[0044] d i <μ+3σ.

[0045] Furthermore, this invention also proposes a road marking extraction system for insufficient reflection intensity of vehicle-mounted laser point clouds, including...

[0046] The vehicle-mounted laser point cloud data indexing module is used to build a data index for the vehicle-mounted laser point cloud based on the distance between the driving trajectory points and the corresponding trajectory points.

[0047] The near-ground point cloud module is used to perform elevation filtering based on the data index in the vehicle-mounted laser point cloud data index module to obtain the near-ground point cloud.

[0048] The road surface point cloud module is used to extract point clouds representing a plane using principal component analysis, and performs DBSCAN clustering based on the point cloud density to obtain the road surface point cloud.

[0049] The lane marking candidate point cloud module is used to perform K-means++ clustering based on the reflection intensity of the neighborhood points selected from the road surface point cloud and its Kd-tree data structure to obtain the lane marking candidate point cloud.

[0050] The road marking extraction module is used to obtain a cluster of candidate points for road markings based on clustering, calculate the average distance from any point in the cluster to its neighboring points, and use a three-standard-deviation judgment criterion to remove noise points and extract the road markings.

[0051] Furthermore, the present invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the road marking extraction method for insufficient reflection intensity of vehicle-mounted laser point cloud described above.

[0052] Furthermore, the present invention also proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the method for extracting road markings with insufficient reflection intensity from vehicle-mounted laser point clouds.

[0053] The present invention adopts the above technical solution, and its significant technical effects compared with the prior art are as follows:

[0054] (1) Based on vehicle-mounted laser point cloud, this invention achieves high-precision and complete extraction of markings with weak reflection intensity, which is conducive to digital management and high-definition map generation.

[0055] (2) The present invention constructs a data index for vehicle-mounted laser point cloud based on the distance between driving trajectory points and corresponding trajectory points, and can determine its relative position on the entire road segment based on the three-dimensional coordinates of the point cloud.

[0056] (3) The present invention avoids the problem of point cloud noise data caused by using scan lines to extract the road surface or extracting a single road surface in conventional methods by first extracting the planar point cloud and then generating the road surface points in a step-by-step extraction method.

[0057] (4) This invention uses the Kd-tree data structure to obtain the road surface point cloud and the surrounding neighborhood points, and compares the reflection intensity of the road surface point cloud and the neighborhood points; the K-means++ clustering algorithm is used on the road surface point cloud to obtain the high reflection intensity point cloud, eliminating the influence of the laser emitter distance and wear degree on the reflection intensity value.

[0058] (5) The present invention eliminates noise points in the candidate point cloud of the grading line according to the three-times standard deviation judgment standard, and obtains the grading line point cloud, which is helpful to solve the problem of misjudgment when there are non-grading line points with abnormal intensity and grading line points in their neighborhood. Attached Figure Description

[0059] Figure 1 This is a flowchart illustrating the overall implementation of the present invention.

[0060] Figure 2 This is a schematic diagram of the data index based on driving trajectory points in this invention.

[0061] Figure 3 This is a schematic diagram illustrating the meaning of matrix eigenvectors in the principal component analysis method of this invention.

[0062] Figure 4 This is the original vehicle-mounted laser point cloud in this embodiment of the invention.

[0063] Figure 5 This is the result of line extraction in an embodiment of the present invention. Detailed Implementation

[0064] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0065] This invention proposes a method for extracting road markings with insufficient reflection intensity from vehicle-mounted laser point clouds, such as... Figure 1 As shown, it includes the following steps:

[0066] S1. Construct a data index for the vehicle-mounted laser point cloud based on the distance between the driving trajectory points and the corresponding trajectory points. The specific steps are as follows:

[0067] S101. For any point i in the point cloud data, its three-dimensional coordinates are (522287.354, 3638005.577, 39.292). Use the Kd-tree data structure to search for the nearest driving trajectory point T from the driving trajectory points. i and the corresponding shortest distance d min , get T i =1998,d min =3.707.

[0068] Since the x-coordinate of point i is greater than the driving trajectory point T i x-coordinate, then d min A positive sign indicates that the point is on the right side of the driving trajectory.

[0069] S102, Represent the vehicle-mounted laser point cloud data index as (T) i ,d min ), that is (1998, 3.707).

[0070] S103. Repeat steps S101 and S102 until the data index construction of all points in the vehicle-mounted laser point cloud is completed. (Specific details are as follows...) Figure 2 As shown.

[0071] S2. Perform elevation filtering based on the data index in step S1. The elevation of the i-th point in the point cloud data is h. i =39.292, its nearest driving trajectory point T i The elevation is h i =39.292, dynamic elevation threshold h N It is 0.5m, because h i ≤h t +h N If the point is a near-surface point cloud, then that point is a near-surface point cloud.

[0072] S3. The specific steps for extracting the point cloud representing the plane using principal component analysis and generating the road surface point cloud by DBSCAN clustering based on the point cloud density are as follows:

[0073] S301. Using a Kd-tree data structure, search for 50 points in the neighborhood of each point in the near-ground point cloud data. Standardize the data of these 50 points and perform principal component analysis. Taking a point near the ground as an example, its three-dimensional coordinates are (522287.354, 3638005.577, 39.292). Obtain the eigenvalues ​​λ1 = 1.535, λ2 = 0.0038, and λ3 = 0.0029, and arrange them from largest to smallest: λ1 ≥ λ2 ≥ λ3, reflecting the degree of dispersion of the point cloud in the projection direction. The eigenvectors σ1(-0.3596, -0.9328, -0.0249), σ2(-0.9328, 0.3601, -0.0176), and σ3(-0.0254, -0.0169, 0.995) reflect the normal vectors of the point cloud. The plane formed by σ1 and σ2 can be considered as the plane fitted by these points, and σ3 can be considered as the normal vector corresponding to this plane. Figure 3 As shown.

[0074] The coordinates of the eigenvector σ3 are (-0.0254, -0.0169, 0.995), with the Z-axis direction... The angle between the eigenvector σ3 and the Z-axis is θ, and the specific calculation formula is as follows:

[0075]

[0076]

[0077] The calculated value of θ (in radians) is 0.0306.

[0078] S302, make the included angle θ≤θ z The road surface points are separated to generate a planar point cloud; where θ z The included angle threshold is set to 0.1. Based on this, the point belongs to the planar point cloud.

[0079] S303. Based on the generated planar point cloud, use a Kd-tree data structure to search for the 20 nearest neighbor data points and their corresponding nearest neighbor distances. Set the average value of the 20 nearest neighbor distances to D. i Then the formula for calculating the radius r is:

[0080]

[0081] Since Di is 0.082, traversing all generated planar point clouds yields r of 0.107.

[0082] S304. For each point in the planar point cloud, use a Kd-tree data structure to search for the number of points P in its radius r neighborhood. i P iIf the threshold value is 10, then the formula for calculating the threshold minpt is:

[0083]

[0084] We get minpt = 12.

[0085] S305. Use the DBSCAN algorithm to cluster planar point clouds to generate road surface point clouds.

[0086] S4. Based on the reflection intensity of the neighborhood points selected from the road surface point cloud and its Kd-tree data structure, perform K-means++ clustering to obtain the candidate point cloud for the road marking. The specific steps are as follows:

[0087] S401. Using the Kd-tree data structure, select 100 points in the neighborhood of any point i in the road surface point cloud. The three-dimensional coordinates of point i are (522265.451, 3638013.86, 39.146), and the reflection intensity is 193.

[0088] S402. Using K-means++ clustering, divide point i and its n neighbors into two classes. Compare the strength values ​​of the cluster centers of the two classes, and denote the cluster center of the class with the smaller strength value as I. A =190, the number is N A =89; the cluster center of the class with the larger intensity value is denoted as I. B =238, the number is N B =11.

[0089] S403. Determine the cluster centers of each point in the road surface point cloud:

[0090] (1) If I B -I A G Then point i will be assigned to the non-standard line point set;

[0091] (2) If I B -I A ≥I G And N A ≤N G Then point i will be assigned to the non-standard line point set;

[0092] (3) If I B -I A ≥I G And N A >N G If point i is in category A, then point i is assigned to the candidate point cloud of the datum line; if point i is in category B, then point i is assigned to the non-datum line point cloud.

[0093] Among them, I G ​The threshold representing the difference in reflection intensity is 20; N G This represents the threshold for the number of clusters, which is 3.

[0094] Based on the judgment criteria, this point should belong to a cluster of non-marked line points.

[0095] S404. Based on the judgment results, a cluster of candidate points for the grading line is generated.

[0096] S5. Based on the clustered candidate point set of road markings, calculate the average distance from any point in the set to its neighboring points, and use a three-standard-deviation criterion to remove noise points to extract the road markings. The specific steps are as follows:

[0097] S501, any point S in the candidate point set S of the datum line. i Calculate S for each of the 10 points in its neighborhood. i The average distance d from these 10 points i Let S be the set D. i The three-dimensional coordinates are (522197.333, 3638047.560, 37.164), and the reflection intensity is 235.

[0098] S502. Solve for the mean μ and standard deviation σ of the data in set D, which are 0.375 and 0.181, respectively.

[0099] S503, based on the average distance d i The numerical value is used to determine the following for the candidate point cloud set of the datum line:

[0100] (1) When any point S of the candidate points of the gradation line is clustered together i Average neighborhood distance d i If a point is a noise point, it is discarded. Noise points satisfy the following formula:

[0101] d i ≥μ+3σ.

[0102] (2) When any point S of the candidate points of the gradation line is clustered together i Average neighborhood distance d i If it is a marking point, then extract it. Marking points satisfy the following formula:

[0103] d i <μ+3σ.

[0104] To determine this point, d i =0.288, d i Points less than μ+3σ are not considered noise points. Based on the judgment result, the grading point cloud is extracted.

[0105] The experimental data used in this embodiment is road point cloud data from a highway in Nanjing, such as... Figure 4 As shown, the road markings are faint and blurry. After extraction using the method provided by this invention, the road marking recognition results are clearly visible, as detailed below. Figure 5 As shown.

[0106] By comparing with manually marked road markings, the accuracy was evaluated using recall, precision, and overall accuracy, as shown in Table 1. It can be concluded that the extraction accuracy of all markings is above 90%, demonstrating good application performance.

[0107] Table 1 Accuracy of Road Marking Extraction

[0108]

[0109] This invention also proposes a road marking extraction system for insufficient reflection intensity of vehicle-mounted laser point clouds, including a vehicle-mounted laser point cloud data indexing module, a near-ground point cloud module, a road surface point cloud module, a road marking candidate point cloud aggregation module, a road marking extraction module, and a computer program that can run on a processor. It should be noted that each module in the above system corresponds to a specific step of the method provided in this invention embodiment, possessing the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in this invention embodiment.

[0110] This invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. It should be noted that each module in the above system corresponds to a specific step of the method provided in this invention, possessing the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in this invention.

[0111] This invention also proposes a computer-readable storage medium storing a computer program. It should be noted that each module in the above system corresponds to a specific step of the method provided in this invention, possessing the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in this invention.

[0112] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A method for extracting road marking lines with insufficient reflection intensity of laser point cloud on a vehicle, characterized in that, The method comprises the following steps: S1, constructing a data index of the vehicle-mounted laser point cloud according to the distance between the driving track points and the corresponding track points; specifically, S101、For any point i in the point cloud data, search the nearest driving track point T from the driving track points by using the Kd-tree data structure i and the corresponding shortest distance d min ; Wherein, if the horizontal coordinate of the i point is greater than the horizontal coordinate of the driving track point T i horizontal coordinate, then d min is positive, indicating that the point is on the right side of the driving track; otherwise, d min is negative, indicating that the point is on the left side of the driving track; In S102, the vehicle-mounted laser point cloud data index is expressed as (T i ,d min ). S103, repeating steps S101 and S102 until the data index construction of all points in the vehicle-mounted laser point cloud is completed; S2, performing elevation filtering according to the data index in step S1 to obtain a near-ground point cloud; S3, extracting the point cloud of the presented plane by using the principal component analysis method, and obtaining the road surface point cloud by performing DBSCAN clustering according to the point cloud density; S4, performing K-means++ clustering according to the reflection intensity of the neighbor points selected according to the road surface point cloud and the Kd-tree data structure to obtain a candidate point cloud set of the marking line; specifically, S401, selecting n points in the neighborhood of any point i of the road surface point cloud by using the Kd-tree data structure; S402、Utilize K-means++ clustering, divide the n points in the neighborhood of point i into two categories; compare the intensity values of the two cluster centers, and the cluster center of the category with small intensity value is recorded as I A , The number is N A ; The cluster center of the category with large intensity value is recorded as I B , The number is N B ; S403, judging the cluster center of each point of the road surface point cloud: (1) If I B - I A < I G then point i is assigned to the non-planar point cloud set; (2) If I B - I A ≥ I G and N A ≤ N G then point i is assigned to the non-planar point cloud set; (3) if I B - I A ≥ I G and N A > N G then if point i is in class A, point i is assigned to the ruler candidate point cloud set, and if point i is in class B, point i is assigned to the non-ruler point cloud set. where I G represents a threshold value of the difference of reflection intensity, N G represents a threshold value of the number of clusters; S404, according to the judgment result, the candidate point cloud set of the marking line is generated; S5, according to the candidate point cloud set of the marking line, calculating the average distance from any point in the set to its neighbor points, and removing the noise points by adopting the three times standard deviation judgment standard to extract the road marking line. 2.The method according to claim 1, wherein, In step S2, the specific content of obtaining the near-ground point cloud is: setting the height of the i-th point in the point cloud data as h i , the height of the nearest driving track point T i of the vehicle is h t , the dynamic height threshold is h N , if h i ≤ h t + h N , the point is a near-ground point cloud. 3.The method according to claim 1, wherein, In step S3, the specific steps for obtaining the road surface point cloud are as follows: S301, searching for k points in the neighborhood of each point of the near-ground point cloud data by using the Kd-tree data structure, performing data standardization on the k points, and performing principal component analysis to obtain characteristic values λ1, λ2, λ3, and defining the minimum characteristic value as λ3, and the characteristic vector corresponding to λ3 as σ3, which represents the normal vector corresponding to the plane; The coordinates of the characteristic vector σ3 are set as (Xσ3, Yσ3, Zσ3), and the direction of the Z axis The angle between the characteristic vector σ3 and the Z axis is θ, and the specific calculation formula is: S302, separate the road points with the included angle θ ≤ θ z to obtain a planar point cloud; wherein, θ z is an included angle threshold; S303、According to the plane point cloud, search out l nearest neighbor data points and corresponding nearest neighbor distances through the Kd-tree data structure, set the average value of the l nearest neighbor distances as D i The calculation formula of the radius r is: S304、For each point in the planar point cloud, search the number P of points in its radius r neighborhood through the Kd-tree data structure i The calculation formula of the threshold minpt is: S305, performing clustering on the plane point cloud by using the DBSCAN algorithm to obtain the road surface point cloud. 4.The method of claim 1, wherein, In step S5, the specific steps for extracting the road marking line are as follows: S501、For any point S in the candidate point cloud set of the reticle i and m points in its neighborhood, respectively calculate S i The average distance d of S and m points i , denoted as set D; S502, solving the average value μ and the standard deviation σ of the data in the set D; S503、According to the numerical size of the average distance d i The following judgment is made on the candidate point cloud set in the reticle according to the numerical size of the average distance d (1) When the average neighborhood point distance d i of point S i is a noise point, then it is eliminated; the noise point satisfies the following formula: d i ≥ μ + 3σ; (2) When the point S i is an average neighborhood point distance d i of the mark point, then extraction; the mark point satisfies the following formula: d i <μ+3σ.

5. The system for applying the method for extracting road marking lines with insufficient reflection intensity from a vehicle-mounted laser point cloud according to claim 1, characterized in that, The method comprises the following steps: The vehicle-mounted laser point cloud data index module is configured to construct a data index of the vehicle-mounted laser point cloud according to the distance between the driving track points and the corresponding track points; The near-ground point cloud module is configured to perform elevation filtering according to the data index in the vehicle-mounted laser point cloud data index module to obtain a near-ground point cloud; The road surface point cloud module is configured to extract the point cloud of the presented plane by using the principal component analysis method, and obtain the road surface point cloud by performing DBSCAN clustering according to the point cloud density; The candidate point cloud set of the marking line module is configured to perform K-means++ clustering according to the reflection intensity of the neighbor points selected according to the road surface point cloud and the Kd-tree data structure to obtain a candidate point cloud set of the marking line; The road marking line extraction module is configured to calculate the average distance from any point in the set to its neighbor points according to the candidate point cloud set of the marking line, remove the noise points by adopting the three times standard deviation judgment standard, and extract the road marking line.

6. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the steps of the method in any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, the computer-readable storage medium being characterized by, The computer program is run by the processor to execute the method in any one of claims 1 to 4.

Images