Lane detection method, device and system

Through the filtering of point cloud data and hash table search, combined with geometric feature screening, accurate detection of lane lines is achieved, and the problem of low detection accuracy in the existing technology is solved.

CN115049826BActive Publication Date: 2024-11-26HENAN YUZHIXING INTELLECTUAL PROPERTY OPERATION CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210542120.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-05-18
Publication Date
2024-11-26
Estimated Expiration
2042-05-18

AI Technical Summary

Technical Problem

Existing lane line detection technology is susceptible to other road surface identification and changes in light intensity, resulting in lower detection accuracy.

Method used

By obtaining point cloud data, the initial ground point cloud is filtered out and divided into ground point sets and lane marking points sets, dynamic distance search is performed based on the hash table, connecting areas are divided, and lane line areas are filtered out according to geometric features to establish lane lines curve.

Benefits of technology

It reduces the interference of other road surface markings on lane line detection, improves the accuracy of lane line detection, and is not affected by changes in environmental parameters such as light intensity.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115049826B_ABST
    Figure CN115049826B_ABST
Patent Text Reader

Abstract

The present invention provides a lane line detection method, device and system, the method comprising: obtaining point cloud data of an area to be detected, filtering out an initial ground point cloud from the point cloud data of the area to be detected; dividing the initial ground point cloud into a ground point set and a lane marking point set according to the coordinates and pulse width of the initial ground point cloud; establishing a hash table based on the lane marking point set, performing a dynamic distance search on the hash table to divide the lane marking point set into multiple connected areas; filtering out a lane line area based on the geometric features of the connected area, filtering out lane line feature points from the lane line area, and establishing a lane line curve based on the lane line feature points. The present invention can reduce the interference of other road surface markings on lane line detection, and will not be affected by changes in environmental parameters such as light intensity, thereby improving the accuracy of lane line detection.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of 3D sensing technology, and in particular to a lane line detection method, device and system. Background Art

[0002] With the development of unmanned driving technology, improving the ability of vehicle environment perception and autonomous decision-making is a key issue. In order to reduce the frequency of vehicle safety accidents in unmanned driving, effectively assisting the planning of vehicle driving routes to constrain the boundaries of the vehicle's drivable area and ensure that the vehicle drives normally within the lane line has become one of the key technologies that need to be solved. Existing lane line detection technology mainly detects lane lines by collecting road images, which is easily affected by changes in environmental parameters such as other road signs or light intensity, resulting in low lane line detection accuracy. Summary of the invention

[0003] In view of this, the purpose of the present invention is to provide a lane line detection method, device and system, which can reduce the interference of other road surface markings on lane line detection and will not be affected by changes in environmental parameters such as light intensity, thereby improving the accuracy of lane line detection.

[0004] In order to achieve the above purpose, the technical solution adopted by the embodiment of the present invention is as follows:

[0005] In a first aspect, an embodiment of the present invention provides a lane line detection method, comprising: acquiring point cloud data of an area to be detected, and filtering out an initial ground point cloud from the point cloud data of the area to be detected; dividing the initial ground point cloud into a ground point set and a lane marking point set according to the coordinates and pulse width of the initial ground point cloud; establishing a hash table based on the lane marking point set, and performing a dynamic distance search on the hash table to divide the lane marking point set into multiple connected areas; filtering out a lane line area based on the geometric features of the connected area, filtering out lane line feature points from the lane line area, and establishing a lane line curve based on the lane line feature points.

[0006] Further, an embodiment of the present invention provides a first possible implementation of the first aspect, wherein the step of dividing the initial ground point cloud into a ground point set and a lane marking point set according to the coordinates and pulse width of the initial ground point cloud includes: obtaining the coordinates and actual pulse width of each point cloud in the initial ground point cloud, and determining the collection distance of each point cloud in the initial ground point cloud based on the coordinates of each point cloud in the initial ground point cloud; obtaining a first pulse width and distance function and a second pulse width and distance function; wherein the first pulse width and distance function is a functional relationship between the pulse width of a ground material point and the collection distance of the ground material point, and the second pulse width and distance function is a functional relationship between the pulse width of a lane line or marking line material point and the collection distance of the lane line or marking line The functional relationship between the collection distance of the marking line material points; determining the first pulse width prediction value of each point cloud based on the collection distance of each point cloud in the initial ground point cloud and the first pulse width and distance function, and determining the second pulse width prediction value of each point cloud based on the collection distance of each point cloud in the initial ground point cloud and the second pulse width and distance function; respectively calculating the first pulse width error between the actual pulse width of each point cloud and the first pulse width prediction value and the second pulse width error between the actual pulse width of each point cloud and the second pulse width prediction value, and judging the material corresponding to each point cloud in the initial ground point cloud based on the first pulse width error and the second pulse width error, so as to divide the initial ground point cloud into a ground point set and a lane marking point set.

[0007] Further, an embodiment of the present invention provides a second possible implementation of the first aspect, wherein the first pulse width and distance function is: PW=a×ln(Dis)+b; the second pulse width and distance function is PW=c×ln(Dis)+d; wherein PW is the pulse width, Dis is the acquisition distance, and a, b, c and d are all fitting coefficients.

[0008] Furthermore, an embodiment of the present invention provides a third possible implementation of the first aspect, wherein the step of establishing a hash table based on the lane marking point set and performing a dynamic distance search on the hash table to divide the lane marking point set into multiple connected areas includes: generating a corresponding hash index according to the coordinates of each point cloud in the lane marking point set, and storing each point cloud in the lane marking point set in the hash table; determining the dynamic search distance of each point cloud in the lane marking point set, and performing a neighborhood search on the point cloud in the hash table based on the dynamic search distance to obtain the connected area in the lane marking point set.

[0009] Furthermore, an embodiment of the present invention provides a fourth possible implementation of the first aspect, wherein the lane line detection method also includes: obtaining the coordinates and pulse width of each point cloud in the connected area, and determining the collection distance of each point cloud based on the coordinates of each point cloud in the connected area; establishing a connected area pulse width and distance function based on the coordinates and pulse width of each point cloud in the connected area, and determining the pulse width error of each point cloud in the connected area based on the connected area pulse width and distance function; filtering out false detection points in the connected area based on the pulse width error, and obtaining the connected area after eliminating the false detection points.

[0010] Furthermore, an embodiment of the present invention provides a fifth possible implementation of the first aspect, wherein the step of screening out lane line areas based on the geometric features of the connected areas comprises: performing feature decomposition on each of the connected areas based on a preset feature decomposition algorithm to obtain decomposition features; determining the geometric shape of each of the connected areas based on the decomposition features, and determining the category of each of the connected areas based on the geometric shape of each of the connected areas; wherein the types of the connected areas include lane line connected areas and road surface marking connected areas; and establishing a minimum bounding box of the lane line connected areas to obtain a lane line area.

[0011] Further, an embodiment of the present invention provides a sixth possible implementation of the first aspect, wherein the step of establishing a minimum bounding box of the lane line connected area to obtain the lane line area includes: establishing a lane line minimum bounding box of the lane line connected area, and determining the yaw angle of the lane line based on the coordinates of the lane line minimum bounding box; performing posture correction on the point cloud within the lane line minimum bounding box based on the yaw angle of the lane line; obtaining a histogram corresponding to the lane line minimum bounding box after posture correction, and determining the lane line area based on the peak distribution of the histogram.

[0012] Further, an embodiment of the present invention provides a seventh possible implementation of the first aspect, which also includes: obtaining the longitudinal spacing distance between each of the lane line areas, and when the longitudinal spacing distance is greater than a preset distance, determining that the lane line corresponding to the lane line area is a dotted lane line, otherwise, determining that the lane line corresponding to the lane line area is a solid lane line.

[0013] In the second aspect, an embodiment of the present invention also provides a lane line detection device, including: a screening module, used to obtain point cloud data of the area to be detected, and screen out an initial ground point cloud from the point cloud data of the area to be detected; a division module, used to divide the initial ground point cloud into a ground point set and a lane marking point set according to the coordinates and pulse width of the initial ground point cloud; a search module, used to establish a hash table based on the lane marking point set, and perform dynamic distance search on the hash table to divide the lane marking point set into multiple connected areas; an establishment module, used to filter out lane line areas based on the geometric features of the connected areas, filter out lane line feature points from the lane line areas, and establish lane line curves based on the lane line feature points.

[0014] In a third aspect, an embodiment of the present invention provides a lane detection system, comprising: a laser radar and a controller, the controller comprising a processor and a storage device; a computer program is stored on the storage device, and the computer program, when executed by the processor, executes the method described in any one of the first aspects.

[0015] The embodiment of the present invention provides a lane line detection method, device and system, the method comprising: obtaining point cloud data of an area to be detected, filtering out an initial ground point cloud from the point cloud data of the area to be detected; dividing the initial ground point cloud into a ground point set and a lane marking point set according to the coordinates and pulse width of the initial ground point cloud; establishing a hash table based on the lane marking point set, performing a dynamic distance search on the hash table to divide the lane marking point set into multiple connected areas; filtering out lane line areas based on the geometric features of the connected areas, filtering out lane line feature points from the lane line areas, and establishing lane line curves based on the lane line feature points. The present invention preferentially filters out an initial ground point set, further divides the initial ground point set into a ground point set and a lane marking point set according to the coordinates and pulse width of the initial ground point set, filters out interference from ground points, obtains multiple connected areas by performing a dynamic distance search based on the hash table, and determines the lane line area according to the geometric features to establish a lane line curve, thereby reducing interference from other road surface markings on lane line detection, and will not be affected by changes in environmental parameters such as light intensity, thereby improving the accuracy of lane line detection.

[0016] Other features and advantages of the embodiments of the present invention will be described in the following description, or some features and advantages can be inferred or determined without doubt from the description, or can be learned by implementing the above-mentioned techniques of the embodiments of the present invention.

[0017] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 A flow chart of a lane line detection method provided by an embodiment of the present invention is shown;

[0020] Figure 2 A curve diagram showing the relationship between a collection distance and a pulse width provided by an embodiment of the present invention is shown;

[0021] Figure 3 A histogram of lane lines provided by an embodiment of the present invention is shown;

[0022] Figure 4 A schematic structural diagram of a lane line detection device provided in an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be described below in conjunction with the drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments.

[0024] At present, since the road is printed with not only lane lines but also lane signs such as lane indicator arrows, and the lane lines and lane signs are usually printed with the same material, the existing lane line detection technology does not accurately distinguish lane lines from other road surface signs, which is prone to false detection and interference, reducing the accuracy of lane line detection.

[0025] To improve the above problems, the embodiments of the present invention provide a lane line detection method, device and system, which can be applied to improve the accuracy of lane line detection. The embodiments of the present invention are described in detail below.

[0026] This embodiment provides a lane line detection method, which is applied to electronic devices such as computers. Figure 1 The lane line detection method flow chart shown in the figure includes the following steps:

[0027] Step S102, obtaining point cloud data of the area to be detected, and screening out an initial ground point cloud from the point cloud data of the area to be detected.

[0028] The point cloud data of the area to be detected is collected based on the sensor. The sensor can be a device such as a laser radar or a depth camera that can collect and output point cloud data. The ROI area (region of interest) is set according to the collection position of the sensor to extract the point cloud data of the near-ground area. The smaller the Z-axis coordinate value of the point cloud, the higher the probability of belonging to the ground point.

[0029] The ROI area is divided into multiple grid areas, and ground simulation is performed on the point clouds with smaller Z-axis coordinate values ​​in each grid area in the ROI area based on a preset algorithm to obtain a simulated ground. The preset algorithm may include ground simulation algorithms such as cross product, least squares method, PCA (Principal Component Analysis) algorithm and SVD (Singular Value Decomposition) algorithm. According to the vertical distance from the point cloud in each grid area to the simulated ground, the ground points are screened out from the point cloud in the ROI area based on the vertical distance, and the screened ground points are recorded as initial ground points. In a specific implementation, the point cloud with a vertical distance less than the preset distance can be used as the initial ground point cloud.

[0030] In one embodiment, when the above-mentioned sensor is set at the top of the vehicle, a three-dimensional coordinate system is established with the sensor as the origin and the vehicle's driving direction as the X-axis direction (that is, the direction in which the sensor points to the front of the vehicle is the positive direction of the X-axis), the lateral direction of the vehicle is the Y-axis, and the vertical upward direction of the sensor on the top of the vehicle is the positive direction of the Z-axis. The height of the sensor from the ground is set to h. In order to extract the point cloud data of the near-ground area, the point cloud data that satisfies the X-axis coordinate value being in the range of [-w, w], the Y-axis coordinate value being in the range of [-L, L], and the Z-axis coordinate value being less than or equal to -h+1 is set as the point cloud data of the ROI area.

[0031] Step S104, dividing the initial ground point cloud into a ground point set and a lane marking point set according to the coordinates and pulse width of the initial ground point cloud.

[0032] The pulse width reflected by each point cloud in the initial ground point cloud is obtained. Due to the different materials of the ground and the lane lines printed on the ground, the reflected pulse width is also different; when the collection distance between the point cloud data and the sensor is different, the pulse width of the point cloud will also be different. In order to accurately distinguish between ground points and lane marking points, the collection distance between each point cloud and the sensor (i.e., the coordinate origin) is calculated according to the coordinates of the initial ground point cloud. According to the collection distance and pulse width of the initial ground point cloud, each point cloud is judged to be a ground point or a lane marking point, thereby dividing the initial ground point cloud into a ground point cloud set and a lane marking point set.

[0033] Step S106: establish a hash table based on the lane marking point set, and perform dynamic distance search on the hash table to divide the lane marking point set into multiple connected areas.

[0034] The lane marking point set includes lane line point clouds and road surface marking (also referred to as lane marking) point clouds. Each point cloud in the lane marking point set is stored in a hash table. By performing a neighborhood dynamic distance search on the hash table, the point cloud is detected for connected areas, thereby improving the neighborhood search efficiency.

[0035] Step S108, filtering out lane line areas based on the geometric features of the connected areas, filtering out lane line feature points from the lane line areas, and establishing lane line curves based on the lane line feature points.

[0036] Since the above-mentioned connected areas include the connected areas formed by lane lines and the connected areas formed by road surface markings, in order to achieve effective segmentation of the lane line areas and the road surface marking areas, the geometric shapes of each connected area are judged. Since the lane lines are usually approximately straight lines and the lane line width is usually 10 to 15 centimeters, the lane line areas can be screened out from the connected areas according to the shape and size of the lane lines and the geometric shapes of each connected area.

[0037] A plurality of point cloud data with unequal X-axis coordinate values ​​are selected from the lane line area as lane line feature points, or point cloud data on the edge of the lane line area are selected as lane line feature points, and curve fitting is performed on the selected lane line feature points to obtain the lane line curve in the area to be detected.

[0038] The lane line detection method provided in this embodiment preferentially screens out an initial ground point set, further divides the initial ground point set into a ground point set and a lane marking point set according to the coordinates and pulse width of the initial ground point set, filters out interference from ground points, obtains multiple connected areas through dynamic distance search detection based on a hash table, and determines the lane line area based on geometric features to establish a lane line curve, thereby reducing interference from other road surface markings on lane line detection, and will not be affected by changes in environmental parameters such as light intensity, thereby improving the accuracy of lane line detection.

[0039] In one embodiment, this embodiment provides an implementation method for dividing the initial ground point cloud into a ground point set and a lane marking point set according to the coordinates and pulse width of the initial ground point cloud. Specifically, the implementation method may refer to the following steps (1) to (4):

[0040] Step (1): Obtain the coordinates and actual pulse width of each point cloud in the initial ground point cloud, and determine the collection distance of each point cloud in the initial ground point cloud based on the coordinates of each point cloud in the initial ground point cloud.

[0041] The coordinates of each point cloud in the currently collected initial ground point cloud are obtained, the collection distance between each point cloud and the sensor is calculated based on the coordinates of each point cloud in the initial ground point cloud, and the pulse width reflected by each point cloud in the initial ground point cloud is obtained, which is recorded as the actual pulse width.

[0042] Step (2): Obtain a first pulse width and distance function and a second pulse width and distance function.

[0043] The first pulse width and distance function and the second pulse width and distance function are obtained by fitting based on the acquisition distance and pulse width of the sample. The first pulse width and distance function is the functional relationship between the pulse width of the ground material point and the acquisition distance of the ground material point, and the second pulse width and distance function is the functional relationship between the pulse width of the lane line or marking line material point and the acquisition distance of the lane line or marking line material point.

[0044] The collection distance and pulse width of ground point sampling and the collection distance and pulse width of lane marking point sampling are respectively used as samples. By setting the sample ratio of random sampling, for example: extracting 50% of the samples by iteration N times, based on the principle of least squares, the function of the collection distance and pulse width of the samples is fitted, and according to the number of errors between the predicted value and the actual value less than a threshold, the optimal relationship function in the N fitting models is screened out, and the first pulse width and distance function is obtained based on the samples of the ground points, and the second pulse width and distance function is obtained based on the samples of the lane marking points.

[0045] In one implementation, the first pulse width and distance function is: PW=a×ln(Dis)+b; the first pulse width and distance function can reflect the functional relationship between the acquisition distance of the ground point and the pulse width.

[0046] The second pulse width and distance function is PW=c×ln(Dis)+d; the second pulse width and distance function can reflect the functional relationship between the collection distance of the lane marking point on the lane line printed on the ground or on the road surface marking and its pulse width.

[0047] Where PW is the pulse width, Dis is the acquisition distance, a, b, c and d are all fitting coefficients. ln(.) represents the logarithmic function with e as the base. Figure 2 The relationship curve between the acquisition distance and the pulse width is shown in the figure. Figure 2 The relationship between the acquisition distance and pulse width of the ground point and the acquisition distance and pulse width of the lane marking point are shown in Figure 2 It can be seen that the acquisition distance and pulse width of the point cloud under each material approximately obey the logarithmic function relationship. As the acquisition distance of the ground point and the lane marking point increases, the difference between the pulse width of the ground point and the pulse width of the lane marking point becomes larger and larger.

[0048] Step (3): Determine the first pulse width prediction value of each point cloud based on the acquisition distance of each point cloud in the initial ground point cloud and the first pulse width and distance function, and determine the second pulse width prediction value of each point cloud based on the acquisition distance of each point cloud in the initial ground point cloud and the second pulse width and distance function.

[0049] In order to distinguish whether each point cloud in the initial ground point cloud is a ground point or a lane marking point, the collection distance of each point cloud is respectively input into the above-mentioned first pulse width and distance function and the second pulse width and distance function, and the calculated pulse width is respectively recorded as the first pulse width prediction value and the second pulse width prediction value, and the first pulse width prediction value and the second pulse width prediction value corresponding to each point cloud in the initial ground point cloud are calculated.

[0050] Step (4): Calculate the first pulse width error between the actual pulse width of each point cloud and the first pulse width prediction value, and the second pulse width error between the actual pulse width of each point cloud and the second pulse width prediction value, and determine the material corresponding to each point cloud in the initial ground point cloud based on the first pulse width error and the second pulse width error, so as to divide the initial ground point cloud into a ground point set and a lane marking point set.

[0051] Since each point cloud has an actual pulse width of the detected reflection, for each point cloud, the difference between its actual pulse width and the first pulse width prediction value is calculated, and the difference is recorded as the first pulse width error. The difference between its actual pulse width and the second pulse width prediction value is calculated, and the difference is recorded as the second pulse width error.

[0052] If the first pulse width error is smaller than the second pulse width error, it indicates that the acquisition distance of the point cloud and the reflected pulse width are more consistent with the first pulse width and distance function relationship fitted by the ground point, and the point cloud is determined to be a ground point, and the point cloud is added to the ground point set.

[0053] If the second pulse width error is smaller than the first pulse width error, it indicates that the collection distance of the point cloud and the reflected pulse width are more consistent with the second pulse width and distance function relationship obtained by fitting the lane marking point. The point cloud is determined to be a lane marking point, and the lane marking point is added to the lane marking point set.

[0054] In one embodiment, this embodiment provides an implementation method of establishing a hash table based on the lane marking point set, performing a dynamic distance search on the hash table to divide the lane marking point set into multiple connected areas, which can be specifically performed with reference to the following steps 1) to 3):

[0055] Step 1): Generate a corresponding hash index according to the coordinates of each point cloud in the lane marking point set, and store each point cloud in the lane marking point set in the hash table.

[0056] Each point cloud in the lane marking point set is stored in a hash table. The creation process is to generate a hash index based on the three-dimensional coordinates of the point cloud and store the corresponding point in the hash table. According to the ROI area set above, the hash index corresponding to each point cloud in the lane marking point set is calculated. The calculation formula of the hash index is:

[0057] id=int(x)×(L×H)+int(z)×L+int(y)

[0058] Among them, id is the hash index corresponding to the point cloud, int(.) is the rounding function, (x, y, z) is the coordinate of the point cloud, L and H are the length and height of the ROI area respectively. By calculating the hash index based on the above formula and storing each point cloud in the lane identification point set into the hash table, the efficiency of neighborhood search can be improved when performing connected area detection.

[0059] Step 2): Determine the dynamic search distance of each point cloud in the lane marking point set, and perform a neighborhood search on the point cloud in the hash table based on the dynamic search distance to obtain a connected area in the lane marking point set.

[0060] The dynamic search distance calculation formula for any point cloud in the lane marking point set is:

[0061]

[0062] Where D is the dynamic search distance of the point cloud, dis is the acquisition distance of the point cloud (i.e., the distance from the point cloud to the lidar), α is the neighborhood scaling factor, α≥1, and Δθ is the vertical or horizontal angular resolution of the lidar.

[0063] Use breadth-first or width-first to dynamically search the point cloud in the lane marking point set, calculate the dynamic search distance of the current search point cloud to find the hash table, and check whether there is a valid point cloud belonging to the same type of target within the dynamic search distance in the hash table. If there is a valid point cloud within the dynamic search distance range of the current search point cloud, determine that the current search point cloud and the valid point cloud within the dynamic search distance range belong to the same type of target, and form a connected area with the current search point cloud and the valid point cloud belonging to the same type of target. The connected area in the lane marking point set can be determined by dynamically searching the hash table, and the connected area can be one or more.

[0064] Step 3): Obtain the coordinates and pulse width of each point cloud in the connected area, and determine the collection distance of each point cloud based on the coordinates of each point cloud in the connected area; establish a pulse width and distance function of the connected area based on the coordinates and pulse width of each point cloud in the connected area, and determine the pulse width error of each point cloud in the connected area based on the pulse width and distance function of the connected area; filter out the false detection points in the connected area based on the pulse width error, and obtain multiple connected areas after eliminating the false detection points.

[0065] Fit the relationship between the acquisition distance and the pulse width of the point cloud inside each connected area, record the fitted function as the connected area pulse width and distance function, input the acquisition distance of each point cloud in the connected area into the corresponding connected area pulse width and distance function to calculate the predicted pulse width, calculate the difference between the predicted pulse width and the pulse width actually reflected by the detected point cloud, and obtain the pulse width error of the point cloud. When the pulse width error of the point cloud is greater than the preset error, the point cloud is determined to be a false detection point and the point cloud is removed from the connected area to which it belongs. When the pulse width error of the point cloud is less than the preset error, the point cloud is determined not to be a false detection point and the point cloud is retained in the connected area.

[0066] The preset error may be set according to the pulse width error value of each point cloud in the connected area, such as the median of the pulse width errors of each point cloud in the connected area (ie, the median of the pulse width errors of all point clouds in the connected area is used as the preset error).

[0067] In one embodiment, this embodiment provides a specific implementation for filtering out lane line areas based on the geometric features of connected areas: performing feature decomposition on each connected area based on a preset feature decomposition algorithm to obtain decomposition features; determining the geometric shape of each connected area based on the decomposition features, and determining the category of each connected area based on the geometric shape of each connected area; wherein the types of connected areas include lane line connected areas and road surface marking connected areas; establishing a minimum bounding box of the lane line connected area to obtain the lane line area.

[0068] The above-mentioned preset feature decomposition algorithm can be a decomposition algorithm such as PCA (Principal Component Analysis) algorithm or SVD (Singular Value Decomposition). Since the lane line is usually approximately a straight line, PCA decomposition or SVD decomposition is performed on each connected area based on the preset feature decomposition algorithm to obtain multiple eigenvalues, and the relationship between the eigenvalues ​​is compared to determine the geometric shape of the connected area.

[0069] In a specific implementation, PCA decomposition can be performed on the connected area to obtain the eigenvalues ​​λ1 and λ2 of the geometric set. When the ratio of the eigenvalues ​​λ1 to λ2 is greater than or equal to a preset threshold T, it indicates that the characteristic decomposition result meets the threshold, and the geometric shape of the connected area meets the linear characteristics. The connected area is determined to be a lane line connected area. Otherwise, the connected area is determined to be a road surface marking connected area.

[0070]

[0071] Based on the minimum bounding box algorithm, the minimum bounding box is fitted for the lane line connected area in the above connected area, and the lane line area is further screened out from the above lane line connected area based on the actual size information of the lane line. According to the new road marking specifications, the width of the lane line is usually 10 to 15 cm, the width of the zebra crossing similar to the shape of the lane line is about 45 cm, and the shapes of other road markings and lane lines are quite different. The above lane line connected areas are fitted into the minimum bounding box, and the size information of each minimum bounding box is judged according to the vertex coordinates of each minimum bounding box to see whether it meets the actual size of the lane line, and the minimum bounding box that meets the actual size of the lane line is used as the lane line area.

[0072] In one embodiment, this embodiment provides a specific implementation method for establishing a minimum bounding box of a lane line connected area to obtain a lane line area: establishing a lane line minimum bounding box of a lane line connected area, and determining the yaw angle of the lane line based on the coordinates of the lane line minimum bounding box; performing posture correction on the point cloud within the lane line minimum bounding box based on the yaw angle of the lane line; obtaining a histogram corresponding to the lane line minimum bounding box after posture correction, and determining the lane line area based on the peak distribution of the histogram.

[0073] Lane lines usually appear in pairs and are parallel to each other. For example, when lane lines are detected based on sensors on the top of a vehicle, lane lines will appear on both sides of the vehicle, forming a pair of lane lines. The yaw angle of the lane line can be determined based on the border slope of the minimum bounding box corresponding to each lane line area. The correction matrix for point cloud posture correction can be calculated based on the average yaw angle of the lane lines that appear in pairs. The point cloud in the lane line area is posture corrected based on the correction matrix, and the histogram of the lane lines after posture correction is calculated. The sum of the histogram peaks on the left and right sides of the vehicle in the histogram is calculated, and the lane line corresponding to the maximum peak sum is the final lane line area. See, for example, Figure 3 The histogram of lane lines shown, Figure 3 The maximum peaks on both sides and the corresponding long straight lines are the lane line areas.

[0074] In one embodiment, the lane line detection method provided in this embodiment also includes: obtaining the longitudinal spacing distance between each lane line area, and when the longitudinal spacing distance is greater than a preset distance, determining that the lane line corresponding to the lane line area is a dotted lane line, otherwise, determining that the lane line corresponding to the lane line area is a solid lane line.

[0075] Since the solid lane line is usually a connected long straight line, and the dotted lane line is usually a discontinuous line segment, it is determined whether the longitudinal spacing distance between the lane line area and the lane line area in the longitudinal direction (i.e., the extension direction of the lane line) is greater than the preset distance. If so, the lane line corresponding to the lane line area is determined to be a dotted lane line, otherwise, the lane line corresponding to the lane line area is determined to be a solid lane line. The above preset distance can be determined based on the actual spacing distance of the dotted lane lines, such as 1 to 2 meters.

[0076] The lane line detection method provided in this embodiment filters out the interference of ground points by selecting a set of lane marking points according to the collection distance and pulse width of the point cloud, and realizes the segmentation of lane lines and road surface markings by performing connected area detection based on a hash table and dynamic search distance. By performing feature decomposition on the connected area and determining the lane line area based on the geometric features of the connected area, the interference of other road surface markings on lane line detection is reduced, thereby improving the accuracy of lane line detection.

[0077] Corresponding to the lane line detection method provided in the above embodiment, the embodiment of the present invention provides a lane line detection device, see Figure 4 A schematic diagram of the structure of a lane line detection device is shown, and the device includes the following modules:

[0078] The screening module 41 is used to obtain the point cloud data of the area to be detected, and screen out the initial ground point cloud from the point cloud data of the area to be detected.

[0079] The division module 42 is used to divide the initial ground point cloud into a ground point set and a lane marking point set according to the coordinates and pulse width of the initial ground point cloud.

[0080] The search module 43 is used to establish a hash table based on the lane marking point set, and perform dynamic distance search on the hash table to divide the lane marking point set into multiple connected areas.

[0081] The establishment module 44 is used to filter out the lane line area based on the geometric features of the connected area, filter out the lane line feature points from the lane line area, and establish the lane line curve based on the lane line feature points.

[0082] The lane line detection device provided in the present embodiment filters out interference from ground points by preferentially screening out an initial ground point set, further dividing the initial ground point set into a ground point set and a lane marking point set according to the coordinates and pulse width of the initial ground point set, and obtains multiple connected areas through dynamic distance search detection based on a hash table. Lane line areas are determined based on geometric features to establish lane line curves, thereby reducing interference from other road surface markings on lane line detection, and will not be affected by changes in environmental parameters such as light intensity, thereby improving the accuracy of lane line detection.

[0083] In one embodiment, the division module 42 is further used to obtain the coordinates and actual pulse width of each point cloud in the initial ground point cloud, and determine the collection distance of each point cloud in the initial ground point cloud based on the coordinates of each point cloud in the initial ground point cloud; obtain a first pulse width and distance function and a second pulse width and distance function; wherein the first pulse width and distance function is a functional relationship between the pulse width of a ground material point and the collection distance of the ground material point, and the second pulse width and distance function is a functional relationship between the pulse width of a lane line or marking line material point and the collection distance of a lane line or marking line material point; based on the initial The first pulse width prediction value of each point cloud in the ground point cloud is determined based on the collection distance of each point cloud in the ground point cloud and the first pulse width and the distance function. The second pulse width prediction value of each point cloud is determined based on the collection distance of each point cloud in the initial ground point cloud and the second pulse width and the distance function. The first pulse width error between the actual pulse width of each point cloud and the first pulse width prediction value and the second pulse width error between the actual pulse width of each point cloud and the second pulse width prediction value are calculated respectively. The material corresponding to each point cloud in the initial ground point cloud is judged based on the first pulse width error and the second pulse width error, so as to divide the initial ground point cloud into a ground point set and a lane marking point set.

[0084] In one embodiment, the first pulse width and distance function is: PW=a×ln(Dis)+b; the second pulse width and distance function is PW=c×ln(Dis)+d; wherein PW is the pulse width, Dis is the acquisition distance, and a, b, c and d are all fitting coefficients.

[0085] In one embodiment, the above-mentioned search module 43 is further used to generate a corresponding hash index according to the coordinates of each point cloud in the lane marking point set, and store each point cloud in the lane marking point set in a hash table; determine the dynamic search distance of each point cloud in the lane marking point set, and perform a neighborhood search on the point cloud in the hash table based on the dynamic search distance to obtain a connected area in the lane marking point set.

[0086] In one embodiment, the above device further comprises:

[0087] The filtering module is used to obtain the coordinates and pulse width of each point cloud in the connected area, determine the collection distance of each point cloud based on the coordinates of each point cloud in the connected area; establish the pulse width and distance function of the connected area based on the coordinates and pulse width of each point cloud in the connected area, and determine the pulse width error of each point cloud in the connected area based on the pulse width and distance function of the connected area; filter out the false detection points in the connected area based on the pulse width error, and obtain the connected area after eliminating the false detection points.

[0088] In one embodiment, the above-mentioned establishment module 44 is further used to perform feature decomposition on each connected area based on a preset feature decomposition algorithm to obtain decomposition features; determine the geometric shape of each connected area based on the decomposition features, and determine the category of each connected area based on the geometric shape of each connected area; wherein the types of connected areas include lane line connected areas and road surface marking connected areas; establish the minimum bounding box of the lane line connected area to obtain the lane line area.

[0089] In one embodiment, the above-mentioned establishment module 44 is further used to establish a lane line minimum bounding box for the lane line connection area, determine the yaw angle of the lane line based on the coordinates of the lane line minimum bounding box; perform posture correction on the point cloud within the lane line minimum bounding box based on the yaw angle of the lane line; obtain the histogram corresponding to the lane line minimum bounding box after posture correction, and determine the lane line area based on the peak distribution of the histogram.

[0090] In one embodiment, the above device further comprises:

[0091] The determination module is used to obtain the longitudinal spacing distance between each lane line area. When the longitudinal spacing distance is greater than a preset distance, the lane line corresponding to the lane line area is determined to be a dotted lane line. Otherwise, the lane line corresponding to the lane line area is determined to be a solid lane line.

[0092] The lane line detection device provided in this embodiment filters out the interference of ground points by selecting a set of lane marking points according to the collection distance and pulse width of the point cloud, and realizes the segmentation of lane lines and road surface markings by performing connected area detection based on a hash table and dynamic search distance. By performing feature decomposition on the connected area and determining the lane line area based on the geometric features of the connected area, the interference of other road surface markings on lane line detection is reduced, thereby improving the accuracy of lane line detection.

[0093] The implementation principle and technical effects of the device provided in this embodiment are the same as those of the aforementioned embodiments. For the sake of brief description, for matters not mentioned in the device embodiment, reference may be made to the corresponding contents in the aforementioned method embodiment.

[0094] An embodiment of the present invention provides a lane line detection system, including: a laser radar and a controller, the laser radar is communicatively connected to the controller, and the laser radar can scan point cloud data of the area to be detected. The controller includes a processor and a storage device; a computer program is stored on the storage device, and when the computer program is run by the processor, the lane line detection method provided in the above embodiment is executed.

[0095] An embodiment of the present invention provides a computer-readable medium, wherein the computer-readable medium stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the method described in the above embodiment.

[0096] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned embodiment, and will not be repeated here.

[0097] The computer program product of the lane line detection method, device and system provided in the embodiments of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the previous method embodiments. The specific implementation can be found in the method embodiments, which will not be repeated here.

[0098] In addition, in the description of the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0099] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0100] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0101] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A lane line detection method, characterized in that: include: Acquire point cloud data of the area to be detected, and filter out an initial ground point cloud from the point cloud data of the area to be detected; Dividing the initial ground point cloud into a ground point set and a lane marking point set according to the coordinates and pulse width of the initial ground point cloud; Establishing a hash table based on the lane marking point set, and performing a dynamic distance search on the hash table to divide the lane marking point set into a plurality of connected areas; Filtering a lane line area based on the geometric features of the connected area, filtering a lane line feature point from the lane line area, and establishing a lane line curve based on the lane line feature point; The step of selecting lane line areas based on the geometric features of the connected areas comprises: performing feature decomposition on each of the connected areas based on a preset feature decomposition algorithm to obtain decomposition features; determining the geometric form of each of the connected areas based on the decomposition features, and determining the type of each of the connected areas based on the geometric form of each of the connected areas; wherein the types of the connected areas include lane line connected areas and road surface marking connected areas; A lane line minimum bounding box for the lane line connected area is established, and the yaw angle of the lane line is determined based on the coordinates of the lane line minimum bounding box; a point cloud within the lane line minimum bounding box is posture corrected based on the yaw angle of the lane line; a histogram corresponding to the lane line minimum bounding box after posture correction is obtained, and the lane line area is determined based on the peak distribution of the histogram.

2. The method according to claim 1, characterized in that The step of dividing the initial ground point cloud into a ground point set and a lane marking point set according to the coordinates and pulse width of the initial ground point cloud comprises: Acquire the coordinates and actual pulse width of each point cloud in the initial ground point cloud, and determine the collection distance of each point cloud in the initial ground point cloud based on the coordinates of each point cloud in the initial ground point cloud; Obtain a first pulse width and distance function and a second pulse width and distance function; wherein the first pulse width and distance function is a functional relationship between a pulse width of a ground material point and a collection distance of the ground material point, and the second pulse width and distance function is a functional relationship between a pulse width of a lane line or marking line material point and a collection distance of the lane line or marking line material point; Determine a first pulse width prediction value of each point cloud based on the acquisition distance of each point cloud in the initial ground point cloud and the first pulse width and distance function, and determine a second pulse width prediction value of each point cloud based on the acquisition distance of each point cloud in the initial ground point cloud and the second pulse width and distance function; The first pulse width error between the actual pulse width of each point cloud and the first pulse width prediction value and the second pulse width error between the actual pulse width of each point cloud and the second pulse width prediction value are calculated respectively, and the material corresponding to each point cloud in the initial ground point cloud is determined based on the first pulse width error and the second pulse width error, so as to divide the initial ground point cloud into a ground point set and a lane marking point set.

3. The method according to claim 2, characterized in that The first pulse width and distance function is: PW = a × ln (Dis) + b; The second pulse width and distance function is PW=c×ln(Dis)+d; Wherein, PW is the pulse width, Dis is the acquisition distance, and a, b, c and d are all fitting coefficients.

4. The method according to claim 1, characterized in that: The step of establishing a hash table based on the lane marking point set and performing a dynamic distance search on the hash table to divide the lane marking point set into a plurality of connected areas comprises: Generate a corresponding hash index according to the coordinates of each point cloud in the lane marking point set, and store each point cloud in the lane marking point set in a hash table; Determine a dynamic search distance of each point cloud in the lane marking point set, and perform a neighborhood search on the point cloud in the hash table based on the dynamic search distance to obtain a connected area in the lane marking point set.

5. The method according to claim 4, characterized in that Also includes: Acquire the coordinates and pulse width of each point cloud in the connected area, and determine the collection distance of each point cloud based on the coordinates of each point cloud in the connected area; Establishing a connected area pulse width and distance function based on the coordinates and pulse width of each point cloud in the connected area, and determining the pulse width error of each point cloud in the connected area based on the connected area pulse width and distance function; The false detection points in the connected area are filtered out based on the pulse width error to obtain the connected area after the false detection points are removed.

6. The method according to any one of claims 1 to 5, characterized in that: Also includes: The longitudinal spacing distance between each lane line area is obtained. When the longitudinal spacing distance is greater than a preset distance, the lane line corresponding to the lane line area is determined to be a dotted lane line. Otherwise, the lane line corresponding to the lane line area is determined to be a solid lane line.

7. A lane line detection device, characterized in that: include: A screening module is used to obtain point cloud data of the area to be detected, and screen out an initial ground point cloud from the point cloud data of the area to be detected; A division module, used for dividing the initial ground point cloud into a ground point set and a lane marking point set according to the coordinates and pulse width of the initial ground point cloud; A search module, used to establish a hash table based on the lane marking point set, and perform a dynamic distance search on the hash table to divide the lane marking point set into a plurality of connected areas; An establishment module, used for screening out a lane line area based on the geometric features of the connected area, screening out lane line feature points from the lane line area, and establishing a lane line curve based on the lane line feature points; The filtering out of lane line areas based on the geometric features of the connected areas includes: performing feature decomposition on each of the connected areas based on a preset feature decomposition algorithm to obtain decomposition features; determining the geometric form of each of the connected areas based on the decomposition features, and determining the category of each of the connected areas based on the geometric form of each of the connected areas; wherein the types of the connected areas include lane line connected areas and road surface marking connected areas; A lane line minimum bounding box for the lane line connected area is established, and the yaw angle of the lane line is determined based on the coordinates of the lane line minimum bounding box; a point cloud within the lane line minimum bounding box is posture corrected based on the yaw angle of the lane line; a histogram corresponding to the lane line minimum bounding box after posture correction is obtained, and the lane line area is determined based on the peak distribution of the histogram.

8. A lane detection system, characterized in that: include: A laser radar and a controller, wherein the controller includes a processor and a storage device; The storage device stores a computer program, which, when executed by the processor, executes the method according to any one of claims 1 to 6.