A lane line detection method and device
By acquiring the intensity and elevation maps of LiDAR point cloud data and combining them with vehicle trajectory points, lane line regions are extracted and vectorized, solving the difficulty of lane line detection in high-precision maps and achieving high-precision and automated lane line detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA AUTOMOTIVE INNOVATION CORP
- Filing Date
- 2022-10-27
- Publication Date
- 2026-04-17
AI Technical Summary
Existing lane detection algorithms mostly rely on camera images, which are not suitable for creating high-precision maps based on LiDAR point cloud data. Furthermore, the point cloud reflection intensity map does not have a fixed direction for lane lines, making detection difficult.
By acquiring point cloud intensity maps and elevation maps, and combining them with vehicle trajectory points, the main direction interval of the target preset lane line is determined. Point cloud intensity and elevation sub-maps that meet preset conditions are extracted, and lane line segmentation maps are generated and vectorized.
It enables accurate detection of lane lines in high-precision maps, removes interference from vegetation and impurity spots, and improves the efficiency and automation of high-precision map making.
Smart Images
Figure CN115690710B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer vision, and in particular to a lane line detection method and apparatus. Background Technology
[0002] For high-definition map production, lane lines, as one of the main road elements, are particularly important for detection. Current mainstream lane line detection algorithms are mostly applied to autonomous driving scenarios, relying on camera image acquisition and recognition. They assume lane lines follow a near-to-far trend in the image, which is unsuitable for high-definition map lane line detection. Furthermore, the basic data for high-definition map production is mostly point cloud data obtained from LiDAR scanning. After forming an intensity map based on the point cloud reflection intensity, it does not provide a fixed lane line orientation. Summary of the Invention
[0003] To address the problems of existing technologies, this application provides a lane line detection method and apparatus. The technical solution is as follows:
[0004] On the one hand, a lane line detection method is provided, the method comprising:
[0005] Obtain the point cloud intensity map and point cloud elevation map corresponding to the area to be detected; the point cloud intensity map and the point cloud elevation map are obtained based on the ground point cloud data within the area to be detected collected by the lidar on the acquisition vehicle;
[0006] Based on the collected vehicle trajectory points, the target preset lane line main direction interval is determined from multiple preset lane line main direction intervals;
[0007] Based on the main direction interval of the target preset lane line, extract the target point cloud intensity sub-map that meets the preset intensity condition from the point cloud intensity map;
[0008] Extract a target point cloud elevation sub-map from the point cloud elevation map; the elevation values of the pixels in the target point cloud elevation sub-map meet preset elevation conditions;
[0009] Based on the target point cloud intensity sub-map and the target point cloud elevation sub-map, a lane line segmentation map is determined, and based on the coordinate information of the pixels in the lane line segmentation map, the lane line vectorization result is determined.
[0010] On the other hand, a lane line detection device is provided, the device comprising:
[0011] The image acquisition module is used to acquire the point cloud intensity map and the point cloud elevation map corresponding to the area to be detected; the point cloud intensity map and the point cloud elevation map are obtained based on the ground point cloud data within the area to be detected collected by the lidar on the acquisition vehicle;
[0012] The main direction determination module is used to determine the target preset lane line main direction interval from multiple preset lane line main direction intervals based on the driving trajectory points of the collected vehicle.
[0013] The intensity map extraction module is used to extract a target point cloud intensity sub-map that meets the preset intensity conditions from the point cloud intensity map based on the main direction interval of the target preset lane line.
[0014] The elevation map extraction module is used to extract a target point cloud elevation sub-map from the point cloud elevation map; the elevation values of the pixels in the target point cloud elevation sub-map meet preset elevation conditions;
[0015] The lane line determination module is used to determine a lane line segmentation map based on the target point cloud intensity sub-map and the target point cloud elevation sub-map, and to determine the lane line vectorization result based on the coordinate information of the pixels in the lane line segmentation map.
[0016] In one exemplary embodiment, the apparatus further includes an image creation module for generating point cloud intensity maps and point cloud elevation maps, the image creation module comprising:
[0017] The point cloud acquisition module is used to acquire ground point cloud data within the detection range collected by the lidar;
[0018] The sample extraction module is used to extract multiple sample 3D points from the ground point cloud data based on a preset sampling interval;
[0019] The sample 2D conversion module is used to project the multiple sample 3D points onto the first 2D image to obtain the first projected image;
[0020] A region segmentation module is used to divide the first projected image into multiple sampling regions;
[0021] The pixel value determination module is used to determine the average intensity of pixels in each sampling region and the average elevation of pixels in each sampling region.
[0022] The image generation module is used to generate a point cloud intensity map based on the average intensity value corresponding to each sampling region, and to generate a point cloud elevation map based on the average elevation value corresponding to each sampling region.
[0023] In one exemplary embodiment, the plurality of preset lane line main direction intervals are a preset number of equal-angle intervals, and the main direction determination module includes:
[0024] The trajectory point 2D conversion module is used to project the driving trajectory points onto a second 2D image to obtain 2D trajectory points;
[0025] The trajectory point fitting module is used to perform straight line fitting on the two-dimensional trajectory points to obtain the target trajectory straight line;
[0026] Angle determination module is used to determine the absolute value of the angle between the target trajectory line and the positive direction of the horizontal axis of the second two-dimensional image, thereby obtaining the target trajectory angle;
[0027] The main direction interval determination module is used to determine the target isoangular interval that matches the angle between the preset number of isoangular intervals and the target trajectory, and to use the target isoangular interval as the main direction interval of the target preset lane line.
[0028] In one exemplary embodiment, the intensity map extraction module includes:
[0029] The first sliding window determination module is used to determine the first sliding window based on the lane line width; the first sliding window consists of an inner neighborhood and an outer neighborhood on the same axis, the inner neighborhood is a square with a side length equal to the lane line width, and the outer neighborhood is a square ring with an inner side coinciding with the outer side of the inner neighborhood and an outer side length greater than the lane line width;
[0030] The first region extraction module is used to slide the first sliding window along the first preset direction on the point cloud intensity map with a first preset step size, and extract the region in the first coverage area that meets the preset intensity condition at each first preset step size, until the overlap between the area slid by the inner neighborhood of the first sliding window and the point cloud intensity map meets the first preset requirement; the first coverage area is the area on the point cloud intensity map that overlaps with the first sliding window.
[0031] The first sub-image generation module is used to generate a preliminary point cloud intensity sub-image based on the regions that meet the preset intensity conditions corresponding to each first preset step size.
[0032] The second sub-map generation module is used to generate the target point cloud intensity sub-map based on the target preset lane line main direction interval and the preliminary point cloud intensity sub-map.
[0033] In one exemplary embodiment, the first region extraction module includes:
[0034] The first pixel point statistics module is used to determine the first statistical value of the pixels in the area that overlaps with the inner neighborhood in the first coverage area under the current first preset step size;
[0035] The second pixel point statistics module is used to determine the second statistical value of the pixels in the area that overlaps with the outer neighborhood in the first coverage area under the current first preset step size;
[0036] The first pixel extraction module is used to extract pixels in the area that overlaps with the inner neighborhood in the first coverage area under the current first preset step size when the first statistical value and the second statistical value satisfy the preset intensity condition.
[0037] In one exemplary embodiment, the second subgraph generation module includes:
[0038] The second sliding window determination module is used to determine a second sliding window based on the target preset lane line main direction interval and the lane line width; the second sliding window is composed of a first mask and a second mask;
[0039] The second region extraction module is used to slide the second sliding window along the second preset direction on the preliminary point cloud intensity sub-map at a second preset step size, and extract the region in the second coverage area that meets the preset direction condition under each second preset step size, until the overlap between the target area slid by the second sliding window and the point cloud intensity map meets the second preset requirement; the second coverage area is the region on the preliminary point cloud intensity sub-map that overlaps with the second sliding window;
[0040] The third sub-graph generation module is used to generate the target point cloud intensity sub-graph based on the regions that meet the preset direction conditions corresponding to each second preset step size.
[0041] In one exemplary embodiment, the second region extraction module includes:
[0042] The third pixel point statistics module is used to determine the number of pixels with pixel values greater than a preset pixel value in the second coverage area that overlaps with the internal statistics area of the first mask under the current second preset step size, and to obtain the first quantity value;
[0043] The fourth pixel point statistics module is used to determine the number of pixels in the second coverage area that overlaps with the internal statistics area of the second mask under the current second preset step size, whose pixel value is greater than a preset pixel value, and obtain the second quantity value;
[0044] The second pixel extraction module is used to extract pixels in the area that overlaps with the internal statistical area of the second mask in the second coverage area under the current second preset step size when the ratio of the first quantity value and the second quantity value is within a preset range.
[0045] In one exemplary embodiment, the elevation map extraction module includes:
[0046] The third sliding window determination module is used to determine a third sliding window based on the lane line width; the third sliding window is a square with a side length equal to the lane line width;
[0047] The third region extraction module slides the third sliding window along the third preset direction on the point cloud elevation map at a third preset step size, based on the third sliding window, to extract the third coverage region that meets the preset elevation condition at each third preset step size, until the overlap between the area traversed by the third sliding window and the point cloud intensity map meets the third preset requirement; the third coverage region is the region on the point cloud elevation map that overlaps with the third sliding window; the preset elevation condition is that the standard deviation of the pixels in the third coverage region is less than a preset threshold.
[0048] The fourth sub-map generation module is used to generate the target point cloud elevation sub-map based on the third coverage area that meets the preset elevation conditions corresponding to each third preset step size.
[0049] In one exemplary embodiment, the lane line determination module includes:
[0050] A pixel determination module is used to determine the overlapping pixels of the target point cloud intensity sub-image and the target point cloud elevation sub-image;
[0051] A segmentation map generation module is used to generate the lane line segmentation map based on the overlapping pixels;
[0052] A centerline determination module is used to determine the centerline of each lane line in the lane line segmentation diagram;
[0053] The pixel restoration module is used to restore the coordinates of the pixels on the center line to three-dimensional coordinates to obtain the lane line vectorization result.
[0054] On the other hand, an electronic device is provided, including a processor and a memory, wherein the memory stores at least one instruction or at least one program, the at least one instruction or the at least one program being loaded and executed by the processor to implement the lane detection method of any of the above aspects.
[0055] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction or at least one program is stored therein, the at least one instruction or the at least one program being loaded and executed by a processor to implement the lane detection method as described above.
[0056] On the other hand, a computer program product or computer program is provided, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the lane detection method of any of the above aspects.
[0057] This application's embodiments estimate the approximate direction of lane lines in the intensity map by using the trajectory points corresponding to the point cloud, design corresponding intensity distribution rules, and combine other intensity and elevation features of the lane lines to filter out lane line regions. This method not only solves the problem of intensity maps lacking fixed lane line directions, but also removes vegetation on both sides of the lanes and impurity spots on the lanes through multiple condition constraints, resulting in more accurate lane line regions for subsequent lane line vectorization work, assisting manual mapping, greatly improving the efficiency of high-precision map production, and promoting automated high-precision map production. Attached Figure Description
[0058] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0059] Figure 1 This is a schematic flowchart of a lane line detection method provided in an embodiment of this application;
[0060] Figure 2 This is a schematic flowchart of the process for determining the main direction of a lane line according to an embodiment of this application;
[0061] Figure 3 This is a schematic diagram of the lane line main direction section division provided in the embodiments of this application;
[0062] Figure 4 This is a flowchart illustrating the basic statistical rules for setting lane lines in a point cloud intensity map, as provided in an embodiment of this application.
[0063] Figure 5 This is a flowchart illustrating the process of setting lane direction statistics rules for point cloud intensity maps according to an embodiment of this application.
[0064] Figure 6 This is a schematic diagram of a lane line detection mask with defined direction provided in an embodiment of this application;
[0065] Figure 7 This is a flowchart illustrating the basic statistical rules for setting lane lines in a point cloud intensity map, as provided in an embodiment of this application.
[0066] Figure 8 This is a flowchart illustrating the process of determining lane line vectorization results provided in an embodiment of this application;
[0067] Figure 9 This is a structural block diagram of a lane line detection device provided in an embodiment of this application;
[0068] Figure 10This is a hardware structure block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation
[0069] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0070] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.
[0071] It is understood that in the specific embodiments of this application, data such as user information are involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.
[0072] Please see Figure 1 The diagram illustrates a flow chart of a lane line detection method provided in this application. It should be noted that while this specification provides the operational steps of the method as described in the embodiments or flowcharts, more or fewer operational steps may be included based on conventional or non-inventive methods. The order of steps listed in the embodiments is merely one possible execution order among many and does not represent the only possible execution order. In actual system or product execution, the method can be executed sequentially or in parallel (e.g., in a parallel processor or multi-threaded processing environment) as shown in the embodiments or accompanying drawings. Specifically, as... Figure 1 As shown, the method may include:
[0073] S101, obtain the point cloud intensity map and point cloud elevation map corresponding to the area to be detected.
[0074] The point cloud intensity map and point cloud elevation map are obtained based on ground point cloud data within the detection range collected by the lidar on the acquisition vehicle. Specifically, this may include the following steps:
[0075] Acquire ground point cloud data within the detection range collected by lidar;
[0076] Multiple sample 3D points are extracted from the ground point cloud data based on a preset sampling interval;
[0077] The multiple sample 3D points are projected onto the first 2D image to obtain the first projected image;
[0078] The first projected image is divided into multiple sampling regions;
[0079] Determine the mean intensity of pixels in each sampling region, and the mean elevation of pixels in each sampling region;
[0080] A point cloud intensity map is obtained based on the average intensity value corresponding to each sampling area, and a point cloud elevation map is obtained based on the average elevation value corresponding to each sampling area.
[0081] The preset sampling interval is a distance interval. When setting the preset sampling interval, the actual lane width should be taken into account. Ideally, the value of the preset sampling interval should be divisible by the actual lane width. For example, if the actual lane width is 15 centimeters, the preset sampling interval can be 3 centimeters.
[0082] When dividing the first projected image into multiple sampling regions, the lane line width in the first projected image should be taken into account, and the side length of the sampling region should be divisible by the lane line width in the first projected image. Specifically, the lane line width on the point cloud intensity map is the actual lane line width divided by the preset sampling interval. For example, if the actual lane line width is 15 cm, and the preset sampling interval is 3 cm, then the lane line width in the first projected image is 5 cm, and the side length of the sampling region can be 2.5 cm.
[0083] The first two-dimensional image is a newly created blank two-dimensional image, which is used to receive two-dimensional points formed by the projection of the sample three-dimensional points to generate the first projected image.
[0084] In practice, for each sample 3D point P, it is projected onto the first 2D image using formula (1). Where P... x P y Let x and y be the x and y coordinates of point P, respectively, and let xmin and ymin be the minimum x and y coordinates of this point cloud. x Index y Let P be the pixel coordinates of the sample 3D point P falling on the first 2D image, and g be the sampling interval.
[0085]
[0086] Simultaneously, the average intensity and average elevation of the pixels falling into each sampling area are calculated and used as the pixel value of the current pixel, generating a point cloud intensity map and a point cloud elevation map respectively.
[0087] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application sample ground point cloud data and convert the collected sample points into two-dimensional points, and generate point cloud intensity maps and point cloud elevation maps based on these two-dimensional points, providing implementation objects for extracting lane line areas.
[0088] S103, based on the collected vehicle's driving trajectory points, determine the target preset lane line main direction interval from multiple preset lane line main direction intervals.
[0089] Among them, the main direction intervals of multiple preset lane lines are preset number of equal angle intervals.
[0090] In practice, the preset number is 4. The direction of the lane line is divided into four cases according to the interval in which it is located. That is, all possible directions are divided into eight regions with an angle of π / 4. Two centrally symmetrical regions form one interval, which is the preset lane line main direction interval.
[0091] In one exemplary implementation, such as Figure 2 As shown, step S103 above may include:
[0092] S201, Project the driving trajectory points onto the second two-dimensional image to obtain two-dimensional trajectory points.
[0093] Among them, the driving trajectory points are the trajectory points formed by the vehicle when the lidar on the data collection vehicle collects ground point cloud data. The direction of the driving trajectory points can reflect the direction of the lane lines that the data collection vehicle passes through.
[0094] The second two-dimensional image is a newly created two-dimensional blank image, used to receive the two-dimensional coordinates of the vehicle's driving trajectory points, i.e., the two-dimensional trajectory points.
[0095] S203, perform straight line fitting on the two-dimensional trajectory points to obtain the target trajectory straight line.
[0096] In practice, the least squares method is used to fit the two-dimensional coordinates of the two-dimensional trajectory points to a straight line.
[0097] S205, determine the absolute value of the angle between the target trajectory line and the positive direction of the horizontal axis of the second two-dimensional image to obtain the target trajectory angle.
[0098] The horizontal axis is a number line drawn in the horizontal direction of the second two-dimensional image. The right side of the second two-dimensional image is defined as the positive direction of the horizontal axis. The angle between the fitted target trajectory line and the positive direction of the horizontal axis is calculated and the absolute value θ is taken.
[0099] S207, determine the target isoangular interval that matches the angle between the preset number of isoangular intervals and the target trajectory, and use the target isoangular interval as the main direction interval of the target preset lane line.
[0100] The target angle range is the preset main direction range of the lane line to which the lane line direction belongs, and needs to be determined according to the θ corresponding to the lane line.
[0101] In specific implementation, such as Figure 3 The diagram shows the division of the lane lines in the main direction. Since the lane lines in the ground point cloud data are not necessarily straight lines, they may have a certain curvature (except for excessive curvature or circular roads). Therefore, the range θ [0, π] is divided into 4 intervals as shown in formula (2), which correspond to... Figure 3 The intervals are 1, 2, 3 and 4.
[0102]
[0103] By determining the interval where θ is located, which is the main direction interval of the target preset lane line, the limiting direction in the subsequent lane line detection process is determined.
[0104] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application divide the main direction interval of the lane line, and determine the angle interval to which the driving trajectory points of the collected vehicle belong, i.e. the target preset lane line main direction interval, by dividing the main direction interval of the lane line, and corresponding the direction of the driving trajectory points of the collected vehicle to the lane line direction. This provides a basis for subsequently limiting the lane line direction on the point cloud intensity map. While limiting the linear characteristics of the lane line, it can also adapt to the determination of the lane line direction with a small curvature.
[0105] S105, Based on the main direction interval of the target preset lane line, extract the target point cloud intensity sub-map that meets the preset intensity conditions from the point cloud intensity map.
[0106] The preset intensity condition is a numerical requirement set based on the distribution and intensity values of pixels in the point cloud intensity map, and the target point cloud intensity sub-map is an image composed of pixels in the region that meets the above requirements.
[0107] Filter out pixels in non-lane line areas from the pixel intensity map to eliminate interference from non-road areas (such as vegetation and guardrails) and road areas (such as incomplete lane lines, road potholes and patches, and uneven LiDAR reflection intensity) in lane line detection.
[0108] In one exemplary implementation, such as Figure 4 As shown, step S105 above may include:
[0109] S401, determine the first sliding window based on the lane line width.
[0110] The first sliding window consists of an inner neighborhood and an outer neighborhood on the same axis. The inner neighborhood is a square with a side length equal to the width of the lane line, and the outer neighborhood is a square ring with an inner side that coincides with the outer side of the inner neighborhood and an outer side length greater than the width of the lane line.
[0111] Wherein, the lane line width is the lane line width on the point cloud intensity map. In specific implementation, according to formula (1), the lane line width on the point cloud intensity map is the actual lane line width divided by the sampling interval g. Therefore, the side length of the inner neighborhood is equal to the actual lane line width divided by the sampling interval g. The radius of the outer neighborhood has been shown by experiments to be set to 25, which can reflect the surface characteristics to a certain extent.
[0112] S403, based on the first sliding window, slide along the first preset direction on the point cloud intensity map with a first preset step size, and extract the area in the first coverage area under each first preset step size that meets the preset intensity condition, until the overlap between the area slid by the inner neighborhood of the first sliding window and the point cloud intensity map meets the first preset requirement.
[0113] The first coverage area is the region on the point cloud intensity map that overlaps with the first sliding window.
[0114] In this implementation, the first preset step size is set to one pixel.
[0115] The first preset direction may include from left to right or from top to bottom.
[0116] The first preset requirement is the degree of overlap between the area slid by the inner neighborhood of the first sliding window and the point cloud intensity map. In specific implementation, the inner neighborhood of the first sliding window should be made to slide over the entire area of the point cloud intensity map as much as possible, so as to preserve the complete lane line as much as possible. Considering that there may be cases where the inner neighborhood of the first sliding window cannot slide over the entire area of the point cloud intensity map, the first preset requirement is not mandatory.
[0117] Step S403 above may include the following steps:
[0118] Determine the first statistical value of the pixels in the region that overlaps with the inner neighborhood in the first coverage area under the current first preset step size;
[0119] Determine the second statistical value of the pixels in the region that overlaps with the outer neighborhood in the first coverage area under the current first preset step size;
[0120] When the first statistical value and the second statistical value satisfy the preset intensity condition, extract the pixels in the area that overlaps with the inner neighborhood in the first coverage area under the current first preset step size.
[0121] The first statistical value includes the standard deviation and mean intensity of pixels in the region overlapping with the inner neighbor region in the first coverage area, and the second statistical value includes the standard deviation and mean intensity of pixels in the region overlapping with the outer neighbor region in the first coverage area.
[0122] Since point cloud intensity maps differ from color images, some pixels corresponding to lane lines in the point cloud intensity map may have a lot of uneven noise, affecting the overall feature statistics of lane lines. To reduce the impact of noise, the mean of non-zero pixels falling within the inner neighborhood is calculated first, and points with excessive differences from the mean are removed. Then, the mean intensity M1 and standard deviation S1 of non-zero pixels within the inner neighborhood are calculated again. For non-zero pixels falling within the outer neighborhood, after removing outliers within the inner neighborhood, their mean intensity M2, standard deviation S2, and maximum intensity value Ma2 are directly calculated. Through the first sliding window, based on the rules corresponding to formula (3), pixels falling within the inner neighborhood that meet the rules are extracted, thereby filtering out non-lane line areas.
[0123]
[0124] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application combine the differences between intensity maps and color maps to suppress the influence of noise in intensity maps. By loosely limiting multiple conditions, lane line areas can be preserved to a certain extent, preventing excessive lane line information from being filtered out due to strict setting of a certain condition.
[0125] S405, Based on the regions that meet the preset intensity conditions corresponding to each first preset step size, a preliminary point cloud intensity sub-map is generated.
[0126] Among them, the preliminary point cloud intensity sub-map is generated from the point cloud intensity map after only preliminary screening under preset intensity conditions.
[0127] S407, Based on the target preset lane line main direction interval and the preliminary point cloud intensity sub-map, generate the target point cloud intensity sub-map.
[0128] Based on the point cloud intensity map after being filtered by preset intensity conditions, further lane line filtering is performed by setting directional conditions according to the main direction interval of the lane lines.
[0129] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application extract lane line regions from the point cloud intensity map through the first sliding window based on preset intensity conditions, complete the preliminary screening of lane lines, and provide implementation objects for subsequent lane line direction limitation.
[0130] In one exemplary implementation, such as Figure 5 As shown, step S407 above may include:
[0131] S501, based on the target preset lane line main direction interval and the lane line width, determine the second sliding window.
[0132] The second sliding window consists of a first mask and a second mask.
[0133] The first and second masks are squares with sides longer than the lane width. Each mask contains a statistical region whose width is equal to the lane width. The mask and its statistical region share a common center point. In practice, the first and second masks are overlapped to form a second sliding window.
[0134] The determination of the first and second masks needs to be combined with the target preset lane line main direction interval. In specific implementation, such as... Figure 6 The diagram shows a schematic of a lane line detection mask with a defined direction. Before selecting the first and second masks, the main direction of the lane lines in the point cloud intensity map is determined based on the target preset lane line main direction interval. Specifically, the long side of the internal statistical region of the first mask is perpendicular to the main direction of the lane lines, while the long side of the internal statistical region of the second mask is parallel to the main direction of the lane lines. Figure 3 and Figure 6 As shown, interval 1 corresponds to mask a and mask b, interval 2 corresponds to mask c and mask d, interval 3 corresponds to mask b and mask a, and interval 4 corresponds to mask d and mask c.
[0135] Wherein, the lane line width is the lane line width on the point cloud intensity map. In specific implementation, according to formula (1), the lane line width on the point cloud intensity map is the actual lane line width divided by the sampling interval g. Therefore, the width of the internal statistical region of the first mask and the internal statistical region of the second mask is equal to the actual lane line width divided by the sampling interval g. The side length of the first mask and the second mask can be set to 25 according to the experiment.
[0136] S503, based on the second sliding window, slide along the second preset direction with the second preset step size on the preliminary point cloud intensity sub-map, extract the area in the second coverage area under each second preset step size that meets the preset direction condition, until the overlap between the target area slid by the second sliding window and the point cloud intensity map meets the second preset requirement.
[0137] The second coverage area is the region on the initial point cloud intensity submap that overlaps with the second sliding window.
[0138] In practice, the second preset step size is set to one pixel.
[0139] The second preset direction can include from left to right and from top to bottom.
[0140] The second preset requirement is the degree of overlap between the area traversed by the internal statistical region of the second mask and the point cloud intensity map. The target area traversed by the second sliding window is the area traversed by the internal statistical region of the second mask. In specific implementation, the goal is to ensure that the internal statistical region of the second mask traverses the entire area of the point cloud intensity map, thereby preserving as many lane lines as possible. Considering the possibility that the internal statistical region of the second mask may not traverse the entire area of the point cloud intensity map, this second preset requirement is not mandatory.
[0141] Step S503 above may include the following steps:
[0142] Determine the number of pixels in the second coverage area at the current second preset step size whose pixel values are greater than a preset pixel value in the area that overlaps with the internal statistical area of the first mask, and obtain the first quantity value;
[0143] Determine the number of pixels in the second coverage area that overlaps with the internal statistical area of the second mask under the current second preset step size whose pixel value is greater than a preset pixel value, and obtain the second quantity value;
[0144] When the ratio of the first quantity value to the second quantity value is within a preset range, extract the pixels in the area that overlaps with the internal statistical area of the second mask in the second coverage area under the current second preset step size.
[0145] The first quantity value is the number of pixels with a pixel value greater than a preset pixel value in the internal statistical region of the first mask, and the second quantity value is the number of pixels with a pixel value greater than a preset pixel value in the internal statistical region of the second mask.
[0146] The preset threshold is the minimum threshold of the first and second quantity values that need to be set in advance. When setting it, the maximum value of the pixel under the mask can be referenced.
[0147] For the preset interval, the maximum and minimum values of the ratio of the first quantity value and the second quantity value need to be preset, and the preset interval is composed of the maximum value, the minimum value, and all real numbers between the maximum value and the minimum value.
[0148] In specific implementation, such as Figure 3 and Figure 6As shown, taking masks a and b corresponding to interval 1 as examples, the number of pixel values greater than a preset threshold under masks a and b are counted as the first quantity value Ca and the second quantity value Cb, respectively. The preset threshold is set to 0.12 times the maximum value of all non-zero pixel points within the mask to eliminate noise and the influence of uneven reflection from the lidar. Through the second sliding window, based on the rules corresponding to formula (4), the pixel points that meet the rules and fall within the statistical area inside mask b are extracted, thereby filtering out non-lane line areas.
[0149]
[0150] The maximum and minimum value ranges of this formula are loosely defined, and can be amplified or reduced to a certain extent to ensure the integrity of the lane lines after filtering.
[0151] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application limit the direction of lane lines by dividing the lane lines into intervals. While limiting the linear characteristics of lane lines, they can also adapt to lane line detection with a small curvature, and have a wider range of applicable scenarios compared to general straight line detection.
[0152] S505, Based on the regions that satisfy the preset direction conditions corresponding to each second preset step size, generate the target point cloud intensity sub-map.
[0153] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application, through the second sliding window, on the basis of the point cloud intensity map being filtered by preset intensity conditions, set directional conditions according to the main direction interval of the lane lines, and complete the final filtering of lane lines on the point cloud intensity map.
[0154] S107, Extract the target point cloud elevation sub-map from the point cloud elevation map; The elevation values of the pixels in the target point cloud elevation sub-map meet the preset elevation conditions.
[0155] The preset elevation conditions are numerical requirements set based on the distribution of pixels in the point cloud elevation map, while the target point cloud elevation sub-map is an image composed of pixels within the area that meets the above requirements.
[0156] Filter out pixels in the point cloud elevation map that are not in lane lines, and filter out pixels that have large elevation jumps in a small area.
[0157] In one exemplary implementation, such as Figure 7 As shown, step S107 above may include:
[0158] S701, determine the third sliding window based on the lane line width.
[0159] The third sliding window is a square with a side length equal to the width of the lane line.
[0160] S703, based on the third sliding window, slide along the third preset direction on the point cloud elevation map with a third preset step size, and extract the third coverage area that meets the preset elevation conditions under each third preset step size, until the overlap between the area slid by the third sliding window and the point cloud intensity map meets the third preset requirements.
[0161] The preset elevation condition is that the standard deviation of the pixels in the third coverage area is less than a preset threshold.
[0162] The third coverage area is the region on the point cloud elevation map that overlaps with the third sliding window.
[0163] The preset threshold is a limitation on the distribution of pixels in the third coverage area. Since the road surface where the lane line is located will have a large elevation jump in a small area, the preset threshold is 0.01 in this embodiment. The standard deviation Se of non-zero pixels falling in the third sliding window in the point cloud elevation map is calculated. Through the third sliding window, based on the corresponding rule of formula (5), the pixels that meet the rule and fall in the third sliding window are extracted, thereby filtering out non-lane line areas.
[0164] Se < 0.01 (5)
[0165] In practice, the third preset step size is set to one pixel.
[0166] The third preset direction can include from left to right and from top to bottom.
[0167] The third preset requirement is the degree of overlap between the area slid by the third sliding window and the point cloud elevation map. In specific implementation, the third sliding window should be made to slide over the entire area of the point cloud elevation map as much as possible, so as to preserve the complete lane lines as much as possible. Considering that there may be cases where the third sliding window cannot slide over the entire area of the point cloud elevation map, the third preset requirement is not mandatory.
[0168] S705, Based on the third coverage area that meets the preset elevation conditions corresponding to each third preset step size, generate the target point cloud elevation sub-map.
[0169] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application can obtain a more accurate lane line segmentation map by setting lane line extraction rules for point cloud elevation maps, filtering out non-lane line areas, and combining the lane lines extracted from the point cloud intensity map.
[0170] S109, determine the lane line segmentation map based on the target point cloud intensity sub-map and the target point cloud elevation sub-map, and determine the lane line vectorization result based on the coordinate information of the pixels in the lane line segmentation map.
[0171] The lane line segmentation map is a two-dimensional image composed of pixels within a region that simultaneously meets preset intensity, direction, and elevation conditions. Reconstructing the two-dimensional lane line segmentation map with defined directions into a three-dimensional vectorized lane line result can be used for lane line feature representation in high-precision maps.
[0172] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application estimate the approximate direction of lane lines in the intensity map by using the trajectory points corresponding to the point cloud, design corresponding intensity distribution rules, and combine other intensity and elevation features of the lane lines to filter out lane line regions. This method can not only solve the problem of intensity maps lacking fixed lane line directions, but also remove vegetation on both sides of the lane and impurity spots on the lane through multiple condition constraints, resulting in more accurate lane line regions for subsequent lane line vectorization work.
[0173] In one exemplary implementation, such as Figure 8 As shown, step S109 above may include:
[0174] S801, determine the overlapping pixels of the target point cloud intensity sub-map and the target point cloud elevation sub-map.
[0175] S803, Generate the lane line segmentation map based on the overlapping pixels.
[0176] The intersection operation is performed on the point cloud intensity map that has been limited by intensity features and direction, and the point cloud elevation map that has been limited by elevation features. The resulting lane line segmentation map is a more accurate lane line region after triple filtering.
[0177] S805, determine the centerline of each lane line in the lane line segmentation diagram.
[0178] By extracting the skeleton from the lane line segmentation map, a rough lane centerline can be obtained.
[0179] S807, the coordinates of the pixels on the center line are restored to three-dimensional coordinates to obtain the vectorized lane line result.
[0180] The two-dimensional pixel coordinates can be converted into three-dimensional coordinates using formula (1).
[0181] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application extract the skeleton of a more accurate lane line segmentation map and restore it to a three-dimensional coordinate system to obtain the lane line vectorization result, which can assist manual mapping and greatly improve the efficiency of high-precision map making.
[0182] Corresponding to the lane line detection methods provided in the above embodiments, this application also provides a lane line detection device. Since the lane line detection device provided in this application corresponds to the lane line detection methods provided in the above embodiments, the implementation methods of the aforementioned lane line detection methods are also applicable to the lane line detection device provided in this embodiment, and will not be described in detail in this embodiment.
[0183] Please see Figure 9 The diagram shown is a structural schematic of a lane line detection device provided in an embodiment of this application. This device has the function of implementing the lane line detection method described in the above-described method embodiments. This function can be implemented in hardware or by hardware executing corresponding software. Figure 9 As shown, the device may include:
[0184] The image acquisition module is used to acquire the point cloud intensity map and the point cloud elevation map corresponding to the area to be detected; the point cloud intensity map and the point cloud elevation map are obtained based on the ground point cloud data within the area to be detected collected by the lidar on the acquisition vehicle;
[0185] The main direction determination module is used to determine the target preset lane line main direction interval from multiple preset lane line main direction intervals based on the driving trajectory points of the collected vehicle.
[0186] The intensity map extraction module is used to extract a target point cloud intensity sub-map that meets the preset intensity conditions from the point cloud intensity map based on the main direction interval of the target preset lane line.
[0187] The elevation map extraction module is used to extract a target point cloud elevation sub-map from the point cloud elevation map; the elevation values of the pixels in the target point cloud elevation sub-map meet preset elevation conditions;
[0188] The lane line determination module is used to determine a lane line segmentation map based on the target point cloud intensity sub-map and the target point cloud elevation sub-map, and to determine the lane line vectorization result based on the coordinate information of the pixels in the lane line segmentation map.
[0189] In one exemplary embodiment, the apparatus further includes an image creation module for generating point cloud intensity maps and point cloud elevation maps, the image creation module comprising:
[0190] The point cloud acquisition module is used to acquire ground point cloud data within the detection range collected by the lidar;
[0191] The sample extraction module is used to extract multiple sample 3D points from the ground point cloud data based on a preset sampling interval;
[0192] The sample 2D conversion module is used to project the multiple sample 3D points onto the first 2D image to obtain the first projected image;
[0193] A region segmentation module is used to divide the first projected image into multiple sampling regions;
[0194] The pixel value determination module is used to determine the average intensity of pixels in each sampling region and the average elevation of pixels in each sampling region.
[0195] The image generation module is used to generate a point cloud intensity map based on the average intensity value corresponding to each sampling region, and to generate a point cloud elevation map based on the average elevation value corresponding to each sampling region.
[0196] In one exemplary embodiment, the plurality of preset lane line main direction intervals are a preset number of equal-angle intervals, and the main direction determination module includes:
[0197] The trajectory point 2D conversion module is used to project the driving trajectory points onto a second 2D image to obtain 2D trajectory points;
[0198] The trajectory point fitting module is used to perform straight line fitting on the two-dimensional trajectory points to obtain the target trajectory straight line;
[0199] Angle determination module is used to determine the absolute value of the angle between the target trajectory line and the positive direction of the horizontal axis of the second two-dimensional image, thereby obtaining the target trajectory angle;
[0200] The main direction interval determination module is used to determine the target isoangular interval that matches the angle between the preset number of isoangular intervals and the target trajectory, and to use the target isoangular interval as the main direction interval of the target preset lane line.
[0201] In one exemplary embodiment, the intensity map extraction module includes:
[0202] The first sliding window determination module is used to determine the first sliding window based on the lane line width; the first sliding window consists of an inner neighborhood and an outer neighborhood on the same axis, the inner neighborhood is a square with a side length equal to the lane line width, and the outer neighborhood is a square ring with an inner side coinciding with the outer side of the inner neighborhood and an outer side length greater than the lane line width;
[0203] The first region extraction module is used to slide the first sliding window along the first preset direction on the point cloud intensity map with a first preset step size, and extract the region in the first coverage area that meets the preset intensity condition at each first preset step size, until the overlap between the area slid by the inner neighborhood of the first sliding window and the point cloud intensity map meets the first preset requirement; the first coverage area is the area on the point cloud intensity map that overlaps with the first sliding window.
[0204] The first sub-image generation module is used to generate a preliminary point cloud intensity sub-image based on the regions that meet the preset intensity conditions corresponding to each first preset step size.
[0205] The second sub-map generation module is used to generate the target point cloud intensity sub-map based on the target preset lane line main direction interval and the preliminary point cloud intensity sub-map.
[0206] In one exemplary embodiment, the first region extraction module includes:
[0207] The first pixel point statistics module is used to determine the first statistical value of the pixels in the area that overlaps with the inner neighborhood in the first coverage area under the current first preset step size;
[0208] The second pixel point statistics module is used to determine the second statistical value of the pixels in the area that overlaps with the outer neighborhood in the first coverage area under the current first preset step size;
[0209] The first pixel extraction module is used to extract pixels in the area that overlaps with the inner neighborhood in the first coverage area under the current first preset step size when the first statistical value and the second statistical value satisfy the preset intensity condition.
[0210] In one exemplary embodiment, the second subgraph generation module includes:
[0211] The second sliding window determination module is used to determine a second sliding window based on the target preset lane line main direction interval and the lane line width; the second sliding window is composed of a first mask and a second mask;
[0212] The second region extraction module is used to slide the second sliding window along the second preset direction on the preliminary point cloud intensity sub-map at a second preset step size, and extract the region in the second coverage area that meets the preset direction condition under each second preset step size, until the overlap between the target area slid by the second sliding window and the point cloud intensity map meets the second preset requirement; the second coverage area is the region on the preliminary point cloud intensity sub-map that overlaps with the second sliding window;
[0213] The third sub-graph generation module is used to generate the target point cloud intensity sub-graph based on the regions that meet the preset direction conditions corresponding to each second preset step size.
[0214] In one exemplary embodiment, the second region extraction module includes:
[0215] The third pixel point statistics module is used to determine the number of pixels with pixel values greater than a preset pixel value in the second coverage area that overlaps with the internal statistics area of the first mask under the current second preset step size, and to obtain the first quantity value;
[0216] The fourth pixel point statistics module is used to determine the number of pixels in the second coverage area that overlaps with the internal statistics area of the second mask under the current second preset step size, whose pixel value is greater than a preset pixel value, and obtain the second quantity value;
[0217] The second pixel extraction module is used to extract pixels in the area that overlaps with the internal statistical area of the second mask in the second coverage area under the current second preset step size when the ratio of the first quantity value and the second quantity value is within a preset range.
[0218] In one exemplary embodiment, the elevation map extraction module includes:
[0219] The third sliding window determination module is used to determine a third sliding window based on the lane line width; the third sliding window is a square with a side length equal to the lane line width;
[0220] The third region extraction module slides the third sliding window along the third preset direction on the point cloud elevation map at a third preset step size, based on the third sliding window, to extract the third coverage region that meets the preset elevation condition at each third preset step size, until the overlap between the area traversed by the third sliding window and the point cloud intensity map meets the third preset requirement; the third coverage region is the region on the point cloud elevation map that overlaps with the third sliding window; the preset elevation condition is that the standard deviation of the pixels in the third coverage region is less than a preset threshold.
[0221] The fourth sub-map generation module is used to generate the target point cloud elevation sub-map based on the third coverage area that meets the preset elevation conditions corresponding to each third preset step size.
[0222] In one exemplary embodiment, the lane line determination module includes:
[0223] A pixel determination module is used to determine the overlapping pixels of the target point cloud intensity sub-image and the target point cloud elevation sub-image;
[0224] A segmentation map generation module is used to generate the lane line segmentation map based on the overlapping pixels;
[0225] A centerline determination module is used to determine the centerline of each lane line in the lane line segmentation diagram;
[0226] The pixel restoration module is used to restore the coordinates of the pixels on the center line to three-dimensional coordinates to obtain the lane line vectorization result.
[0227] It should be noted that the apparatus provided in the above embodiments is only illustrated by the division of the above functional modules when implementing its functions. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.
[0228] This application provides an electronic device including a processor and a memory. The memory stores at least one instruction or at least one program, which is loaded and executed by the processor to implement any of the lane detection methods provided in the above method embodiments.
[0229] Memory can be used to store software programs and modules. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory. Memory can primarily include a program storage area and a data storage area. The program storage area can store the operating system, application programs required for the functions, etc.; the data storage area can store data created based on the use of the device, etc. Furthermore, memory can include high-speed random access memory, and can also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, memory can also include a memory controller to provide the processor with access to the memory.
[0230] The method embodiments provided in this application can be executed in a computer terminal, server or similar computing device, that is, the above-mentioned electronic device may include a computer terminal, server or similar computing device. Figure 10 This is a hardware structure block diagram of an electronic device that runs a lane line detection method according to an embodiment of this application, such as... Figure 10 As shown, the internal structure of this computer device may include, but is not limited to, a processor, a network interface, and a memory. The processor, network interface, and memory within the computer device can be connected via a bus or other means, as illustrated in the embodiments of this specification. Figure 10 Taking the example of a connection between China and Israel via a bus.
[0231] The processor (or CPU, Central Processing Unit) is the computing and control core of the computer device. The network interface may optionally include a standard wired interface or a wireless interface (such as Wi-Fi, mobile communication interface, etc.). Memory is the storage device in the computer device used to store programs and data. It is understood that the memory here can be a high-speed RAM storage device, or a non-volatile storage device, such as at least one disk storage device; optionally, it can also be at least one storage device located remotely from the aforementioned processor. The memory provides storage space, which stores the operating system of the electronic device, including but not limited to: Windows (an operating system), Linux (an operating system), Android (a mobile operating system), iOS (a mobile operating system), etc., which are not limited in this invention; and the storage space also stores one or more instructions suitable for being loaded and executed by the processor, which can be one or more computer programs (including program code). In the embodiments of this specification, the processor loads and executes one or more instructions stored in the memory to implement the lane detection method provided in the above method embodiments.
[0232] Embodiments of this application also provide a computer-readable storage medium, which can be disposed in an electronic device to store at least one instruction or at least one program related to implementing a lane line detection method. The at least one instruction or the at least one program is loaded and executed by the processor to implement any of the lane line detection methods provided in the above-described method embodiments.
[0233] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.
[0234] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0235] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.
[0236] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.
[0237] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.
Claims
1. A lane line detection method, characterized in that, The method includes: Obtain the point cloud intensity map and point cloud elevation map corresponding to the area to be detected; the point cloud intensity map and the point cloud elevation map are obtained based on the ground point cloud data within the area to be detected collected by the lidar on the acquisition vehicle; Based on the collected vehicle trajectory points, the target preset lane line main direction interval is determined from multiple preset lane line main direction intervals; Based on the main direction interval of the target preset lane line, extract the target point cloud intensity sub-map that meets the preset intensity condition from the point cloud intensity map; Extract a target point cloud elevation sub-map from the point cloud elevation map; the elevation values of the pixels in the target point cloud elevation sub-map meet preset elevation conditions; A lane line segmentation map is determined based on the target point cloud intensity submap and the target point cloud elevation submap, and the lane line vectorization result is determined based on the coordinate information of the pixels in the lane line segmentation map. The step of extracting a target point cloud intensity sub-map that meets preset intensity conditions from the point cloud intensity map based on the main direction interval of the target preset lane line includes: determining a first sliding window based on the lane line width; the first sliding window consists of an inner neighborhood and an outer neighborhood on the same axis, the inner neighborhood being a square with a side length equal to the lane line width, and the outer neighborhood being a square ring with its inner side coinciding with the outer side of the inner neighborhood and its outer side length being greater than the lane line width; sliding the first sliding window on the point cloud intensity map to extract the region that meets the preset intensity conditions and generate a preliminary point cloud intensity sub-map; and generating the target point cloud intensity sub-map based on the main direction interval of the target preset lane line and the preliminary point cloud intensity sub-map.
2. The lane line detection method according to claim 1, characterized in that, The method further includes: Acquire ground point cloud data within the detection range collected by lidar; Multiple sample 3D points are extracted from the ground point cloud data based on a preset sampling interval; The multiple sample 3D points are projected onto the first 2D image to obtain the first projected image; The first projected image is divided into multiple sampling regions; Determine the mean intensity of pixels in each sampling region, and the mean elevation of pixels in each sampling region; A point cloud intensity map is obtained based on the average intensity value corresponding to each sampling area, and a point cloud elevation map is obtained based on the average elevation value corresponding to each sampling area.
3. The lane line detection method according to claim 1, characterized in that, The main direction intervals of the multiple preset lane lines are a preset number of equal angle intervals; The step of determining the target preset lane main direction interval from multiple preset lane main direction intervals based on the collected vehicle's driving trajectory points includes: The driving trajectory points are projected onto the second two-dimensional image to obtain two-dimensional trajectory points; By performing linear fitting on the two-dimensional trajectory points, the target trajectory straight line is obtained; The absolute value of the angle between the target trajectory line and the positive horizontal axis of the second two-dimensional image is determined to obtain the target trajectory angle; Determine the target isoangular intervals that match the angle between the target trajectory and the preset number of isoangular intervals, and use the target isoangular intervals as the main direction intervals of the target preset lane lines.
4. The lane line detection method according to claim 1, characterized in that, The step of sliding the first sliding window on the point cloud intensity map to extract regions that meet the preset intensity conditions and generate a preliminary point cloud intensity sub-map includes: Based on the first sliding window, slide along the first preset direction on the point cloud intensity map with a first preset step size, and extract the area in the first coverage area that meets the preset intensity condition under each first preset step size, until the overlap between the area slid by the inner neighborhood of the first sliding window and the point cloud intensity map meets the first preset requirement; the first coverage area is the area on the point cloud intensity map that overlaps with the first sliding window. Based on the regions that meet the preset intensity conditions corresponding to each first preset step size, the preliminary point cloud intensity sub-map is generated.
5. The lane line detection method according to claim 4, characterized in that, The extraction of regions within the first coverage area that satisfy the preset intensity condition at each first preset step size includes: Determine the first statistical value of the pixels in the region that overlaps with the inner neighborhood in the first coverage area under the current first preset step size; Determine the second statistical value of the pixels in the region that overlaps with the outer neighborhood in the first coverage area under the current first preset step size; When the first statistical value and the second statistical value satisfy the preset intensity condition, extract the pixels in the area that overlaps with the inner neighborhood in the first coverage area under the current first preset step size.
6. The lane line detection method according to claim 4, characterized in that, The step of generating the target point cloud intensity sub-map based on the target preset lane line main direction interval and the preliminary point cloud intensity sub-map includes: Based on the target preset lane line main direction interval and the lane line width, a second sliding window is determined; the second sliding window is composed of a first mask and a second mask; Based on the second sliding window, slide along the second preset direction with a second preset step size on the preliminary point cloud intensity sub-map, and extract the region in the second coverage area under each second preset step size that meets the preset direction condition, until the overlap between the target area slid by the second sliding window and the point cloud intensity map meets the second preset requirement; the second coverage area is the region on the preliminary point cloud intensity sub-map that overlaps with the second sliding window; Based on the regions that satisfy the preset directional conditions corresponding to each second preset step size, the target point cloud intensity sub-map is generated.
7. The lane line detection method according to claim 6, characterized in that, The extraction of regions within the second coverage area that satisfy the preset direction condition under each second preset step size includes: Determine the number of pixels in the second coverage area at the current second preset step size whose pixel values are greater than a preset pixel value in the area that overlaps with the internal statistical area of the first mask, and obtain the first quantity value; Determine the number of pixels in the second coverage area that overlaps with the internal statistical area of the second mask under the current second preset step size whose pixel value is greater than a preset pixel value, and obtain the second quantity value; When the ratio of the first quantity value to the second quantity value is within a preset range, extract the pixels in the area that overlaps with the internal statistical area of the second mask in the second coverage area under the current second preset step size.
8. The lane line detection method according to claim 1, characterized in that, Extracting the target point cloud elevation sub-map from the point cloud elevation map includes: A third sliding window is determined based on the lane line width; the third sliding window is a square with a side length equal to the lane line width. Based on the third sliding window, slide along the third preset direction on the point cloud elevation map with a third preset step size, and extract the third coverage area that meets the preset elevation condition at each third preset step size, until the overlap between the area slid by the third sliding window and the point cloud intensity map meets the third preset requirement; the third coverage area is the area on the point cloud elevation map that overlaps with the third sliding window; the preset elevation condition is that the standard deviation of the pixels in the third coverage area is less than a preset threshold. Based on the third coverage area that meets the preset elevation conditions corresponding to each third preset step size, the target point cloud elevation sub-map is generated.
9. The lane line detection method according to any one of claims 1 to 8, characterized in that, The process of determining a lane line segmentation map based on the target point cloud intensity sub-map and the target point cloud elevation sub-map, and determining the lane line vectorization result based on the coordinate information of pixels in the lane line segmentation map, includes: Identify the overlapping pixels of the target point cloud intensity sub-map and the target point cloud elevation sub-map; The lane line segmentation map is generated based on the overlapping pixels; Determine the centerline of each lane line in the lane line segmentation diagram; The coordinates of the pixels on the center line are restored to three-dimensional coordinates to obtain the vectorized lane line result.
10. A lane line detection device, characterized in that, The device includes: The image acquisition module is used to acquire the point cloud intensity map and the point cloud elevation map corresponding to the area to be detected; the point cloud intensity map and the point cloud elevation map are obtained based on the ground point cloud data within the area to be detected collected by the lidar on the acquisition vehicle; The main direction determination module is used to determine the target preset lane line main direction interval from multiple preset lane line main direction intervals based on the driving trajectory points of the collected vehicle. The intensity map extraction module is used to extract a target point cloud intensity sub-map that meets the preset intensity conditions from the point cloud intensity map based on the main direction interval of the target preset lane line. The elevation map extraction module is used to extract a target point cloud elevation sub-map from the point cloud elevation map; the elevation values of the pixels in the target point cloud elevation sub-map meet preset elevation conditions; The lane line determination module is used to determine a lane line segmentation map based on the target point cloud intensity sub-map and the target point cloud elevation sub-map, and to determine the lane line vectorization result based on the coordinate information of the pixels in the lane line segmentation map. The step of extracting a target point cloud intensity sub-map that meets preset intensity conditions from the point cloud intensity map based on the main direction interval of the target preset lane line includes: determining a first sliding window based on the lane line width; the first sliding window consists of an inner neighborhood and an outer neighborhood on the same axis, the inner neighborhood being a square with a side length equal to the lane line width, and the outer neighborhood being a square ring with its inner side coinciding with the outer side of the inner neighborhood and its outer side length being greater than the lane line width; sliding the first sliding window on the point cloud intensity map to extract the region that meets the preset intensity conditions and generate a preliminary point cloud intensity sub-map; and generating the target point cloud intensity sub-map based on the main direction interval of the target preset lane line and the preliminary point cloud intensity sub-map.
Citation Information
Patent Citations
Laser point cloud lane line extraction method and electronic equipment
CN111783721A
Lane line detection method and device
CN111832536A