Millimeter wave radar original point cloud-based lane line extraction method and computer equipment

Abstract

The invention discloses a millimeter wave radar original point cloud-based lane line extraction method and computer equipment, and the method comprises the steps: S1, obtaining original point cloud data through a millimeter wave radar, and carrying out the preprocessing of the original point cloud data, and forming target point cloud data; s2, constructing an occupied grid map, performing statistical calculation on the target point cloud data, and performing coarse extraction of lane lines; s3, determining the number of road lanes and the positions of the road lanes by using a CFAR algorithm, and obtaining a plurality of detection peak values; and S4, detecting a plurality of grids where peak values are located according to a CFAR algorithm, performing point cloud coordinate statistics on the target point cloud data, and performing fine extraction of lane lines through Gaussian distribution fitting. According to the invention, accurate lane line information can be extracted, the method is applied to the field of intelligent traffic, and important supplement is provided for construction of a lane-level high-precision map and holographic traffic perception.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method for extracting lane lines based on raw point clouds of millimeter-wave radar and a computer device. Background Art

[0002] With the economic development, the number of urban vehicles has been increasing in recent years. The demand for urban road traffic has increased significantly, and the continuous intensification of the urban road load restricts the development of the city. The intelligent transportation system is gradually becoming an effective method to solve urban traffic problems. The intelligent transportation system is a comprehensive information system integrating technologies such as detection, recognition, control, communication, and computer. The cornerstone and key of the construction of intelligent transportation is the acquisition of traffic information. Lane lines are important features of the traffic environment and the basis for a series of traffic applications. After obtaining the lane lines, vehicle lane-level positioning can be realized, and traffic safety management technologies such as accident warning and traffic monitoring can be achieved. Therefore, lane line detection is one of the key issues in the construction of intelligent transportation.

[0003] In the prior art, the lane line detection method based on millimeter-wave radar mainly extracts lane lines based on the center line of the existing map by using the density of points of the vehicle trajectory in the grid map, or performs overall clustering on the target trajectory to extract lane lines. However, these research methods simply fit lane lines based on the center line of the existing map, which will affect the efficiency and accuracy of the subsequent extraction and update of lane lines. Moreover, they all extract lane lines based on the vehicle trajectory obtained by processing the raw point cloud of the radar through complex steps such as the target tracking module, resulting in a large computational consumption. At the same time, the number of lanes in the road scene needs to be manually input, and the manual workload is also large. Summary of the Invention

[0004] The present invention provides a method for extracting lane lines based on raw point clouds of millimeter-wave radar and a computer device that is more accurate, efficient, and requires less computational and workload input, which can at least solve one of the above technical problems.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions: A method for extracting lane lines based on raw point clouds of millimeter-wave radar includes the following steps: S1. The millimeter-wave radar acquires raw point cloud data and preprocesses the raw point cloud data to form target point cloud data; S2. A occupancy grid map is constructed, and statistical calculations are performed on the target point cloud data to perform rough extraction of lane lines; S3. The CFAR algorithm is used to determine the number of road lanes and the positions of road lanes, and multiple detection peaks are obtained; S4. Detect multiple grids where the peak is located according to the CFAR algorithm, perform point cloud coordinate statistics on the target point cloud data, and perform fine extraction of the lane line through Gaussian distribution fitting.

[0006] Further, in S1, the millimeter-wave radar uses a transmitting antenna to transmit a frequency-modulated continuous wave signal into the front space to cover multiple lane areas. The signal beam is reflected when encountering a target or obstacle during transmission, generating an echo signal. The receiving antenna receives this echo signal and calculates it through the signal processing system of the built-in chip of the millimeter-wave radar to obtain multi-dimensional information of the detection points in the detected road area, that is, the original point cloud data. The original point cloud data at least includes radial distance, azimuth angle, height, radial velocity, and signal strength.

[0007] Further, in S1, with the position center of the millimeter-wave radar as the origin, according to the actual road environment, the detection range is divided into a rectangular area, and the effective targets are screened by restricting the longitudinal and lateral ranges of the millimeter-wave radar targets to clean the original point cloud data and form the target point cloud data. It further includes: S11. Delete the data in the original point cloud data that does not conform to the range of the detected road and the data that does not meet the requirement of the maximum detection distance that the millimeter-wave radar can detect; S12. Eliminate some noise points with small power and echo energy according to the road conditions, as well as the static target point cloud with a speed of 0; S13. Eliminate the point traces with few data items along the road direction and the false ghost point traces.

[0008] Further, S2 further includes: S21. Construct an occupancy grid map as a representation of road traffic information. The color of each grid in the occupancy grid map represents the probability of being occupied by that grid, that is, the number of times the point cloud of different target vehicles occupies. The darker the color, the more times the point cloud of the passing vehicle occupies. Therefore, each grid can be represented by a numerical value. The higher the numerical value, the more likely it is to be the lane center area. At the same time, the occupied grid also records the coordinate information of each point trace; S22. Divide the road environment where the vehicle travels in the detection area of the millimeter-wave radar with grids. The smaller the grid size, the more the number, and the more accurate the lane line information contained; S23. Perform statistics on the target point cloud data after preprocessing, calculate the number of point clouds falling on each grid respectively, and record the coordinate of each point cloud falling on the grid. After the cumulative statistics of the road vehicle point cloud in a period of time frame, an occupancy grid map containing the current road traffic information is obtained.

[0009] Further, S3 further includes: S31. Perform data statistics on the abscissa of the statistical point cloud of all grids in each row of the occupied grid map longitudinally. The independent variable is the lateral coordinate of the vehicle position, and the dependent variable is the number of abscissa values of the vehicle point cloud positions detected by the millimeter-wave radar. S32. Each row of grids includes multiple peaks, and each peak corresponds to a lane, which is consistent with the actual quantity. The lane peak quantity and the reliable position of the lane are detected through the CFAR algorithm. S33. For the center position of a lane, multiple peaks will be generated after detection by the CFAR algorithm. It is set that starting from the first detected peak in the CFAR algorithm detection peak set, the next detected peak whose lateral distance from the first peak is less than 3m belongs to the same lane peak set, and those greater than 3m are the first detected peaks of the next lane. Thus, the reliable lane quantity of the occupied grids in this row and the positions of the grids where the center of each lane is located are obtained.

[0010] Further, the S4 further includes: S41. After the reliable lane quantity and lane positions are detected by the CFAR algorithm, select multiple grids where the CFAR algorithm detection peaks are located for point cloud coordinate statistics. Since the vehicle point clouds on each lane are mostly distributed in the center of the lane and a small number are distributed on both sides of the lane, which conforms to the Gaussian distribution condition, the Gaussian distribution is used to fit the point cloud distribution data. The Gaussian data distribution model formula is as follows:

[0011] Among them, μ represents the mean value, and σ represents the standard deviation; S42. Solve the Gaussian data distribution model of each lane, and obtain the accurate center position of the lane from the Gaussian distribution center of each lane. S43. Repeatedly extract and fuse each row of grids longitudinally to obtain continuous road lane lines in the longitudinal direction of the road.

[0012] A computer device, including a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes the steps of the above-mentioned lane line extraction method based on the original point cloud of the millimeter-wave radar.

[0013] The beneficial effects of the present invention are reflected in: 1. Low computational effort input: This method only analyzes, statistics, and extracts the road lane lines on the original point cloud layer of the millimeter-wave radar, without the need for complex processing of steps such as the target tracking module, which can save many processing steps and reduce resource consumption.

[0014] 2. Real-time and fast response: Since the millimeter-wave radar can provide continuous point cloud data sampling frames at the millisecond level, and the grid map can be processed based on real-time data, this method can quickly accumulate lane line information on the radar-detected road in a short time and can respond in real time quickly.

[0015] 3. High accuracy: This method uses the CFAR detection algorithm to detect the point cloud distribution in the grid map, which can accurately detect the number of lanes in the radar-detected road without the need for artificial input of the number of lanes in the road scene in advance. At the same time, it also provides more reliable and accurate lane number and position data for the subsequent fine extraction of lane lines based on Gaussian distribution fitting, improving the accuracy of the subsequent fine extraction of lane lines.

[0016] 4. Can better identify some curved sections: The lane lines extracted by this method can be fused longitudinally in the grid map. The smaller the longitudinal length of the grid, the more accurate the lane lines for curved sections. Description of the Drawings

[0017] The drawings described herein are used to provide a further understanding of the present application and form a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application.

[0018] Figure 1 It is a schematic flow chart of the lane line extraction method based on the original point cloud of the millimeter-wave radar in the embodiment of the present invention.

[0019] Figure 2 It is a flow block diagram of the lane line extraction method based on the original point cloud of the millimeter-wave radar in the embodiment of the present invention.

[0020] Figure 3 It is a scatter diagram of the position of the traffic target point cloud data detected by the millimeter-wave radar in the embodiment of the present invention.

[0021] Figure 4 It is a schematic diagram of the point cloud statistical distribution of all grids in a certain row in the embodiment of the present invention.

[0022] Figure 5 It is a schematic flow chart of the CFAR algorithm processing in the embodiment of the present invention.

[0023] Figure 6 It is a schematic diagram of the CFAR algorithm detection result of the vehicle target trajectory statistics with the radial distance [120, 130] in the embodiment of the present invention.

[0024] Figure 7 It is a block diagram of the structure of the computer device in the embodiment of the present invention. Detailed Embodiments

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0026] It should be noted that the meaning of "and / or" appearing throughout the text includes three parallel solutions. Taking "A and / or B" as an example, it includes solution A, or solution B, or a solution where A and B are satisfied simultaneously. In addition, "a plurality of" means two or more. In addition, the technical solutions between the embodiments can be combined with each other, but it must be based on the fact that those of ordinary skill in the art can implement it. When the combination of technical solutions appears to be contradictory or unable to be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0027] See Figures 1-6 , an embodiment of the present invention provides a lane line extraction method based on millimeter-wave radar raw point cloud, including the following steps: S1. The millimeter-wave radar acquires raw point cloud data and preprocesses the raw point cloud data to form target point cloud data; S2. Construct an occupancy grid map, perform statistical calculations on the target point cloud data, and perform rough extraction of lane lines; S3. Use the CFAR algorithm to determine the number and position of road lanes and obtain multiple detection peaks; S4. According to the multiple grids where the detection peaks of the CFAR algorithm are located, perform point cloud coordinate statistics on the target point cloud data, and perform fine extraction of lane lines through Gaussian distribution fitting.

[0028] The present invention provides a lane line recognition and extraction method based on millimeter-wave radar raw point cloud. Without other additional detection information, this method only relies on the raw point cloud data provided by the radar and does not require complex processing steps such as target tracking and association. A dual-lane line precise extraction method based on an occupancy grid map and Gaussian distribution fitting is established, which can detect lane lines under different external environmental conditions. Even if the road lane lines are worn and unclear, this method can still perform precise extraction of lane lines.

[0029] In this embodiment, in S1, the millimeter-wave radar uses a transmitting antenna to transmit a frequency-modulated continuous-wave signal into the forward space to cover multiple lane areas. When the signal beam encounters a target or an obstacle during transmission, it is reflected to generate an echo signal. The receiving antenna receives this echo signal and calculates it through the signal processing system of the built-in chip of the millimeter-wave radar to obtain multi-dimensional information of the detection points in the detected road area, that is, the original point cloud data. The original point cloud data includes at least radial distance, azimuth angle, height, radial velocity, and signal strength.

[0030] In this embodiment, in S1, with the position center of the millimeter-wave radar as the origin, according to the actual road environment, the detection range is divided into a rectangular area, and the effective targets are screened by restricting the longitudinal and lateral ranges of the millimeter-wave radar targets to clean the original point cloud data and form the target point cloud data. It further includes: S11. Delete the data in the original point cloud data that does not meet the range of the detected road and the data that does not meet the requirement of the maximum detection distance that the millimeter-wave radar can detect; S12. Eliminate some noise points with small power and echo energy, and static target point clouds with a speed of 0 according to the road conditions; S13. Eliminate the point traces with few data items in the direction of the along-road and the false ghost point traces.

[0031] In addition to containing the measurements of effective targets, the original point cloud data also contains a large amount of noise clutter and false targets. These measurement points that are significantly inconsistent with the target movement trend are called "outlier points", and data cleaning is required to obtain the target point cloud data with higher accuracy.

[0032] See Figure 3 , in this embodiment, S2 further includes: S21. Construct an occupancy grid map as a representation of road traffic information. The color of each grid in the occupancy grid map represents the probability of that grid being occupied, that is, the number of times the point cloud of different target vehicles occupies it. The darker the color, the more times the point cloud of the passing vehicle occupies it. Therefore, each grid can be represented by a numerical value. The higher the numerical value, the more likely it is to be the lane center area. At the same time, the occupied grid also records the coordinate information of each point trace; S22. Divide the road environment where the vehicle travels in the detection area of the millimeter-wave radar into grids. The smaller the grid size, the more grids there are, and the more accurate the lane line information contained; However, it should be noted that the smaller the grid division, the higher the required computing and storage performance. Since the point cloud of the detected target vehicle by the millimeter-wave radar is not concentrated at the vehicle center but is uncertainly distributed in an area the size of a vehicle, after comprehensive consideration, the present invention sets the grid to a size of 10x1m in length and width; S23. Statistically analyze the preprocessed target point cloud data, calculate the number of point clouds falling on each grid respectively, and record the coordinates of each point cloud falling on the grid. After the cumulative statistics of the road vehicle point clouds in a period of frames, an occupancy grid map containing the current road traffic information is obtained.

[0033] There are multiple lanes in a road cross-section. According to the objective road traffic regulations and the subjective driving habits of drivers, the probability that a vehicle travels within the lane area is much greater than the probability that a vehicle straddles the lane line. However, vehicle driving is random, and some vehicles will cross lanes instead of driving along the lane. Therefore, most of the point cloud coordinates detected by the millimeter-wave radar are located in the center of the lane, and a small part appears in the middle of two lanes.

[0034] Through the constructed occupancy grid, the grid area with the darkest color on the map is the lane position on the road detected by the millimeter-wave radar. However, at this time, only the rough position of the lane is detected, and the extracted lane line is not accurate. Therefore, further precise extraction is required. The lane line is precisely extracted from the vehicle point cloud data in each individual lane area through Gaussian distribution fitting in the following embodiments.

[0035] See Figures 4-6 , in this embodiment, the S3 further includes: S31. Make a data statistic on the abscissa of the statistical point clouds of all grids in each row longitudinally in the occupancy grid map. The independent variable is the lateral coordinate of the vehicle position, and the dependent variable is the number of abscissa values of the vehicle point cloud positions detected by the millimeter-wave radar; S32. Each row of grids includes multiple peaks, and each peak corresponds to a lane, which is consistent with the actual quantity. The number of lane peaks and the reliable positions of the lanes are detected through the CFAR algorithm; S33. For the center position of a lane, multiple peaks will be generated after being detected by the CFAR algorithm. It is set that starting from the first detected peak in the CFAR algorithm detection peak set, the next detected peak whose lateral distance from the first peak is less than 3m belongs to the same lane peak set, and the one greater than 3m is the first detected peak of the next lane. Thus, the number of reliable lanes in the occupancy grid of this row and the positions of the grids where the centers of each lane are located are obtained.

[0036] Determining the number and position of lanes only by the depth of the occupancy grid color does not have a threshold standard for judgment and is also unreliable and inaccurate. This is because the grids in the middle position between two lanes may also have a high point cloud quantity statistic, and it cannot be excluded that the number of point clouds statistically counted in the center grid of some lanes is much less than that of other lane center grids. The above situations may cause omissions and incorrect increases in the number and position of lanes. In this case, the precise extraction of lane lines using Gaussian fitting will lead to incorrect extraction of lane lines.

[0037] Therefore, in order to exclude the darker grids in the middle of the two lanes and not miss the lighter grids with less vehicle movement on the lanes, a simple and effective method is needed to detect the number of lanes, providing more reliable and accurate lane position data for subsequent steps and improving the accuracy of the lane line fine extraction method based on Gaussian fitting distribution. Therefore, after the occupancy grid map of the road surface detected by the millimeter-wave radar is constructed, it is first necessary to clarify the accurate and reliable number of lanes and lane positions included in the detected road surface, and then the subsequent accurate lane line extraction can be carried out. Therefore, the present application proposes a detection method combining the practical CFAR algorithm.

[0038] The CFAR algorithm is a common means of radar target detection, and its processing flow is as Figure 5 shown. The input of the CFAR detection algorithm is the detection unit (Clutter Under Test, CUT) and a total of 2n reference units, each with n on both sides of the detection unit. The role of the protection unit is to prevent the target energy from leaking into the reference unit when a single target spans multiple range units, resulting in an error in the estimation of the noise clutter power level and affecting the detection of the target. The reference threshold level can be expressed as:

[0039] where Z is the estimated clutter noise power level and α is the threshold factor. Then, when CUT > T, it can be considered that there is a target; otherwise, it is considered that there is no target.

[0040] Using the CFAR algorithm to perform peak detection on the point cloud distribution of a certain row of occupancy grids obtained in the previous embodiment, the result is as Figure 6 shown.

[0041] In this embodiment, step S4 further includes: S41. After the CFAR algorithm detects the reliable number of lanes and lane positions, select multiple grids where the peaks detected by the CFAR algorithm are located (if the peak set is completely contained within one grid, then additionally select one grid on each side of this grid, for a total of 3 grid data) for point cloud coordinate statistics. Since the vehicle point clouds on each lane are mostly distributed in the center of the lane and a small number are distributed on both sides of the lane, meeting the Gaussian distribution conditions, use the Gaussian distribution to fit the point cloud distribution data. The Gaussian data distribution model formula is as follows:

[0042] where μ represents the mean and σ represents the standard deviation; S42. Solve the Gaussian data distribution model for each lane, and obtain the accurate center position of the lane from the Gaussian distribution center of each lane; S43. Repeatedly extract and fuse each vertical row of grids to obtain continuous road lane lines in the longitudinal direction of the road.

[0043] An embodiment of the present invention also provides a computer-readable storage medium storing a computer program, which when executed by a processor causes the processor to perform the steps of the above-described lane line extraction method based on the original point cloud of a millimeter-wave radar.

[0044] See Figure 7 , an embodiment of the present invention also provides a computer device including a memory and a processor, the memory storing a computer program, which when executed by the processor causes the processor to perform the steps of the above-described lane line extraction method based on the original point cloud of a millimeter-wave radar.

[0045] An embodiment of the present invention also provides a computer program product containing instructions, which when run on a computer causes the computer to perform the steps of the above-described lane line extraction method based on the original point cloud of a millimeter-wave radar.

[0046] It can be understood that the system, device, and storage medium provided by the embodiments of the present invention correspond to the method provided by the embodiments of the present invention. For the explanations, examples, and beneficial effects of related content, reference can be made to the corresponding parts in the above-described lane line extraction method based on the original point cloud of a millimeter-wave radar.

[0047] It should be noted that those of ordinary skill in the art can understand that all or part of the steps implemented in the embodiments of the present invention can be implemented in whole or in part through software, hardware, firmware, or any combination thereof. When implemented using hardware, it can be implemented in whole or in part in the form of purchasing standard parts or modified parts. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server or data center that includes one or more integrated available media. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid state disk (SSD)), etc.

[0048] In summary, the lane line extraction method based on millimeter-wave radar provided by the present invention has the advantages of simple calculation, fast response, accuracy, high precision, etc. It can be widely applied to the field of intelligent transportation, has important theoretical significance for the development of intelligent transportation, and has high application value and innovative research results.

[0049] It should be understood that the examples and embodiments described herein are for illustrative purposes only and are not intended to limit the present invention. Those skilled in the art can make various modifications or changes according to it. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A lane line extraction method based on the original point cloud of a millimeter-wave radar, characterized in that, Including the following steps: S1. The millimeter-wave radar acquires the original point cloud data and preprocesses the original point cloud data to form target point cloud data; S2. Construct an occupancy grid map, perform statistical calculations on the target point cloud data, and perform rough extraction of lane lines; S3. Use the CFAR algorithm to determine the number and position of road lanes and obtain multiple detection peaks; S4. According to the multiple grids where the CFAR algorithm detection peaks are located, perform point cloud coordinate statistics on the target point cloud data, and perform fine extraction of lane lines through Gaussian distribution fitting.

2. The lane line extraction method based on the millimeter-wave radar raw point cloud according to claim 1, characterized in that, In S1, the millimeter-wave radar uses a transmitting antenna to transmit a frequency-modulated continuous wave signal into the front space to cover multiple lane areas. The signal beam reflects when encountering a target or obstacle during transmission, generating an echo signal. The receiving antenna receives this echo signal and calculates it through the signal processing system of the millimeter-wave radar's built-in chip to obtain multi-dimensional information of the detection points in the detected road area, that is, the original point cloud data. The original point cloud data at least includes radial distance, azimuth angle, height, radial velocity, and signal strength.

3. The lane line extraction method based on the original point cloud of millimeter-wave radar according to claim 1, wherein In S1, with the position center of the millimeter-wave radar as the origin, according to the actual road environment, the detection range is divided into a rectangular area, and effective targets are screened by restricting the longitudinal and lateral ranges of the millimeter-wave radar targets to clean the original point cloud data and form the target point cloud data. It further includes: S11. Delete the data in the original point cloud data that does not meet the range of the detected road and the data that does not meet the requirements of the maximum detection distance that the millimeter-wave radar can detect; S12. Eliminate some noise points with small power and echo energy, and static target point clouds with a speed of 0 according to the road conditions; S13. Eliminate the point traces with few data items along the road direction and false ghost point traces.

4. The lane line extraction method based on the millimeter-wave radar raw point cloud according to claim 1, characterized in that, S2 further includes: S21. Construct an occupancy grid map as a representation of road traffic information. The color of each grid in the occupancy grid map represents the probability that the grid is occupied, that is, the number of times the point cloud of different target vehicles occupies. The darker the color, the more times the point cloud of the passing vehicle occupies. Therefore, each grid can be represented by a numerical value. The higher the numerical value, the more likely it is to be the lane center area. At the same time, the occupied grid also records the coordinate information of each point trace; S22. Divide the vehicle driving road environment in the detection area of the millimeter-wave radar with grids. The smaller the grid size, the more the number, and the more accurate the lane line information contained; S23. Perform statistics on the preprocessed target point cloud data, calculate the number of point clouds falling on each grid respectively, and record the coordinate of each point cloud falling on the grid. After the cumulative statistics of the road vehicle point clouds in a period of time frame, an occupancy grid map containing the current road traffic information is obtained.

5. The lane line extraction method based on the millimeter-wave radar raw point cloud according to claim 4, characterized in that S3 further includes: S31. Perform data statistics on the abscissa of the statistical point clouds of all grids in each row of the occupancy grid map longitudinally. The independent variable is the lateral coordinate of the vehicle position, and the dependent variable is the number of abscissa values of the millimeter-wave radar detected vehicle point cloud positions; S32. Each row of grids includes multiple peaks, and each peak corresponds to a lane, which is consistent with the actual quantity. The number of lane peaks and the reliable positions of the lanes are detected through the CFAR algorithm; S33. For the central position of a lane, multiple peaks will be generated after being detected by the CFAR algorithm. It is set that starting from the first detected peak in the CFAR algorithm detection peak set, the next detected peak whose lateral distance from the first peak is less than 3 m belongs to the same lane peak set, and the one greater than 3 m is the first detected peak of the next lane. Thus, the number of reliable lanes occupied by this row of grids and the positions of the grids where the centers of each lane are located are obtained.

6. The lane line extraction method based on the millimeter-wave radar original point cloud according to claim 5, wherein The said S4 further includes: S41. After the CFAR algorithm detects the reliable number of lanes and lane positions, select multiple grids where the CFAR algorithm detection peaks are located for point cloud coordinate statistics. Since the vehicle point clouds on each lane are mostly distributed in the center of the lane and a small number are distributed on both sides of the lane, which conforms to the Gaussian distribution condition, the Gaussian distribution is used to fit the point cloud distribution data. The Gaussian data distribution model formula is as follows:

7. Among them, μ represents the mean value, and σ represents the standard deviation; S42. Solve the Gaussian data distribution model of each lane, and obtain the accurate central position of the lane from the Gaussian distribution center of each lane; S43. Repeatedly extract and fuse each row of grids longitudinally to obtain continuous road lane lines in the longitudinal direction of the road.

8. A computer device, characterized in that, It includes a memory and a processor. When the computer program stored in the memory is executed by the processor, the processor executes the steps of the lane line extraction method based on the original point cloud of the millimeter-wave radar according to any one of claims 1 - 6.

Images