A method for identifying road lane lines based on millimeter-wave radar point cloud

Abstract

The present invention discloses a lane line recognition method based on millimeter-wave radar point cloud, which mainly includes four stages: millimeter-wave radar traffic target information parsing, traffic target effective driving trajectory extraction, lane line clustering, and lane line fitting. Millimeter-wave radar traffic target information parsing mainly processes the acquired original point cloud into the required data format. Traffic target effective driving trajectory extraction mainly extracts the trajectories in the characteristic direction from a large number of trajectories, which can avoid a large number of original trajectories from participating in the calculation and improve the calculation efficiency. According to the DBSCAN density clustering algorithm, the overall clustering of traffic target trajectories is carried out, and the trajectories belonging to the same lane line are divided into one category, and the average trajectory of this category is obtained. Finally, the average trajectory is curve-fitted, and the bisector of the two adjacent lane centerlines can be taken to obtain the lane line.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent transportation, and particularly relates to a method for identifying road lane lines based on millimeter-wave point clouds. Background Art

[0002] Lane lines are an important part of the transportation field and play an indispensable role in intelligent transportation fields such as vehicle-road collaboration and autonomous driving. Currently, millimeter-wave radars are widely used in the field of intelligent transportation. By installing millimeter-wave radars at intersections or on roads and adjusting the positions of the radars according to the detection ranges, the detection of multiple targets, multiple lanes, and multiple traffic parameters can be achieved, such as traffic flow, driving speed, vehicle type, queue length, etc., and all-round detection of the traffic status at intersections can be realized. Lane line detection provides road surface information for the intelligent transportation system and is of great significance for realizing the detection of driving behaviors and continuous tracking of vehicles.

[0003] Currently, the technical solutions for lane line recognition mainly fall into two categories: one is based on video image detection. Most of these methods adopt edge extraction technology or machine vision technology, such as a lane line detection method based on point clustering (201610195295), which uses roadside camera data, performs point clustering, and applies the coordinate transformation idea used in Hough line detection to lane line detection. However, since lane lines will wear out over time and cameras are vulnerable to external factors such as light, haze, night, rain, and snow, the images will be blurred, resulting in the inability to detect road lane lines.

[0004] The other is based on lidar detection. Most of these methods extract lane line features according to the intensity information reflected by the lidar or using the characteristics of echo pulse widths. They are not easily affected by the external environment and have high detection accuracy. For example, a lane line detection method based on lidar (201910127877) performs hierarchical processing on point cloud data and uses the local variance method to extract lane lines. However, when the lidar scans lane lines, it will be interfered by green belts or other vehicles, resulting in deviations in the detection results. At the same time, the price of lidar is also very expensive. Millimeter-wave radars, as sensors that can work all-weather, have high detection accuracy and strong anti-interference ability, can well adapt to various different scenarios, overcome the defects of cameras under external factors, and at the same time, the price is lower than that of lidar. Summary of the Invention

[0005] The object of the present invention is to address the drawbacks and deficiencies of existing lane line technologies, and provide a method for extracting lane lines based on millimeter-wave radar point clouds. Without relying on other additional detection information, and only depending on the vehicle trajectory data provided by the radar, a lane line recognition method based on millimeter-wave radar 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 extract lane lines and is applicable to various scenarios.

[0006] The present invention is achieved by at least one of the following technical solutions.

[0007] A method for recognizing road lane lines based on millimeter-wave radar point clouds includes the following steps:

[0008] S1. Obtain the original point cloud of traffic targets within the detection range, parse the collected original point cloud data to obtain traffic parameter information, and at the same time correct the tilted vehicle movement trajectory through coordinate transformation to make it vertical;

[0009] S2. Use the discrete point cloud captured by the millimeter-wave radar to characterize the continuous movement trajectory of traffic targets in the detection area through association, and screen effective trajectories from the trajectories based on length features, movement angle features, and speed features to obtain a traffic target trajectory density map;

[0010] S3. Perform trajectory clustering on the selected effective trajectories of traffic targets, classify the trajectories belonging to the same lane line into one category, and calculate the average driving trajectory of traffic targets in each category to obtain the center line of each lane;

[0011] S4. Take the bisector of two adjacent lane center lines, and perform least squares curve fitting on the center point coordinates in the bisector set to obtain the lane lines.

[0012] Furthermore, the parsing in step S1 includes:

[0013] Parse the collected data according to the protocol to obtain the detection response set as:

[0014] D = {d1, d2,..., d n}, d i = {c i , t i} (1)

[0015] where d n represents the detection response, c i represents the information of the detection target response, c i = (p xi , p yi , v xi , v yi ), p xi, p yi are the position attribute components in the observations of target i; v xi , v yi are the velocity attribute components corresponding to target i; t i represents the time of the detection response, i.e., the number of frames of the detection sequence;

[0016] Based on the positive and negative values of the parsed v x , determine the driving direction of the road to distinguish between oncoming vehicles and departing vehicles. The parsing formula is as follows:

[0017]

[0018] In the formula, XH and XL are the high byte and low byte of the target position in the X direction respectively; YH and YL are the high byte and low byte of the target position in the Y direction respectively; XVH and XVL are the high byte and low byte of the target velocity in the X direction respectively; YVH and YVL are the high byte and low byte of the target velocity in the Y direction respectively.

[0019] Furthermore, assume that the angle between the traffic target P and the vertical axis of the radar rectangular coordinate system is β, and the angle between the lane coordinate system and the radar coordinate system is α. The coordinate conversion relationship between (x′, y′) in the lane rectangular coordinate system and (x, y) in the radar rectangular coordinate system is:

[0020]

[0021] After coordinate conversion of the position and velocity information (p, v) measured by the actually installed radar, the (p′, v′) information obtained is:

[0022]

[0023] Furthermore, the characterization of step S2 includes:

[0024] Select the velocity and position obtained by the millimeter-wave radar sensor as the similarity metric for association matching. After association, a total of N target trajectories are obtained, denoted as the set T = {TR1, TR2,..., TR N}, and for any trajectory A in the trajectory set, it is represented as:

[0025] TR A = {a k , k = 1, 2,..., K}, a k = [x k , y k , vx k , vy k (5)

[0026] In the formula, a kis the k-th sampling point of trajectory A, K is the number of frames for the target movement to last, (x k , y k ) is the position coordinate of the target centroid in the k-th frame, (vx k , vy k ) is the velocity coordinate of the target centroid in the k-th frame.

[0027] Furthermore, based on velocity features, length features, motion angle features, and standard deviation features, filter the effective trajectories of traffic targets:

[0028]

[0029] In the formula, are respectively the average velocities of trajectory TR i in the horizontal and vertical directions; (x s , y s ) is the coordinate of the starting point of the trajectory, (x e , y e ) is the coordinate of the ending point of the trajectory; N(x, y) is the number of trajectory coordinate values included in the target trajectory segment; Y st is the horizontal standard deviation; y i (k) is the distance value of the k-th point of trajectory TR i relative to the radar in the horizontal direction; μ represents the average value of the distance values of all points of trajectory TR i relative to the radar in the horizontal direction; ξ v , ξ θ , ξ st are respectively the velocity difference threshold, the angle threshold, and the horizontal standard deviation threshold.

[0030] Furthermore, step S3 specifically includes:

[0031] Perform clustering processing on the effective trajectories of traffic targets obtained by screening through the DBSCAN density clustering algorithm, and at the same time set the neighborhood size and density threshold, classify the trajectory segments belonging to the same lane line into one category, and calculate the average driving trajectory of traffic targets in each category to obtain the center line of each lane;

[0032] Find out the representative trajectory in each lane, that is, the center line of each lane. The specific steps are as follows:

[0033] (1) Establish a set k_point of trajectories in the same lane, and store the traffic target trajectories belonging to the same lane obtained by clustering into the established set;

[0034] (2) Number the traffic targets passing through this lane in chronological order, traverse to find the mapping equivalent points from the starting point to the ending point of the path of the first vehicle and from the starting point to the ending point of the path of the last vehicle, and put the mapping equivalent points into the set mean_points of average points;

[0035] (3) The central mapping equivalent points of the vehicles in the remaining lanes Obtained according to step (2), Indicating the i-th central mapping equivalent point of the k-th lane, k = 1, 2, …, K;

[0036] (4) Use the least squares method to fit the equivalent points in mean_points obtained for each lane into a line to obtain the center line of each lane.

[0037] Furthermore, after initially obtaining the position of the center line, the bisector of the center lines of two adjacent lanes can be taken to obtain the lane lines:

[0038]

[0039] Wherein Is the intersection point of the center line of the middle lane And the straight line with the ordinate Hi, Is the intersection point of the center line of the middle lane And the straight line with the ordinate Hi, Is the coordinate of the center point of the lane line. For the center point coordinates in the set of center point coordinates Perform curve fitting to obtain the lane lines between adjacent lanes.

[0040] Furthermore, in step S1, it also includes screening the radar targets within the specified recognition range and eliminating the invalid targets.

[0041] Furthermore, the screening process includes: taking the center of the radar position as the origin, dividing the research range into rectangular areas according to the actual road environment, screening the effective target points by restricting the longitudinal and lateral ranges of the radar targets, establishing a judgment criterion as shown in Equation (8). If the judgment criterion is satisfied, the effective target points are included in the effective samples, and if not, they are filtered:

[0042]

[0043] In the formula, Y dist Is the set lateral range; X max Is the maximum longitudinal range that the radar can detect, that is, the maximum effective distance; X min Is the minimum longitudinal range that the radar can detect, that is, the minimum effective distance, (p x 、p y ) are the effective target points.

[0044] Furthermore, in step S2, the continuous movement trajectory of the traffic target is generated and terminated by associating the detection data of adjacent frames.

[0045] Furthermore, the extraction of lane lines is an iterative process. Vehicle trajectories are continuously input, and lane lines are continuously recognized. The more vehicle trajectory data there is, the more accurate the recognition accuracy of lane lines will be.

[0046] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0047] (1) The present invention uses a roadside-fixed millimeter-wave radar, makes full use of the returned point cloud data, and recognizes lane lines by obtaining the average driving trajectories of traffic targets, realizing the recognition of lane lines using millimeter-wave radar points.

[0048] (2) The lane line recognition method proposed by the present invention does not rely on the video image information required by traditional methods. It can detect lane lines under different external environmental conditions. Even if the road traffic markings are worn and unclear, lane line extraction can still be carried out without relying on external conditions and is applicable to various scenarios. It can timely, accurately, and effectively recognize the lane lines of the road only relying on the traffic target point cloud data obtained by the millimeter-wave radar. It can also have good performance even under the influence of low light intensity, bad weather, etc. The required cost is low, and it has universal applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a schematic flow chart of a lane line recognition method based on millimeter-wave radar point cloud in an embodiment;

[0050] Figure 2 is a schematic diagram of radar installation and sensing range in an embodiment;

[0051] Figure 3 is a schematic diagram of radar coordinate transformation in an embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] The present invention will be further described in detail below in conjunction with the embodiments and the drawings. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0053] The present invention relates to a method for recognizing road lane lines based on millimeter-wave radar point cloud. This method can make full use of the point cloud data detected by the millimeter-wave radar to obtain the continuous motion trajectories of traffic targets, extract the trajectories in the characteristic direction from a large number of trajectories, and avoid a large number of original trajectories from participating in the calculation. Then, the density clustering algorithm is used to perform overall clustering on the traffic target trajectories, classify the trajectories belonging to the same lane line into one category, calculate the average trajectory of this category, and finally perform curve fitting on the average trajectory line. The bisector of the two adjacent lane centerlines can be taken to obtain the lane lines.

[0054] Example 1

[0055] As Figure 1 shown, the millimeter-wave radar-based lane line recognition method of this embodiment specifically includes the following steps:

[0056] First step: Install the millimeter-wave radar detector on a pole at a certain height above the detection road and adjust the angle according to the actual situation. In this embodiment, the most original data received by the traffic millimeter-wave radar used is in hexadecimal form, and each frame is counted as a packet of data. The refresh frequency of this radar is 70 ms, that is, the radar receives 14 frames of data per second.

[0057] The millimeter-wave radar is installed in the center of the crossbar on the side of an intersection road. Its longitudinal sensing range is 200 m, and the lateral range exceeds the overall width of the lane. It can detect and track up to 200 real-time targets simultaneously. Analyze the collected data according to the specified protocol to obtain the detection response set obtained by the radar detector as D = {d1, d2,..., d n}, d i = {c i , t i}, where c i represents the information of the detection target response, mainly speed and position information, that is, c i = (p xi , p yi , v xi , v yi ), p xi , p yi are the position attribute components in the observation value of target i respectively; v xi , v yi are the corresponding speed attribute components respectively; t i represents the time of the detection response, that is, the number of frames of the detection sequence. Determine the driving direction of the road according to the positive and negative values of the parsed v x to distinguish oncoming vehicles and outgoing vehicles. For oncoming vehicles, the position coordinate of the vehicle in the x-axis direction gradually becomes smaller, and the position coordinate in the y-axis direction remains basically unchanged; for vehicles that decelerate, the speed of the vehicle in the x-axis direction gradually decreases, and the speed in the y-axis direction approaches zero. The parsing formula is as follows:

[0058]

[0059] In the formula, XH and XL are the high byte and low byte of the target position in the X direction respectively; YH and YL are the high byte and low byte of the target position in the Y direction respectively; XVH and XVL are the high byte and low byte of the target speed in the X direction respectively; YVH and YVL are the high byte and low byte of the target speed in the Y direction respectively.

[0060] Perform coordinate transformation on the traffic target information parameters and correct the vehicle motion trajectory that is tilted at a certain angle. In the actual scenario, there is often an angle between the projection of the radar irradiation direction and the lane direction, and the radar may also deviate from the center of the lane. Assume that the angle between the traffic target P and the vertical axis of the radar rectangular coordinate system is β, and the angle between the lane coordinate system and the radar coordinate system is α. Since the subsequent processing of radar data is carried out in the lane rectangular coordinate system, after performing coordinate transformation on the position and speed information (p, v) measured by the actually installed radar, the obtained (p′, v′) information is as follows:

[0061]

[0062] In this embodiment, the angle α between the lane coordinate system and the radar coordinate system is taken as 4.5 degrees, and the schematic diagram of radar coordinate transformation is as shown in Figure 3 shown.

[0063] Step 2: Extract effective trajectory data of traffic targets. Some vehicles may change lanes on the section near the intersection, and the standard deviation of the trajectory of the target data detected by the radar in the Y direction will increase, and the trajectory intersects with the lane line. This type of data will affect the result of lane line recognition. In addition, there may be problems such as pedestrians crossing the street and trajectory interruption during the target detection process by the radar, which will all affect the result of lane line recognition. This type of data needs to be excluded during the recognition process.

[0064] First, obtain the continuous motion trajectory of traffic targets in the radar detection area through data association. The millimeter-wave radar sensor is relatively accurate in both position and speed perception. Therefore, its speed and position can be selected as the similarity metric for association matching. After association, a total of N target trajectories are obtained, which are denoted as the set T = {TR1, TR2,..., TR N}, and any trajectory A in the trajectory set is represented as:

[0065] TR A ={a k , k = 1, 2,..., K}, a k =[x k , y k , vx k , vy k (3)

[0066] In the formula, a k is the k-th sampling point of the trajectory TR A , K is the number of frames for the target motion to continue, (x k , y k ) is the position coordinate of the target in the k-th frame, and (vx k , vy k ) is the speed coordinate of the target in the k-th frame.

[0067] Specifically, screening valid trajectories based on features includes:

[0068] [1] Speed limit, filtering pedestrian crossing trajectories and stationary vehicle data according to the lateral speed;

[0069] [2] Length limit, a minimum value limit needs to be given to the number of coordinate value pairs included in the generated target trajectory to eliminate noise;

[0070] [3] Movement angle limit, in order to limit some trajectories turning left or right at intersections, a maximum trajectory turning angle can be set;

[0071] [4] Standard deviation limit, some vehicles may change lanes on the section near the intersection, the lateral standard deviation of the target trajectory will increase, and the trajectory intersects with the lane line.

[0072]

[0073] In the formula, are respectively the average speeds of the trajectory TR i in the lateral and longitudinal directions; (x s , y s ) are the coordinates of the starting point of the trajectory, (x e , y e ) are the coordinates of the end point of the trajectory; N(x, y) is the number of trajectory coordinate values included in the target trajectory segment; Y st is the lateral standard deviation; y i (k) is the distance value of the k-th point of the trajectory TR i relative to the radar in the lateral direction; μ represents the average value of the distance values of all points in the trajectory TR i relative to the radar in the lateral direction; ξ v , ξ θ , ξ st are respectively the speed difference threshold, the angle threshold, and the lateral standard deviation threshold.

[0074] Step 3: The vehicle trajectories obtained after screening will cover the lane area. The clustering algorithm can effectively identify the range of the vehicle driving interval. The more vehicle trajectory data there is, the more accurate the recognition accuracy is. The DBSCAN clustering algorithm is used to cluster the screened valid traffic target trajectories, with the neighborhood size taken as 1 and the density threshold taken as 1.3. The trajectory segments belonging to the same lane line are divided into one category. The average driving trajectory of the traffic target obtained by processing the classified clusters is fitted by the least squares method to obtain the lane center line. The main task of obtaining the average driving trajectory is to calculate the clustering result and find out the representative trajectory in each lane. The specific steps of the center line of each lane are as follows:

[0075] (1) Establish the set of trajectories in the same lane, k_point, and store the traffic target trajectories belonging to the same lane obtained by clustering into the established set.

[0076] (2) Number the traffic targets passing through this lane in chronological order, traverse to find the mapping equivalent points from the starting point to the ending point of the path of the first vehicle and from the starting point to the ending point of the path of the last vehicle, and put the mapping equivalent points into the set of mean points, mean_points.

[0077] (3) Obtain the central mapping equivalent points of the vehicles in the remaining lanes according to step (2). Let represent the i-th central mapping equivalent point in the k-th lane, where k = 1, 2, …, K;

[0078] (4) Use the least squares method to fit the equivalent points in mean_points obtained for each lane into a line, and thus obtain the center line of each lane.

[0079] Step 4: The lane lines on the actual road are generally composed of symmetric left and right edges, and the lengths of the left and right edges are fixed and unchanged. After obtaining the positions of the center lines of each lane, take the bisector of the two adjacent lane center lines to obtain the lane lines. The set of center lines obtained by fitting in the third step is L C , represents the center line of the middle lane. Let the center line intersect with the line with the ordinate H i at the point The center line intersects with the line with the ordinate Hi at the point Then the calculation formula for the coordinates of the center point of the lane line is:

[0080]

[0081] Perform curve fitting on the obtained center point coordinates to obtain the lane lines between adjacent lanes.

[0082] Through the above steps, the lane lines within the detection range of the millimeter-wave radar can be output.

[0083] Example 2

[0084] To test the applicability of the method disclosed in the present invention to different time periods, an experiment is carried out by changing the straight-line scenario to a curved road based on Example 1. The explanations of the same or corresponding terms as those in the above examples are not repeated here.

[0085] S1. Obtain the original point cloud of traffic targets within the detection range. Analyze the collected original point cloud data according to the corresponding protocol to obtain traffic target distance, speed, and motion parameter information. At the same time, perform coordinate transformation to correct the motion trajectory of vehicles tilted at a certain angle (such as 3 - 5 degrees) to make it vertical;

[0086] Specifically, install the millimeter - wave radar detector on a pole at a certain height above the detection road and adjust the angle according to the actual situation. In this embodiment, a section of the road with a bend is selected as the scenario.

[0087] Specifically, the parsing protocol is designed to enable the device - side to upload the collected target vehicle position and motion information in a unified packet format to improve the real - time performance of tracking. Each frame of data received by the radar device contains a frame header, a frame tail, additional information, and the main content. The specific protocol format is shown in Table 1:

[0088] Table 1 Example of data protocol format

[0089]

[0090] S2. Utilize the discrete point cloud captured by the millimeter - wave radar to characterize the continuous motion trajectory of traffic targets in the detection area through association. Screen the effective trajectories from the trajectories based on length features, motion angle features, and speed features to obtain the traffic target trajectory density map;

[0091] S3. Perform trajectory clustering on the screened effective trajectories of traffic targets, classify the trajectories belonging to the same lane line into one category, and calculate the average driving trajectory of traffic targets in each category to obtain the center line of each lane;

[0092] S4. Take the bisector of two adjacent lane center lines, and perform curve fitting on the center point coordinates in the bisector set to obtain the lane line.

[0093] Embodiment 3

[0094] S1. Obtain the original point cloud of traffic targets within the detection range. Analyze the collected original point cloud data according to the specified protocol to obtain traffic target distance, speed, and motion parameter information. At the same time, perform coordinate transformation to correct the motion trajectory of vehicles tilted at a certain angle to make it vertical;

[0095] Specifically, install the millimeter - wave radar detector on a pole at a certain height above the detection road and adjust the angle according to the actual situation. In this embodiment, the data in the time period from 0:00:00 to 06:00:00 is selected.

[0096] S2. Using the discrete point cloud captured by the millimeter-wave radar, the continuous motion trajectory of traffic targets in the detection area is characterized through association. Effective trajectories are screened from the trajectories based on length features, motion angle features, and speed features to obtain a traffic target trajectory density map;

[0097] S3. Perform trajectory clustering on the effective trajectories of traffic targets obtained by screening, classify the trajectories belonging to the same lane line into one category, and obtain the average driving trajectory of traffic targets in each category to obtain the center line of each lane;

[0098] S4. Take the bisector of two adjacent lane center lines, and perform curve fitting on the center point coordinates in the bisector set to obtain the lane line.

[0099] Specifically, the fitting algorithm of the lane line can be but is not limited to any one of the least squares method, the gradient descent method, and the Gauss-Newton algorithm.

[0100] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited by the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principle of the present invention shall be equivalent replacement methods and are all included in the protection scope of the present invention.

Claims

1. A method for identifying road lane lines based on millimeter-wave radar point cloud, characterized in that, It includes the following steps: S1. Obtain the original point cloud of traffic targets within the detection range, parse the collected original point cloud data to obtain traffic parameter information, and at the same time correct the tilted vehicle movement trajectory through coordinate transformation to make it vertical; S2. Use the discrete point cloud captured by the millimeter-wave radar, and characterize the continuous movement trajectory of traffic targets in the detection area through association. Screen the effective trajectories from the trajectories based on length features, movement angle features, and speed features to obtain the traffic target trajectory density map; S3. Perform trajectory clustering on the effective trajectories of traffic targets obtained by screening, classify the trajectories belonging to the same lane line into one category, and calculate the average driving trajectory of traffic targets in each category to obtain the center line of each lane; Screen the effective trajectories of traffic targets based on speed features, length features, movement angle features, and standard deviation features: In the formula, are respectively the average speeds of the trajectory TR i in the horizontal and vertical directions; (x s , y s ) is the coordinate of the starting point of the trajectory, and (x e , y e ) is the coordinate of the ending point of the trajectory; N(x, y) is the number of trajectory coordinate values included in the target trajectory segment; Y st is the standard deviation in the horizontal direction; y i (k) is the distance value of the k-th point of the trajectory TR i relative to the radar in the horizontal direction; μ represents the average value of the distance values of all points of the trajectory TR i relative to the radar in the horizontal direction; ξ v , ξ θ , and ξ st are the speed difference threshold, the angle threshold, and the horizontal standard deviation threshold respectively; S4. Take the bisector of the center lines of two adjacent lanes, and perform least squares curve fitting on the center point coordinates in the bisector set to obtain the lane line.

2. The road lane line recognition method based on millimeter-wave radar point cloud according to claim 1, characterized in that, The parsing of step S1 includes: Parse the collected data according to the protocol to obtain the detection response set as: D = {d1, d2,..., d n}, d i = {c i , t i} (1) Among them, d n represents the detection response, and c i represents the information of the detection target response, and c i =(p xi , p yi , v xi , v yi ), where p xi and p yi are respectively the position attribute components in the observation values of target i; v xi and v yi are respectively the velocity attribute components corresponding to target i; t i represents the time of the detection response, that is, the number of frames of the detection sequence. Based on the positive and negative values of the parsed v x Determine the driving direction of the road to distinguish between oncoming vehicles and vehicles going in the other direction. The parsing formula is as follows: In the formula, XH and XL are the high byte and low byte of the target position in the X direction respectively; YH and YL are the high byte and low byte of the target position in the Y direction respectively; XVH and XVL are the high byte and low byte of the target speed in the X direction respectively; YVH and YVL are the high byte and low byte of the target speed in the Y direction respectively.

3. A method for identifying road lane lines based on millimeter-wave radar point cloud according to claim 1, characterized in that, Assume that the angle between the traffic target P and the vertical axis of the radar rectangular coordinate system is β, and the angle between the lane coordinate system and the radar coordinate system is α. The coordinate transformation relationship between (x′, y′) in the lane rectangular coordinate system and (x, y) in the radar rectangular coordinate system is: After performing coordinate transformation on the position and speed information (p, v) measured by the actually installed radar, the obtained (p′, v′) information is:

4. A method for identifying road lane lines based on millimeter-wave radar point cloud according to claim 1, characterized in that, The characterization of step S2 includes: Select the speed and position obtained by the millimeter-wave radar sensor as the similarity metric for correlation matching. After correlation, a total of N target trajectories are obtained, denoted as the set T = {TR1, TR2,..., TR N}, and any trajectory A in the trajectory set is represented as: TR A = {a k , k = 1, 2, ..., K}, a k = [x k , y k , vx k , vy k (5) where a k is the k-th sampling point of trajectory A, K is the number of frames for the target movement to last, (x k , y k ) is the position coordinate of the target centroid in the k-th frame, (vx k , vy k ) is the velocity coordinate of the target centroid in the k-th frame.

5. A method for identifying road lane lines based on millimeter-wave radar point cloud according to claim 1, characterized in that Step S3 specifically includes: Perform clustering processing on the effective trajectories of traffic targets obtained by screening through the DBSCAN density clustering algorithm, and at the same time set the neighborhood size and density threshold, classify the trajectory segments belonging to the same lane line into one category, and calculate the average driving trajectory of traffic targets in each category to obtain the center line of each lane; Find the representative trajectory in each lane, that is, the center line of each lane. The specific steps are as follows: (1) Establish a set k_point of trajectories in the same lane, and store the traffic target trajectories belonging to the same lane obtained by clustering into the established set; (2) Number the traffic targets passing through this lane in chronological order, traverse to find the mapping equivalent points from the starting point to the ending point of the path of the first vehicle and the starting point to the ending point of the path of the last vehicle, and put the mapping equivalent points into the set mean_points of average points; (3) The central mapping equivalent points of vehicles in the remaining lanes Obtained according to step (2), Indicating the i-th central mapping equivalent point of the k-th lane, where k = 1, 2, …, K; (4) Fit the equivalent points in mean_points obtained for each lane into a line using the least squares method to obtain the center line of each lane.

6. The method for identifying road lane lines based on millimeter-wave radar point cloud according to claim 1, characterized in that After initially obtaining the center line position, take the bisector of the center lines of two adjacent lanes to obtain the lane line: Among them is the center line of the middle lane the intersection point with the straight line whose ordinate is Hi, is the center line of the middle lane the intersection point with the straight line whose ordinate is Hi, is the coordinate of the center point of the lane line. For the center point coordinates in the set of center point coordinates perform curve fitting, and the lane lines between adjacent lanes can be obtained.

7. A method for identifying road lane lines based on millimeter-wave radar point cloud according to claim 1, characterized in that, In step S1, it also includes screening the radar targets within the specified recognition range and eliminating the invalid targets.

8. A method for identifying road lane lines based on millimeter-wave radar point cloud according to claim 7, characterized in that, The screening process includes: taking the center of the radar position as the origin, dividing the research scope into rectangular areas according to the actual road environment, screening valid target points by restricting the longitudinal and lateral ranges of radar targets, establishing a judgment criterion as shown in Equation (8), if the judgment criterion is satisfied, the valid target points are included in the valid samples, and if not, they are filtered: Where Y dist is the set horizontal range; X max is the maximum longitudinal range detectable by the radar, i.e., the maximum value of the effective distance; X min is the minimum longitudinal range detectable by the radar, i.e., the minimum value of the effective distance, (p x , p y ) is the effective target point.

9. A method for identifying road lane lines based on millimeter-wave radar point cloud according to claim 1, characterized in that, In step S2, the continuous motion trajectory of the traffic target is generated and terminated by associating the detection data of adjacent frames.

Images