CN114779235A - Road boundary detection method and system based on vehicle-mounted millimeter-wave radar - Google Patents

Abstract

The invention discloses a road boundary detection method and a system based on a vehicle-mounted millimeter wave radar.

Description

Road boundary detection method and system based on vehicle millimeter wave radar

technical field

The invention relates to the technical field of road detection, in particular to a road boundary detection method and system based on a vehicle-mounted millimeter-wave radar.

Background technique

Traditional road boundary detection is generally realized by using lidar, but lidar is expensive and cannot be popularized, while vehicle-mounted millimeter-wave radar is cheap, has good ranging and speed measurement capabilities for targets, and has good penetration of rain and fog. Capability, and not being affected by light intensity, has become an irreplaceable sensor choice in intelligent driving solutions.

Due to the interference of vehicles on highways and urban roads, the output of road boundary lines obtained by existing detection methods based on radar data collected by vehicle-mounted millimeter-wave radars is often unstable in the time domain, resulting in low detection accuracy.

SUMMARY OF THE INVENTION

In order to solve the above technical problems, the present invention provides a road boundary detection method and system based on a vehicle-mounted millimeter-wave radar, which solves the problem in the prior art that the output of the road boundary line obtained according to the radar data collected by the vehicle-mounted millimeter-wave radar is in the time domain. It is unstable and the detection accuracy is low.

In one aspect, the present invention provides a road boundary detection method based on a vehicle-mounted millimeter-wave radar, comprising:

Obtain all static target points of the current frame output by the vehicle-mounted millimeter-wave radar;

Screen a plurality of stable static target points from all the static coordinate points according to the initial screening range, and the initial screening range is determined according to the radar detection capability;

The stable static target point located in the preset screening range is used as the starting point of the alternative boundary; wherein, taking the central axis of the radar as the symmetry axis, two search ranges are set within the field of view of the vehicle-mounted millimeter-wave radar, and the The two set search ranges are used as the preset screening ranges, and each search range takes the sector radius boundary corresponding to the field of view range of the vehicle-mounted millimeter-wave radar as one of the boundaries, and the other boundary is parallel to the sector radius boundary;

Sort the candidate boundary starting points according to the distance from the radar origin from small to large, and use the candidate boundary starting point with the smallest distance as the boundary starting point;

Take the boundary starting point as the starting point, search for the boundary point from the stable static target point within the preset search range, and use the found boundary point as the new starting point to continue searching for the next boundary point until the search fails. until the next boundary point;

Use the least squares method to perform quadratic curve fitting on all the obtained boundary points to obtain the quadratic curve coefficient of the current frame;

The quadratic curve coefficients of the current frame and the quadratic curve coefficients of the previous frame are filtered by using the Kalman filtering algorithm, so as to obtain the target quadratic curve coefficients and the target road boundary curve.

Optionally, it also includes: if a plurality of boundary points are found within the preset search range, calculating the distance between each boundary point and the origin of the radar;

Sort all the boundary points in ascending order of the obtained distance, and use the boundary point corresponding to the minimum distance as the new starting point.

Optionally, the distance calculation formula between the boundary point and the radar origin is expressed as:

d ² =αP _x ² +βP _y ² ---(1)

In the formula: d is the distance between the calculated boundary point and the origin of the radar, α is the horizontal scale factor, β is the vertical scale factor, P _x is the abscissa of the boundary point, and P _y is the ordinate of the boundary point.

Optionally, also include:

If all the boundary points obtained by the search meet the set quantity threshold and length threshold, then transfer to the step of using the least squares method to perform quadratic curve fitting on all the boundary points obtained by the search to obtain the quadratic curve coefficient of the current frame;

Otherwise, the current boundary starting point is removed, and the candidate boundary starting point located next to the current boundary starting point in the sorting is used as a new boundary starting point.

Optionally, the step of removing the current boundary starting point and using the candidate boundary starting point that is located next to the current boundary starting point in the sorting as a new boundary starting point includes:

count the number of the remaining candidate boundary starting points;

If the number of the remaining candidate boundary starting points is greater than zero, the candidate boundary starting point located next to the current boundary starting point in the sorting is used as a new boundary starting point.

Optionally, in the radar coordinate system, the initial screening range is obtained by setting a horizontal threshold and a vertical threshold; wherein, the horizontal threshold is obtained according to the road width and radar detection capability in a real scene; the vertical threshold is obtained according to the radar detection capability. Sure.

Optionally, use the least squares method to perform quadratic curve fitting on all the obtained boundary points to obtain the quadratic curve coefficients of the current frame, including:

Get all the boundary points found in the previous frame;

Fusion of all boundary points found in the current frame and all boundary points found in the previous frame;

The quadratic curve fitting is performed on all the fused boundary points using the least squares method to obtain the quadratic curve coefficients of the current frame.

Optionally, the larger the vertical coordinate of the starting point, the larger the horizontal width and vertical width of the corresponding preset search range.

On the other hand, the present invention also provides a road boundary detection system based on a vehicle-mounted millimeter-wave radar, comprising:

The target point acquisition module is configured to acquire all static target points of the current frame output by the vehicle-mounted millimeter wave radar;

a preliminary screening module, configured to screen out a plurality of stable static target points from all the static coordinate points according to an initial screening range, the initial screening range being determined according to the radar detection capability;

The alternative starting point acquisition module is configured to use the stable static target point located within the preset screening range as the alternative boundary starting point; wherein, taking the radar center axis as the axis of symmetry, at the field of view of the vehicle-mounted millimeter-wave radar Two search ranges are set within the range, and the two set search ranges are used as the preset screening ranges, and each search range takes the sector radius boundary corresponding to the field of view angle range of the vehicle-mounted millimeter-wave radar as one of the boundaries, and the other a boundary parallel to the sector radius boundary;

The initial starting point determination module is configured to sort the candidate boundary starting points according to the distance from the radar origin from small to large, and use the candidate boundary starting point with the smallest distance as the boundary starting point;

The search module is configured to take the boundary starting point as a starting point, search for a boundary point from the stable static target point within a preset search range, and use the found boundary point as a new starting point to continue searching for the next Boundary point until the next boundary point cannot be found;

a curve fitting module, configured to perform quadratic curve fitting on all boundary points obtained by the least squares method to obtain quadratic curve coefficients of the current frame;

And, the filtering processing module is configured to perform filtering processing on the quadratic curve coefficient of the current frame and the quadratic curve coefficient of the previous frame by using the Kalman filtering algorithm, so as to obtain the target quadratic curve coefficient and the target road boundary curve.

In another aspect, the present invention also provides an electronic device, comprising: a processor and a memory, where computer-readable instructions are stored in the memory, and the computer-readable instructions are implemented as described above when executed by the processor A road boundary detection method based on vehicle-mounted millimeter-wave radar.

In another aspect, the present invention also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the above-mentioned road boundary detection method based on a vehicle-mounted millimeter-wave radar.

The road boundary detection method based on the vehicle-mounted millimeter-wave radar of the present invention first performs preliminary screening of the input vehicle-mounted millimeter-wave radar static target points to obtain relatively stable static target points, and uses the first starting point to cluster to find out by setting the search range. The strategy of all boundary points selects all possible road boundary points in the stable static target points, and then uses the least squares method to fit the quadratic curve combined with the boundary points screened in the previous frame of the radar. Finally, considering that the road boundary will not If a jump occurs in a short time, the Kalman filter algorithm is used to filter the three coefficients in the obtained quadratic curve, and a relatively stable curve output in the time domain is obtained, thereby improving the detection accuracy of road boundaries.

Description of drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings required in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some of the present invention. In the embodiments, for those of ordinary skill in the art, other drawings can also be obtained according to these drawings without any creative effort.

1 is a schematic flowchart of some embodiments of a road boundary detection method based on a vehicle-mounted millimeter-wave radar of the present invention;

2 is a schematic flowchart of other embodiments of a road boundary detection method based on a vehicle-mounted millimeter-wave radar of the present invention;

FIG. 3 is a schematic flowchart of some further embodiments of a road boundary detection method based on a vehicle-mounted millimeter-wave radar of the present invention;

Fig. 4 is the road boundary point tracking schematic diagram of the road boundary detection method based on the vehicle-mounted millimeter wave radar of the present invention;

Fig. 5 is the Kalman filter algorithm flow chart of the road boundary detection method based on vehicle-mounted millimeter wave radar of the present invention;

FIG. 6 is a structural block diagram of some embodiments of the road boundary detection system based on the vehicle-mounted millimeter wave radar of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to 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, but not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art fall within the protection scope of the present invention.

On the one hand, as shown in FIG. 1 , an embodiment of the present invention provides a road boundary detection method based on a vehicle-mounted millimeter-wave radar, including:

Step 100: Acquire all static target points of the current frame output by the vehicle-mounted millimeter-wave radar.

In this step, the vehicle-mounted millimeter-wave radar can be a millimeter-wave radar with the function of directly outputting static coordinate points; of course, it can also be a vehicle-mounted millimeter-wave radar without this function. In this case, it is necessary to first obtain the current frame radar response output by the vehicle-mounted millimeter-wave radar. wave signal, and then analyze all the static target points of the current frame from the radar echo signal of the current frame, and the static target point is the position data/coordinate data of the static target point.

Step 200: Screen out a plurality of stable static target points from all the static coordinate points according to the initial screening range, and the initial screening range is determined according to the radar detection capability.

In this step, the static target points located in the initial screening range are directly screened to obtain a plurality of stable static target points; wherein the initial screening range is determined according to the radar detection capability. For example, if the data reliability of the static target points detected by the vehicle-mounted millimeter-wave radar within the range of L meters is high, the static target points detected within the range of L meters are screened out as stable static target points. Here, the initial screening range can be a rectangle or a triangle. When it is a rectangle, the vertical distance of the initial screening range is set as L meters, the horizontal distance of the initial screening range is set as the preset width, and the The static target point can be screened out, and it does not need to be too wide; it can also be a triangle, and only the vertical distance threshold is set. Generally, the preset width can cover all or most of the roads with guardrails in the real scene.

Step 300: Use the stable static target point within the preset screening range as the starting point of the alternative boundary; wherein, taking the radar center axis as the axis of symmetry, two search ranges are set within the field of view of the vehicle-mounted millimeter-wave radar, The two set search ranges are used as the preset screening ranges, and each search range takes the sector radius boundary corresponding to the field of view range of the vehicle-mounted millimeter-wave radar as one of the boundaries, and the other boundary is parallel to the sector radius boundary.

In this step, the distance between the two boundaries of each search range should meet the detection requirements of borders such as guardrails, and the extension length of the search range in the direction of the sector radius is related to the boundary features of various roads, and is generally an empirical value. , the projection length of the extension length of the search range in the direction of the sector radius in the direction of the radar center axis is smaller than the corresponding length of the initial screening range in the direction of the radar center axis. In practical applications, after obtaining stable static target points, it can be sorted according to the ordinate from small to large, which is convenient for subsequent search of boundary points.

Step 400: Sort the candidate boundary starting points according to the distance from the radar origin from small to large, and use the candidate boundary starting point with the smallest distance as the boundary starting point.

Before searching for boundary points, calculate the distance between each candidate boundary starting point and the radar origin, sort the calculated distances from small to large, and take the candidate boundary starting point with the smallest distance as the boundary starting point.

Step 500: Take the boundary starting point as the starting point, search for the boundary point from the stable static target points within the preset search range, and use the found boundary point as the new starting point to continue to search for the next boundary point until no more can be found. until the next boundary point.

In this step, first determine a boundary starting point as the starting point, search for the next boundary starting point within the preset search range, when the next boundary starting point is found, use this boundary starting point as the new starting point, and search for the next The boundary point, that is, the stable static target point found within the preset search range is used as the boundary point to realize the tracking of the road boundary point. It should be noted that each starting point corresponds to a preset search range.

Step 600: Use the least squares method to perform quadratic curve fitting on all the obtained boundary points to obtain quadratic curve coefficients of the current frame.

In this step, the least squares method is used to perform quadratic curve fitting on the points obtained in step 400, and the fitting formula is as follows:

P _x =aP _y ² +bP _y +c---(2)

In the formula: a is the quadratic term coefficient of the fitted function, b is the linear term coefficient of the fitted function, and c is the constant term of the fitted function. P _x is the abscissa of the boundary point, and P _y is the ordinate of the boundary point.

Step 700 : filter the quadratic curve coefficients of the current frame and the quadratic curve coefficients of the previous frame by using the Kalman filtering algorithm to obtain the target quadratic curve coefficients and the target road boundary curve.

In this step, the three coefficients a, b, and c obtained from the current frame and the previous frame are used to filter and process the three coefficients of the current frame and the previous frame, and the Kalman filtering algorithm is used to obtain a relatively stable coefficient output in the time domain.

The road boundary detection method based on the vehicle-mounted millimeter-wave radar according to the embodiment of the present invention first performs preliminary screening of the input vehicle-mounted millimeter-wave radar static target points to obtain relatively stable static target points, and uses the first starting point to perform clustering by setting the search range The strategy of finding all boundary points is to screen all possible road boundary points in the stable static target points, and then, combined with the boundary points screened in the previous frame of the radar, use the least squares method to perform quadratic curve fitting, and finally, consider the road boundary There will be no jump in a short time, and the Kalman filter algorithm is used to filter the three coefficients in the obtained quadratic curve to obtain a relatively stable curve output in the time domain, thereby improving the accuracy of road boundary detection. The road boundary detection method based on the vehicle-mounted millimeter-wave radar according to the embodiment of the present invention can have very important engineering applications in unmanned driving and assisted driving.

In some embodiments, the road boundary detection method based on the vehicle-mounted millimeter-wave radar of the present invention further comprises:

If multiple boundary points are found within the preset search range, the distance between each boundary point and the origin of the radar is calculated;

Sort all the boundary points in ascending order of the obtained distance, and use the boundary point corresponding to the minimum distance as the new starting point.

In this embodiment, a boundary point is generally found through the preset search range. When multiple boundary points are found in the same preset search range, the distance d between each boundary point and the origin of the radar is calculated, Take the boundary point closest to the radar origin as the new starting point. In addition, in the actual carding process, it can also be set through the program, that is, if one is found, the search will be stopped, and then this point will be used for the next step of screening.

In some embodiments, the calculation formula of the distance d between the boundary point and the origin of the radar in the road boundary detection method based on the vehicle-mounted millimeter wave radar of the present invention is expressed as:

d ² =αP _x ² +βP _y ² ---(1)

In the formula: d is the distance between the calculated boundary point and the origin of the radar, α is the horizontal scale factor, β is the vertical scale factor, the value ranges of α and β are both [0, 1], and α=1 -β, α, β can be empirical values or calibration values, P _x is the abscissa of the boundary point, P _y is the ordinate of the boundary point.

In this embodiment, considering the characteristics of boundary points, an influence factor is added to the calculation formula when calculating the distance, and the values of α and β can be used to adjust the influence of _Px and _Py on d.

In some embodiments, as shown in FIG. 2 , the road boundary detection method based on the vehicle-mounted millimeter-wave radar of the present invention further includes:

Step 800: If all the boundary points obtained by the search satisfy the set quantity threshold and length threshold, then proceed to the step of using the least squares method to perform quadratic curve fitting on all the boundary points obtained by the search to obtain the quadratic curve coefficient of the current frame. ;

Step 900: Otherwise, remove the current boundary starting point, and use the candidate boundary starting point that is located next to the current boundary starting point in the sorting as a new boundary starting point.

In this embodiment, for all the boundary points found in step 500, a validity judgment needs to be performed, that is, it is judged whether all the boundary points satisfy the set quantity threshold and length threshold. If not, the current boundary starting point is removed. If the found boundary points meet the requirements, step 600 is performed. It should be noted that the quantity threshold and length threshold set here are empirical values or calibration values. In some other embodiments, because quadratic curve fitting is used, the boundary point data is at least 3, and the number threshold can also be comprehensively determined in combination with the number of output point clouds of the millimeter wave radar. The number of point clouds is large, and this threshold can be appropriately enlarged. The setting of the length threshold is related to the radar resolution and the actual scene, and the threshold is set to 8m here. In some other embodiments, it is also possible to go to step 600 when all the boundary points obtained by searching meet the set quantity threshold.

In some embodiments, as shown in FIG. 2 , step 900 in the road boundary detection method based on the vehicle-mounted millimeter-wave radar of the present invention includes:

Step 901: Count the number of remaining candidate boundary starting points;

Step 902: If the number of remaining candidate boundary starting points is greater than zero, use the candidate boundary starting point located next to the current boundary starting point in the sorting as a new boundary starting point.

In this embodiment, if all the boundary points obtained by the search do not meet the set quantity threshold and length threshold, the current boundary starting point is removed, and it is determined whether the remaining possible candidate boundary starting points obtained in step 400 are empty. If it is empty, select the next closest point as the starting point of the boundary, and go back to step 500. If it is empty, the output is empty, and then directly go to step 100 to process the static target point output by the vehicle-mounted millimeter-wave radar in the next frame.

Optionally, in the road boundary detection method based on the vehicle-mounted millimeter-wave radar in the embodiment of the present invention, in the radar coordinate system, the initial screening range is obtained by setting a horizontal threshold and a vertical threshold. Among them, the horizontal threshold is obtained according to the road width and radar detection ability in the real scene; the vertical threshold is determined according to the radar detection ability. That is, the static target points corresponding to the radar detection capability are screened out, so as to obtain reliable static target points as stable static target points. It should be noted that in this embodiment, the horizontal threshold should cover all or most of the roads with borders such as guardrails in the real scene, and the vertical threshold should cover the distance range that the radar can accurately detect.

In some embodiments, referring to FIG. 3 , step 600 of the road boundary detection method based on the vehicle-mounted millimeter-wave radar of the present invention includes:

Step 601: Acquire all boundary points obtained by searching in the previous frame;

Step 602: fuse all the boundary points obtained by the current frame search with all the boundary points obtained by the previous frame search;

Step 603: Use the least squares method to perform quadratic curve fitting on all the fused boundary points to obtain quadratic curve coefficients of the current frame.

In this embodiment, each static target point has an ID or a mark, and static target points with repeated IDs or marks are removed by fusion.

Optionally, the larger the ordinate of the starting point in the road boundary detection method based on the vehicle-mounted millimeter-wave radar in the embodiment of the present invention, the larger the horizontal and vertical widths of the corresponding preset search range. It should be noted that the correspondence between the ordinate and the preset horizontal width and vertical width of the search range can be obtained based on experience or calibration. The preset search range in this embodiment is a threshold value box, and the starting point is the starting point of the boundary, and a threshold value box is set on the left and right and upward of the threshold value box, and the next point is searched in the box.

The following is a detailed description of the implementation process of the road boundary detection method based on the vehicle-mounted millimeter-wave radar, as shown in Figure 4:

1) Obtain all static target points in the current frame of the radar;

2) In the radar coordinate system, by setting the horizontal and vertical thresholds, an initial screening range is obtained, and all input static target points are preliminarily screened to obtain stable static target points; among them, the horizontal and vertical thresholds are based on the road in the real scene. Width, radar detection ability and other indicators are determined; for example, the horizontal threshold should cover all or most of the roads with borders such as guardrails in the real scene, and the vertical threshold should cover the distance range that the radar can accurately detect. Sort by the size of the ordinate.

3) Determine the area within the range of the radar's field of view (FOV, Field of view) within the initial screening range, and set a search range on both sides of the area with the central axis of the radar as the symmetry axis. The static target point within is defined as the alternative boundary starting point of the road, and in the case that the alternative boundary starting point is non-empty, step 4) is performed; wherein, each search range takes the sector radius boundary corresponding to the field of view angle range as One of the borders and the other border are parallel to the sector radius border, and the distance between the two borders should meet the detection requirements of borders such as guardrails;

4) Calculate the distances from all the obtained alternative boundary starting points to the radar origin. Due to the characteristics of the boundary starting points, the present invention adds an influence factor when calculating the distance, and its calculation formula is shown in formula (1). :

d ² =αP _x ² +βP _y ² ---(1)

In the formula: d is the calculated distance from the starting point of the alternative boundary to the origin of the radar, α is the horizontal scale factor, β is the vertical scale factor, P _x is the abscissa of the starting point of the alternative boundary, and P _y is the alternative boundary The ordinate of the starting point of the boundary, so the value of α and β can be used to adjust the influence of P _x and P _y on d. Take the minimum value of the distance d from all alternative boundary starting points as the boundary starting point;

5) Use the obtained boundary starting point to find the next boundary point in the stable static target point within a certain horizontal and vertical range, and then use the found boundary point as a new starting point, and then search for the next boundary point, By analogy, when the next boundary point cannot be found, go to step 6);

6) Judging the validity of the boundary points found in step 5, determine whether the number and length of the boundary points found in step 5) meet the set threshold; if not, remove the current alternative boundary starting point, and determine the step Whether the remaining possible candidate boundary starting points obtained in (3) are empty, if not, select the next closest candidate boundary starting point as the boundary starting point, go back to step 5), if it is empty, output Is empty. If the found boundary points meet the requirements, then proceed to step 7);

7) Use the least squares method to perform quadratic curve fitting on the points obtained in step 6), and the fitting formula is shown in formula (2):

P _x =aP _y ² +bP _y +c---(2)

where a is the quadratic term coefficient of the fitted function, b is the linear term coefficient of the fitted function, and c is the constant term of the fitted function; P _x is the abscissa of the boundary point, and P _y is the Y-axis.

8) Then use the three coefficients a, b, and c obtained from the current frame and the previous frame to filter them with the idea of Kalman filtering to obtain a relatively stable coefficient output in the time domain.

Referring to Figure 5, taking a certain side boundary of the road as an example to explain the principle of Kalman filtering, let the quadratic curve fitting the road boundary at time k and time k-1 is shown in formula (3):

为每一帧聚类得到的所有边界点的横坐标值，n为边界点的数量。得到两个状态向量与测量向量之后，其状态方程与观测方程分别如式(4)与式(5)所示：In the formula, a _k , b _k , c _k and a _k-1 , b _k-1 , c _k-1 are the polynomial coefficients obtained by fitting at time k and time k-1 respectively, P ^k _x , P ^{k -1} _x respectively is the abscissa of the boundary point at time k and time k-1, and P ^k _y and P ^k-1 _y are the ordinate of the boundary point at time k and time k-1, respectively. Therefore, the state vector x=[abc] ^T can be set. Since the vehicle-mounted millimeter-wave radar can measure the coordinate information of the road boundary point, the measurement vector can be set.

The abscissa values of all boundary points obtained by clustering for each frame, n is the number of boundary points. After two state vectors and measurement vectors are obtained, the state equation and observation equation are shown in equations (4) and (5), respectively:

x _k =Ax _k-1 +w _k-1 ---(4)

z _k = Hx _k +v _k ---(5)

In the formula, A is the state transition matrix, W _k-1 is the covariance matrix of the predicted noise, H is the observation matrix, V _k is the covariance matrix of the measurement noise, X _k and X _k-1 are the k time and k- The state vector at time 1, Z _k and Z _k-1 are the measurement vectors at time k and time k-1, respectively. In the process of driving the car, since there will be a certain rotation angle θ between the k-1 radar coordinate system and the k-1 radar coordinate system (counterclockwise is positive, clockwise is negative), the coordinate system needs to be rotated and transformed to obtain the coordinates of k-1 at time k. The curve equation under the system. Here x _k-1 we take the state vector after coordinate transformation, therefore, the state transition matrix A is the identity matrix. The observation matrix can be obtained from the observation vector as shown in formula (6):

为k时刻状态向量的先验估计，P_k|k-1为k时刻先验估计的协方差矩阵，P_k-1为k-1时刻后验估计误差的协方差矩阵，Q为状态估计噪声的协方差矩阵，R为观测噪声的协方差矩阵。卡尔曼滤波的详细步骤为：In the formula, P ^y is the longitudinal coordinates of all boundary points obtained by clustering the boundary points of each frame, thus obtaining the necessary parameters of the Kalman filter. The specific process is shown in Figure 5. pictured

is the prior estimation of the state vector at time k, P _k|k-1 is the covariance matrix of the prior estimation at time k, P _k-1 is the covariance matrix of the posterior estimation error at time k-1, and Q is the state estimation noise The covariance matrix of , R is the covariance matrix of the observation noise. The detailed steps of the Kalman filter are:

Determine whether the state vector at the last moment (k-1) is empty, if it is empty, then determine whether the boundary point selected at the current moment (k) meets the threshold, if not, end directly, otherwise, perform quadratic function fitting, and output features;

If the state vector at time (k-1) is not empty, then judge whether the boundary point selected at time (k) satisfies the threshold, if not, end directly, otherwise, go to step 4);

The Kalman filter prediction process is performed, as shown in equations (7) and (8):

P _k|k-1 =AP _k-1 A ^T +Q---(8)

Use the boundary points screened at the current moment to generate the observation matrix H and the state vector z _k at the current moment;

The correction stage of Kalman filter, the formula is as formula (9), (10) and (11):

K _k =P _k|k-1 H ^T (HP _k|k-1 H ^T +R) ^-1 ---(9)

P _k =(IK _k H)P _k|k-1 ---(11)

为卡尔曼滤波得到的最优估计值，P_k为k时刻后验估计误差的协方差矩阵，I为1的矩阵。where K _k is the Kalman gain,

is the optimal estimated value obtained by Kalman filtering, P _k is the covariance matrix of the posterior estimation error at time k, and I is a matrix of 1.

Finally, the optimal estimated value obtained by Kalman filtering is output.

On the other hand, referring to FIG. 6 , an embodiment of the present invention further provides a road boundary detection system 1 based on a vehicle-mounted millimeter-wave radar, including:

The target point acquisition module 10 is configured to acquire all static target points of the current frame output by the vehicle-mounted millimeter-wave radar;

The preliminary screening module 20 is configured to screen out a plurality of stable static target points from all the static coordinate points according to the initial screening range, and the initial screening range is determined according to the radar detection capability;

The alternative starting point acquisition module 30 is configured to use a stable static target point located within a preset screening range as an alternative boundary starting point; wherein, taking the radar center axis as the axis of symmetry, within the range of the field of view of the vehicle-mounted millimeter-wave radar Two search ranges are set inside, and the set two search ranges are used as preset screening ranges. Each search range takes the sector radius boundary corresponding to the field of view of the vehicle-mounted millimeter-wave radar as one of the boundaries, and the other boundary is the same as the The sector radius boundaries are parallel;

The initial starting point determination module 40 is configured to sort the candidate boundary starting points according to the distance from the radar origin from small to large, and use the candidate boundary starting point with the smallest distance as the boundary starting point;

The search module 50 is configured to take the boundary starting point as the starting point, search for the boundary point from the stable static target points within the preset search range, and use the found boundary point as a new starting point to continue to search for the next boundary point, until the next boundary point cannot be found;

The curve fitting module 60 is configured to use the least squares method to perform quadratic curve fitting on all boundary points obtained by searching to obtain quadratic curve coefficients of the current frame;

And, the filtering processing module 70 is configured to perform filtering processing on the quadratic curve coefficient of the current frame and the quadratic curve coefficient of the previous frame by using the Kalman filtering algorithm, so as to obtain the target quadratic curve coefficient and the target road boundary curve.

The specific details of each module of the above-mentioned vehicle-mounted millimeter-wave radar-based road boundary detection system have been described in detail in the corresponding vehicle-mounted millimeter-wave radar-based road boundary detection method, so they will not be repeated here.

In another aspect, an embodiment of the present invention also provides an electronic device, including: a processor and a memory, where computer-readable instructions are stored in the memory, and when the computer-readable instructions are executed by the processor, the based Road boundary detection method for vehicle-mounted millimeter-wave radar.

Specifically, the above-mentioned memory and processor can be a general-purpose memory and processor, which is not specifically limited here. When the processor executes the computer-readable instructions stored in the memory, it can execute the vehicle-mounted millimeter-wave radar-based method described in the above embodiment. Road boundary detection method.

In yet another aspect, an embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the road boundary detection based on the vehicle-mounted millimeter-wave radar as described in the above-mentioned embodiment is implemented method.

Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium can include: Flash disk, read-only memory (read-only memory, ROM), random access memory (random access memory, RAM), magnetic disk or optical disk, etc.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of the present invention. within the scope of protection.

Claims (10)

1. a road boundary detection method based on vehicle-mounted millimeter wave radar, is characterized in that, comprises: Obtain all static target points of the current frame output by the vehicle millimeter-wave radar; Screen a plurality of stable static target points from all the static coordinate points according to the initial screening range, and the initial screening range is determined according to the radar detection capability; The stable static target point located in the preset screening range is used as the starting point of the alternative boundary; wherein, taking the central axis of the radar as the symmetry axis, two search ranges are set within the field of view of the vehicle-mounted millimeter-wave radar, and the The two set search ranges are used as the preset screening ranges, and each search range takes the sector radius boundary corresponding to the field of view range of the vehicle-mounted millimeter-wave radar as one of the boundaries, and the other boundary is parallel to the sector radius boundary; Sort the candidate boundary starting points according to the distance from the radar origin from small to large, and use the candidate boundary starting point with the smallest distance as the boundary starting point; Take the boundary starting point as the starting point, search for the boundary point from the stable static target point within the preset search range, and use the found boundary point as the new starting point to continue searching for the next boundary point until the search fails. until the next boundary point; Use the least squares method to perform quadratic curve fitting on all the boundary points obtained by the search to obtain the quadratic curve coefficient of the current frame; The quadratic curve coefficients of the current frame and the quadratic curve coefficients of the previous frame are filtered by using the Kalman filtering algorithm, so as to obtain the target quadratic curve coefficients and the target road boundary curve. 2. the road boundary detection method based on vehicle-mounted millimeter wave radar according to claim 1, is characterized in that, also comprises: If multiple boundary points are found within the preset search range, the distance between each boundary point and the origin of the radar is calculated; Sort all the boundary points in ascending order of the obtained distance, and use the boundary point corresponding to the minimum distance as the new starting point. 3. the road boundary detection method based on vehicle-mounted millimeter wave radar according to claim 2, is characterized in that, the distance calculation formula between boundary point and radar origin is expressed as: d ² =αP _x ² +βP _y ² ---(1) In the formula: d is the distance between the calculated boundary point and the origin of the radar, α is the horizontal scale factor, β is the vertical scale factor, P _x is the abscissa of the boundary point, and P _y is the ordinate of the boundary point. 4. the road boundary detection method based on vehicle-mounted millimeter wave radar according to claim 1, is characterized in that, also comprises: If all the boundary points obtained by the search meet the set quantity threshold and length threshold, then transfer to the step of using the least squares method to perform quadratic curve fitting on all the boundary points obtained by the search to obtain the quadratic curve coefficient of the current frame; Otherwise, the current boundary starting point is removed, and the candidate boundary starting point located next to the current boundary starting point in the sorting is used as a new boundary starting point. 5. The road boundary detection method based on the vehicle-mounted millimeter-wave radar according to claim 4, wherein, the current boundary starting point is removed, and the current boundary starting point is located next to the current boundary starting point in the sorting. The step of using the alternative boundary start point of the bit as a new boundary start point includes: count the number of the remaining candidate boundary starting points; If the number of the remaining candidate boundary starting points is greater than zero, the candidate boundary starting point located next to the current boundary starting point in the sorting is used as a new boundary starting point. 6 . The road boundary detection method based on vehicle-mounted millimeter-wave radar according to claim 1 , wherein, in the radar coordinate system, the initial screening range is obtained by setting a horizontal threshold and a vertical threshold; wherein, the horizontal threshold Obtained according to the road width and radar detection capability in the real scene; the longitudinal threshold is determined according to the radar detection capability. 7. the road boundary detection method based on vehicle-mounted millimeter-wave radar according to claim 1, is characterized in that, utilize least squares method to carry out quadratic curve fitting to obtain the quadratic curve coefficient of current frame to all boundary points that search obtains, include: Get all the boundary points found in the previous frame; Fusion of all boundary points found in the current frame and all boundary points found in the previous frame; Use the least squares method to perform quadratic curve fitting on all the fused boundary points to obtain the quadratic curve coefficients of the current frame. 8. The road boundary detection method based on the vehicle-mounted millimeter-wave radar according to any one of claims 1-7, characterized in that, the larger the ordinate of the starting point is, the greater the horizontal width and the vertical width of the corresponding preset search range bigger. 9. A road boundary detection system based on a vehicle-mounted millimeter-wave radar, characterized in that, comprising: The target point acquisition module is configured to acquire all static target points of the current frame output by the vehicle-mounted millimeter-wave radar; a preliminary screening module, configured to screen out a plurality of stable static target points from all the static coordinate points according to an initial screening range, the initial screening range being determined according to the radar detection capability; The alternative starting point acquisition module is configured to use the stable static target point located within the preset screening range as the alternative boundary starting point; wherein, taking the radar center axis as the axis of symmetry, at the field of view of the vehicle-mounted millimeter-wave radar Two search ranges are set within the range, and the two set search ranges are used as the preset screening ranges, and each search range takes the sector radius boundary corresponding to the field of view angle range of the vehicle-mounted millimeter-wave radar as one of the boundaries, and the other a boundary parallel to the sector radius boundary; The initial starting point determination module is configured to sort the candidate boundary starting points according to the distance from the radar origin from small to large, and use the candidate boundary starting point with the smallest distance as the boundary starting point; The search module is configured to take the boundary starting point as a starting point, search for a boundary point from the stable static target point within a preset search range, and use the found boundary point as a new starting point to continue searching for the next Boundary point until the next boundary point cannot be found; a curve fitting module, configured to perform quadratic curve fitting on all boundary points obtained by the least squares method to obtain quadratic curve coefficients of the current frame; And, the filtering processing module is configured to perform filtering processing on the quadratic curve coefficients of the current frame and the quadratic curve coefficients of the previous frame by using the Kalman filtering algorithm, so as to obtain the target quadratic curve coefficients and the target road boundary curve. 10. An electronic device, characterized in that it comprises: a processor and a memory, wherein computer-readable instructions are stored on the memory, and when the computer-readable instructions are executed by the processor, implementation as in claims 1 to 8 Any one of the road boundary detection methods based on vehicle-mounted millimeter-wave radar.

Images