Millimeter wave radar guardrail detection and identification method

Abstract

The invention provides a millimeter wave radar guardrail detection and identification method, which comprises the steps of receiving radar echo data, processing the radar echo data to obtain a target distance, a target speed and a target angle of a single-frame image, calculating a target coordinate and performing coordinate system conversion to obtain a plurality of static trace points; when detecting that the guardrail exists in the current frame, obtaining a boundary containing a plurality of static trace points, and judging whether the guardrail is established in the previous frame or not; if not, setting a guardrail initial value, traversing all static trace points of the current frame, taking the static trace points as expansion center points to perform guardrail expansion to obtain expansion points, recording the expansion center points and trace point information of the expansion points, and when the recorded trace point number reaches a preset number, performing polynomial fitting on the recorded static trace points to obtain guardrail information; and if yes, acquiring the previous frame of guardrail information to perform trace point absorption, and updating the guardrail according to the absorbed static trace points to obtain guardrail updating information. The method can achieve the precise recognition of the guardrail information, and is suitable for linear and curved guardrails.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle-mounted millimeter-wave radars, and in particular to a millimeter-wave radar guardrail detection and recognition method. Background Art

[0002] With the continuous development of autonomous driving technology, vehicle-mounted millimeter-wave radar has been widely used in target tracking and detection, vehicle alarm, scene recognition, etc. In actual urban roads, radar detection scenes are sometimes more complicated, with targets such as guardrails, building walls, and stationary vehicles on both sides of the road. At this time, the radar receiver will not only receive the echo of the real target, but will also most likely receive other multipath reflection echoes generated by the target, thereby detecting some false targets. These false targets will affect the radar's detection and tracking of targets. For the radar system function, false alarms and missed alarms are prone to occur, which will cause more safety hazards during vehicle driving. Therefore, how to effectively reduce these false targets is one of the important tasks in current vehicle-mounted millimeter-wave radars.

[0003] Existing false target suppression technologies can be roughly divided into model-based methods and data-driven methods. Model-based methods are commonly used to identify road boundaries, such as guardrails, tunnel boundaries, etc. Points outside the identified road boundaries can be considered as false target points. Its advantages are small computational complexity and easy understanding. Its disadvantages are that the actual road environment is changeable, it is difficult to implement a universal strategy under any conditions, and the robustness may be low. Data-driven methods are based on machine learning or deep learning methods. This method requires a large amount of computation and is difficult to implement in embedded systems. At present, the most common and easy-to-implement method is still the recognition of road boundaries.

[0004] In summary, the existing technology uses a single frame of static point traces for aggregation when judging guardrails. If there are other targets next to the radar (such as passing vehicles), the static target of the guardrail is blocked, and the static target is difficult to aggregate, making it difficult to successfully identify the guardrail. In addition, the existing technology relies more on the detection results of a single frame of point traces. If the detection of a certain frame of point traces is poor, it is also easy to cause unsuccessful guardrail detection or sudden changes in the shape and position of the guardrail. At the same time, for scenes where the vehicle turns or the guardrail is curved, guardrail recognition is more difficult. Summary of the invention

[0005] Based on this, it is necessary to provide a millimeter-wave radar guardrail detection and identification method to address the above technical problems.

[0006] A millimeter-wave radar guardrail detection and recognition method includes the following steps: receiving radar echo data, processing the echo data to obtain the target distance, target speed, and target angle of a single-frame image, calculating the coordinates of all targets and performing coordinate system conversion to obtain multiple stationary traces; when it is detected that there is a guardrail in the current frame, obtaining the boundary containing multiple stationary traces and determining whether a guardrail has been established in the previous frame; if not, setting the initial values of the guardrail, traversing all the stationary traces in the current frame, and using the stationary traces as the expansion center points to expand the guardrail to obtain expansion points, recording the trace information of the expansion center points and expansion points, and when the number of recorded traces reaches a preset number, performing polynomial fitting on the recorded stationary traces to obtain guardrail information; if so, obtaining the guardrail information of the previous frame, performing trace absorption, and updating the guardrail according to the absorbed stationary traces to obtain the guardrail update information.

[0007] In one embodiment, the step of receiving radar echo data, processing the echo data to obtain the target distance, target speed, and target angle of a single-frame image, calculating the coordinates of all targets and performing coordinate system conversion to obtain multiple stationary traces includes: receiving radar echo data, performing two-dimensional fast Fourier transform on the echo data to obtain a detection matrix; performing constant false alarm rate detection on the detection matrix to obtain over-detected targets, calculating the target distance and target speed of the over-detected targets according to the distance unit and speed unit where the over-detected targets are located; performing fast Fourier transform in the angle dimension on the over-detected targets to obtain the target angle; calculating the coordinates of all over-detected targets according to the target distance, target speed, and target angle of the over-detected targets, and converting from the radar coordinate system to the vehicle's own coordinate system to obtain multiple stationary traces.

[0008] In one embodiment, the initial values of the guardrail include the initial lateral nearest distance, initial lateral farthest distance, initial longitudinal nearest distance, and initial longitudinal farthest distance; the trace information includes the lateral distance and longitudinal distance of the trace; the guardrail information includes the guardrail lateral nearest distance, guardrail lateral farthest distance, guardrail longitudinal nearest distance, guardrail longitudinal farthest distance, guardrail parameters, the number of guardrail expansion points, guardrail length, and the left-right direction of the guardrail.

[0009] In one embodiment, expanding the guardrail with the static trace as the expansion center point to obtain expansion points, and recording the trace information of the expansion center point and the expansion points includes: selecting a static trace as the expansion center point, and setting the horizontal expansion threshold and the vertical expansion threshold of the expansion center point; traversing the static traces other than the expansion center point as alternative expansion points, calculating the horizontal distance difference and the vertical distance difference between the alternative expansion points and the expansion center point, if the horizontal distance difference is less than the horizontal expansion threshold and the vertical distance difference is less than the vertical expansion threshold, then taking the corresponding alternative expansion point as an expansion point, and recording the trace information of the expansion point; judging whether the horizontal distance of the expansion point is less than the initial horizontal nearest distance or greater than the initial horizontal farthest distance, and whether the vertical distance of the expansion point is less than the initial vertical nearest distance or greater than the initial vertical farthest distance, if so, updating the guardrail horizontal nearest distance, guardrail horizontal farthest distance, guardrail vertical nearest distance and guardrail vertical farthest distance with the vertical distance and the horizontal distance of the expansion point; traversing each expansion point, and respectively setting all expansion points as the expansion center point, and repeating the above steps to obtain all the trace information that meets the requirements in the current frame.

[0010] In one embodiment, when the number of recorded traces reaches a preset number, performing polynomial fitting on the recorded static traces to obtain guardrail information further includes: counting the total number of expansion points, and judging whether the total number of expansion points is within a preset number range; if so, performing second-order fitting on the expansion points, calculating to obtain guardrail parameters, the guardrail parameters including a first parameter, a second parameter and a third parameter, obtaining guardrail information according to the guardrail parameters, when the third parameter is less than a set parameter threshold, determining that the guardrail is a straight guardrail, performing first-order fitting on the expansion points of the guardrail to obtain guardrail information; otherwise, determining that the guardrail is a curved guardrail; if not, the expansion fails, and returning to the guardrail expansion step.

[0011] In one embodiment, the second-order fitting includes: constructing a matrix y = bX, where, y = [ypos1, ypos2,..., ypos n T , b = [coefA, coefB, coefC] T , X = [1, 1,... 1; xpos1, xpos2,..., xpos n ; xpos1 2 , xpos2 2 ,..., xpos n 2 , ypos1, ypos2, ypos n represent the vertical distances of the static traces, and xpos1, xpos2, xpos n ​Indicates the lateral distance of the stationary trace points, coefA is the first parameter, coefB is the second parameter, and coefC is the third parameter; the first parameter, the second parameter, and the third parameter are calculated using the least squares method and used as guardrail parameters.

[0012] In one of the embodiments, the first-order fitting includes: calculating the sum of the lateral distances sumXpos and the sum of the longitudinal distances sumYpos of all the extended points, and the formula is:

[0013] sumXpos = xpos1 + xpos2 + … + xpos n ;

[0014] sumYpos = ypos1 + ypos2 + … + ypos n ;

[0015] Calculating the sum of the squares of the lateral distances sumX2 and the sum of the products of the lateral distances and the longitudinal distances sumXY of all the extended points, and the formula is:

[0016] sumX2 = xpos1 2 + xpos2 2 +... + xpos n 2 ;

[0017] sumXY = xpos1 * ypos1 + xpos2 * ypos2 +... + xpos n * ypos n ;

[0018] Dividing sumXpos, sumYpos, sumX2, and sumXY by the number of extended points respectively to obtain averXpos, averYpos, aversumX2, and averXY; calculating the first parameter and the second parameter of the guardrail information according to averXpos, averYpos, aversumX2, and averXY, and the formula is:

[0019] coefA = averYpos - coefB * averXpos;

[0020] coefB = (sumXY - n * averXpos * averYpos) / (sumX2 - n * averXpos * averXpos).

[0021] In one embodiment, the obtaining of the information of the previous frame of guardrail and the point track absorption include: when the guardrail is a curved guardrail, traversing all stationary point tracks, and calculating the distance d_point_2_farest between the stationary point track and the farthest distance of the previous frame of guardrail and the distance d_point_2_nearest between the stationary point track and the nearest distance of the previous frame of guardrail. The formula is as follows;

[0022] d_point_2_farest = sqrt((xpos - maxXpos_old) 2 +(ypos - maxYpos_old) 2 )

[0023] d_point_2_nearest = sqrt((xpos - minXpos_old) 2 +(ypos - minYpos_old) 2 )

[0024] In the formula, xpos is the horizontal distance of the stationary point track, ypos is the vertical distance of the stationary point track, maxXpos_old is the farthest horizontal distance of the previous frame of guardrail, maxYpos_old is the farthest vertical distance of the previous frame of guardrail, minXpos_old is the nearest horizontal distance of the previous frame of guardrail, and minYpos_old is the nearest vertical distance of the previous frame of guardrail;

[0025] Calculate the horizontal estimated value y_esti according to the vertical distance of the stationary point track and the information of the previous frame of guardrail. The formula is:

[0026] y_esti = coefA + coefB * xpos + coefC * xpos * xpos;

[0027] In the formula, xpos represents the horizontal distance. If d_point_2_farest and d_point_2_nearest of the stationary point track are less than the set value, and the difference between the horizontal distance and the horizontal estimated value is less than the preset threshold, then absorb the stationary point track; otherwise, do not absorb it.

[0028] In one embodiment, the obtaining of the information of the previous frame of guardrail and the point track absorption include: when the guardrail is a straight guardrail, correct the coefA and coefB parameters of the previous frame of guardrail according to the ego vehicle rotation speed egoYawRate. The formula is:

[0029]

[0030] coefA′ = coefA / (cos(egoYawRate * 0.05) + coefB * sin(egoYawRate * 0.05));

[0031] In the formula, egoYawRate represents the self-vehicle rotation speed, coefA and coefB are the guardrail parameters of the previous frame, and coefA' and coefB' are the corrected parameters;

[0032] Calculate the lateral estimated value of the stationary trace according to the corrected parameters coefA' and coefB', and judge whether to absorb the corresponding stationary trace according to the lateral estimated value.

[0033] In one embodiment, updating the guardrail according to the absorbed stationary trace to obtain guardrail update information includes: judging whether the number of stationary traces absorbed in the current frame is within a preset range; if so, determining that the current frame is a valid guardrail frame, incrementing the number of valid guardrail frames by one, performing polynomial fitting on the absorbed stationary traces to obtain guardrail update information; if not, determining that the current frame is an invalid frame, judging whether the guardrail is stable according to the relationship between the number of valid guardrail frames and the valid frame threshold, if the guardrail is stable, inheriting the guardrail information of the previous frame as the guardrail update information, if the guardrail is unstable, deleting the guardrail.

[0034] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows: by receiving radar echo data and processing to obtain the target distance, target speed and target angle of a single-frame image, calculating the coordinates of all targets and performing coordinate system conversion to obtain multiple traces, when detecting that there is a guardrail in the current frame, obtaining a boundary containing multiple stationary traces and judging whether a guardrail has been established in the previous frame; if not, setting the initial value of the guardrail, traversing all the stationary traces in the current frame, and using the stationary traces as the extended center points to expand the guardrail to obtain extended points, recording the trace information of the extended center points and the extended points, when the number of recorded traces reaches the preset number, performing polynomial fitting on the trace information to obtain guardrail information, and the accurate position of the guardrail can be obtained through polynomial fitting; if so, obtaining the guardrail information of the previous frame, performing trace absorption, updating the guardrail according to the absorbed stationary traces to obtain guardrail update information, maintaining the guardrail using multi-frame information, avoiding sudden changes in the shape and position of the guardrail, improving the accuracy of guardrail detection, and being applicable to the identification of straight or curved guardrails. Description of the Drawings

[0035] Figure 1 It is a schematic flow chart of a millimeter-wave radar guardrail detection and identification method in one embodiment;

[0036] Figure 2 It is a flow block diagram of a millimeter-wave radar guardrail detection and identification method in one embodiment;

[0037] Figure 3 It is a schematic diagram of the radar installation position and stationary traces in one embodiment. Detailed Embodiment

[0038] Before describing the specific embodiments of the present invention, the overall concept of the present invention will be described as follows:

[0039] The present invention is mainly developed based on the process of guardrail detection by millimeter-wave radar. At present, single-frame information is used for clustering in guardrail recognition, resulting in low success and accuracy in guardrail recognition and difficulty in applying to curved guardrails.

[0040] Therefore, the present invention proposes a millimeter-wave radar guardrail detection and recognition method. By receiving radar echo data and processing it to obtain the target distance, target speed, and target angle of a single-frame image, calculating the coordinates of all targets and performing coordinate system conversion to obtain multiple stationary traces. When it is detected that there is a guardrail in the current frame, the boundary containing multiple stationary traces is obtained, and it is judged whether a guardrail has been established in the previous frame. If not, the initial value of the guardrail is set, all stationary traces in the current frame are traversed, and the stationary traces are used as the expansion center points for guardrail expansion to obtain expansion points. The trace information of the expansion center points and expansion points is recorded. When the number of recorded traces reaches the preset number, polynomial fitting is performed on the recorded stationary traces to obtain guardrail information. The accurate position of the guardrail can be obtained through polynomial fitting. If so, the guardrail information of the previous frame is obtained, trace absorption is performed, and the guardrail is updated according to the absorbed stationary traces to obtain guardrail update information. Multiple frames of information are used to maintain the guardrail, avoiding sudden changes in the shape and position of the guardrail, improving the accuracy of guardrail detection, and being applicable to the recognition of straight or curved guardrails.

[0041] After introducing the overall concept of the present invention, in order to make the purpose, technical solution, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below through specific embodiments in conjunction with the accompanying drawings. It should be understood that the specific embodiments described here are only used to explain the present invention and are not used to limit the present invention.

[0042] In one embodiment, as Figure 1 shown, a millimeter-wave radar guardrail detection and recognition method is provided, including the following steps:

[0043] Step S110, receive radar echo data, process the echo data to obtain the target distance, target speed, and target angle of a single-frame image, calculate the coordinates of all targets and perform coordinate system conversion to obtain multiple stationary traces.

[0044] Specifically, the radar emits electromagnetic waves and receives radar echo data. The echo data is processed to screen out targets that do not meet the requirements, and the target distance, target speed, and target angle of all over-detected targets in a single-frame image are obtained. The coordinates of the corresponding targets are calculated based on the target distance, target speed, and target angle. At this time, the coordinates are in the radar coordinate system. Coordinate system conversion is performed according to the coordinate correspondence between the radar and the vehicle body to obtain multiple stationary traces in the vehicle body coordinate system, which is convenient for subsequent guardrail detection and recognition. The process of guardrail detection and recognition is as follows Figure 2 shown

[0045] Among them, step S110 includes: receiving radar echo data, performing two-dimensional fast Fourier transform on the echo data to obtain a detection matrix; performing constant false alarm rate detection on the detection matrix to obtain over-detected targets, and calculating the target distance and target speed of the over-detected targets according to the distance unit and speed unit where the over-detected targets are located; performing fast Fourier transform on the over-detected targets in the angle dimension to obtain the target angle; calculating the coordinates of all over-detected targets according to the target distance, target speed, and target angle of the over-detected targets, and converting them from the radar coordinate system to the vehicle's own coordinate system to obtain multiple stationary traces

[0046] Specifically, the vehicle radar emits electromagnetic waves and receives echo data. A two-dimensional fast Fourier transform operation is performed on the echo data to obtain a detection matrix, which converts the input signal into a frequency-domain representation for subsequent image processing; constant false alarm rate detection is performed on the detection matrix to obtain over-detected targets, realizing target detection in a complex signal environment and excluding noise interference; according to the distance unit and speed unit where the over-detected targets are located, the target distance and target speed of the over-detected targets are calculated, and the target angle is obtained by performing fast Fourier transform on the over-detected targets in the angle dimension; the coordinates of all over-detected targets are calculated through the target distance, target speed, and target angle, and they are converted from the radar coordinate system to the vehicle's own coordinate system to obtain multiple stationary traces corresponding to all over-detected targets

[0047] Step S120, when it is detected that there is a guardrail in the current frame, obtain the boundary containing multiple stationary traces and determine whether a guardrail has been established in the previous frame

[0048] Specifically, when it is detected that there is a guardrail on the road in the current frame, a boundary containing multiple stationary traces will appear in the radar detection plane, and at the same time, it is determined whether a guardrail has been established in the previous frame for subsequent guardrail calculation and guardrail maintenance operations. As Figure 3 shown, it shows the installation method of the radar and the positions where the stationary traces of the roadside guardrail that may be observed are located

[0049] Step S130, if not, set the initial value of the guardrail, traverse all the stationary traces in the current frame, use the stationary traces as the expansion center points to expand the guardrail to obtain the expanded points, record the trace information of the expansion center points and the expanded points, and when the number of recorded traces reaches the preset number, perform polynomial fitting on the recorded stationary traces to obtain the guardrail information.

[0050] Specifically, if the guardrail has not been established, it means that the current frame is the initial frame, then set the initial value of the guardrail, traverse all the stationary traces in the current frame, use the stationary traces as the expansion center points to expand the guardrail to obtain the expanded points, and at the same time record the trace information of the expansion center points and the expanded points. When the number of recorded traces meets the preset number, perform polynomial fitting according to the recorded stationary traces to obtain the guardrail information of the current frame, so as to realize the acquisition of the guardrail information when the guardrail has not been established. Through polynomial fitting, the position of the guardrail can be accurately located, improving the accuracy of guardrail detection.

[0051] Among them, the initial value of the guardrail includes the initial minimum horizontal distance, the initial maximum horizontal distance, the initial minimum vertical distance, and the initial maximum vertical distance; the trace information includes the horizontal distance and the vertical distance of the trace; the guardrail information includes the minimum horizontal distance of the guardrail, the maximum horizontal distance of the guardrail, the minimum vertical distance of the guardrail, the maximum vertical distance of the guardrail, the guardrail parameters, the number of guardrail expansion points, the length of the guardrail, and the left - right direction of the guardrail.

[0052] Specifically, for example, the initial value of the guardrail information is set as: the initial minimum horizontal distance minYpos = 999, the initial maximum horizontal distance maxYpos = - 999, the initial minimum vertical distance minXpos = 999, and the initial maximum vertical distance maxXpos = - 999. The trace information includes: the horizontal distance ypos of the trace and the vertical distance xpos of the trace.

[0053] Among them, the calculation formula for the length of the guardrail is:

[0054] length = sqrt((maxXpos - minXpos) 2 +(maxYpos - minYpos) 2 );

[0055] Among them, the guardrail parameters include coefA, coefB, and coefC. The method for judging the left - right direction of the guardrail is: if coefA is negative, the guardrail is on the left side, otherwise it is on the right side.

[0056] Among them, the steps of using a stationary point as the expansion center point to expand the guardrail to obtain the expanded points are as follows: Select a stationary point as the expansion center point, and set the horizontal expansion threshold and vertical expansion threshold of the expansion center point; Traverse the stationary points other than the expansion center point as alternative expansion points, calculate the horizontal distance difference and vertical distance difference between the alternative expansion point and the expansion center point. If the horizontal distance difference is less than the horizontal expansion threshold and the vertical distance difference is less than the vertical expansion threshold, then use the corresponding alternative expansion point as the expansion point and record the trace information of the expansion point; Determine whether the horizontal distance of the expansion point is less than the initial horizontal nearest distance or greater than the initial horizontal farthest distance, and whether the vertical distance of the expansion point is less than the initial vertical nearest distance or greater than the initial vertical farthest distance. If so, update the guardrail horizontal nearest distance, guardrail horizontal farthest distance, guardrail vertical nearest distance, and guardrail vertical farthest distance with the vertical distance and horizontal distance of the expansion point, and set all the expansion points as the expansion center point respectively, and repeat the above steps to obtain all the trace information that meets the requirements in the current frame.

[0057] Specifically, select a stationary point from all the stationary points in the current frame as the expansion center point, and set the horizontal expansion threshold and vertical expansion threshold according to the horizontal distance and vertical distance of the expansion center point from the coordinate origin. For example, if it expands 1m horizontally and 7m vertically based on the expansion center point, then the horizontal expansion threshold is set to 1 and the vertical expansion threshold is set to 7. When setting the threshold, the closer the distance between the expansion center point and the coordinate origin, the smaller the threshold is set, and the farther the distance, the larger the threshold is set.

[0058] Traverse the stationary points other than the expansion center point in the current frame, use the remaining stationary points as alternative expansion points, calculate the horizontal distance difference and vertical distance difference between the alternative expansion point and the expansion center point. When the horizontal distance difference is less than the horizontal expansion threshold and the vertical distance difference is less than the vertical expansion threshold, then recognize the alternative expansion point as the expansion point and record the trace information of the expansion point, so as to realize the screening of the expansion points.

[0059] After screening out the expansion points, determine whether the horizontal distance of the expansion point is less than the initial horizontal nearest distance or greater than the initial horizontal farthest distance, and whether the vertical distance of the expansion point is less than the initial vertical nearest distance or greater than the initial vertical farthest distance. If the horizontal distance is less than the initial horizontal nearest distance and the vertical distance is less than the initial vertical nearest distance, or the horizontal distance is greater than the initial horizontal farthest distance and the vertical distance is greater than the initial vertical farthest distance, then update the guardrail horizontal nearest distance, guardrail horizontal farthest distance, guardrail vertical nearest distance, and guardrail vertical farthest distance with the vertical distance and horizontal distance of the expansion point to facilitate the subsequent generation of guardrail information.

[0060] To obtain sufficient expansion points, it is necessary to expand the expansion points again. By traversing the obtained expansion points, all the expansion points are respectively set as the expansion center points, and the above threshold setting, expansion point screening, and point trace information recording operations are repeated to obtain all the point trace information of the current frame.

[0061] Among them, polynomial fitting is performed on the stationary point traces to obtain guardrail information, including: counting the total number of expansion points and determining whether the total number of expansion points is within a preset quantity range; if so, second-order fitting is performed on the expansion points, the guardrail parameters are calculated, and the guardrail information is obtained according to the guardrail parameters. The guardrail parameters include the first parameter, the second parameter, and the third parameter. When the third parameter is less than the set parameter threshold, the guardrail is determined to be a straight guardrail, and first-order fitting is performed on the expansion points of the guardrail to obtain the guardrail information; otherwise, the guardrail is determined to be a curved guardrail; if not, the expansion fails and returns to the guardrail expansion step.

[0062] Specifically, after the guardrail expansion is completed, the total number of expansion points of all the obtained expansion points is counted, and it is determined whether the total number of expansion points is within a preset quantity range. For example, the preset quantity range is set to be greater than 10. When the total number of expansion points is greater than 10, it indicates that the expansion is successful. Second-order fitting is performed on the obtained expansion points to calculate the guardrail parameters, including the first parameter coefA, the second parameter coefB, and the third parameter coefC. The guardrail information can be calculated according to the guardrail parameters. When the third parameter is less than the set parameter threshold, for example, coefC < 0.001, the guardrail is determined to be a straight guardrail, and first-order fitting is performed on the guardrail again to obtain the guardrail information; otherwise, the guardrail is determined to be a curved guardrail. If it is not within the preset quantity range, it indicates that the total number of expansion points is too small and the expansion fails. Return to the guardrail expansion step for re-expansion until the total number of expansion points meets the requirements.

[0063] Among them, the method of second-order fitting is specifically: construct the matrix y = bX, where, y = [ypos1, ypos2,..., ypos n T , b = [coefA, coefB, coefC] T , X = [1, 1,...1; xpos1, xpos2,..., xpos n ; xpos1 2 , xpos2 2 ,..., xpos n 2 , ypos1, ypos2, ypos n represent the longitudinal distances of the stationary point traces, and xpos1, xpos2, xpos n ​Indicates the lateral distance of the stationary point. coefA is the first parameter, coefB is the second parameter, and coefC is the third parameter. The first parameter, the second parameter, and the third parameter of the coefficient are calculated by the least squares method and used as the guardrail parameters.

[0064] Among them, the method of first-order fitting is specifically as follows: Calculate the sum of the lateral distances sumXpos and the sum of the longitudinal distances sumYpos of all extended points. The formula is:

[0065] sumXpos = xpos1 + xpos2 + … + xpos n ;

[0066] sumYpos = ypos1 + ypos2 + … + ypos n ;

[0067] Calculate the sum of the squares of the lateral distances sumX2 and the sum of the products of the lateral distances and the longitudinal distances sumXY of all extended points. The formula is:

[0068] sumX2 = xpos1 2 + xpos2 2 +... + xpos n 2 ;

[0069] sumXY = xpos1 * ypos1 + xpos2 * ypos2 +... + xpos n * ypos n ;

[0070] Divide sumXpos, sumYpos, sumX2, and sumXY by the number of extended points respectively to obtain averXpos, averYpos, aversumX2, and averXY;

[0071] Calculate the first parameter and the second parameter of the guardrail information according to averXpos, averYpos, aversumX2, and averXY. The formula is:

[0072] coefA = averYpos - coefB * averXpos;

[0073] coefB = (sumXY - n * averXpos * averYpos) / (sumX2 - n * averXpos * averXpos).

[0074] Specifically, since the guardrails on both sides of the road may be straight guardrails or curved guardrails, second-order polynomial fitting can be performed first. If the quadratic term coefficient is too small, first-order polynomial fitting is then used to achieve the identification of both curved and straight guardrails with high accuracy.

[0075] Step S140, if so, obtain the guardrail information of the previous frame, perform track absorption, and update the guardrail based on the stationary tracks after absorption to obtain the guardrail update information.

[0076] Specifically, if the guardrail has been established, it means that the current frame is not the initial frame. Since when the vehicle is driving on a road with guardrails, the attributes such as the length, position, curvature or straightness of the guardrail do not change suddenly, the stationary tracks of the current frame can be clustered according to the guardrail information of the previous frame. By obtaining the guardrail information of the previous frame, track absorption is performed on multiple stationary tracks of the current frame. If the number of absorbed tracks is small, it means that the tracks of the current frame cannot be successfully fitted to the guardrail, and it is determined as an invalid frame; if the number of absorbed tracks is within the preset range, the current frame is determined as a valid frame, and the guardrail is updated based on the stationary tracks after absorption to obtain the guardrail update information, thus avoiding sudden changes in the position and shape of the guardrail, improving the accuracy of guardrail detection, and the method of updating the guardrail based on the guardrail information of the previous frame has high robustness.

[0077] Among them, the track absorption steps for the curved guardrail are specifically as follows: when the guardrail is a curved guardrail, all stationary tracks are traversed, and the distance d_point_2_farest between the stationary track and the farthest distance of the previous frame of the guardrail and the distance d_point_2_nearest between the stationary track and the nearest distance of the previous frame of the guardrail are calculated. The formula is;

[0078] d_point_2_farest = sqrt((xpos - maxXpos_old) 2 +(ypos - maxYpos_old) 2 )

[0079] d_point_2_nearest = sqrt((xpos - minXpos_old) 2 +(ypos - minYpos_old) 2 )

[0080] In the formula, xpos is the horizontal distance of the stationary track, ypos is the vertical distance of the stationary track, maxXpos_old is the farthest horizontal distance of the previous frame of the guardrail, maxYpos_old is the farthest vertical distance of the previous frame of the guardrail, minXpos_old is the nearest horizontal distance of the previous frame of the guardrail, and minYpos_old is the nearest vertical distance of the previous frame of the guardrail;

[0081] Calculate the lateral estimation value y_esti based on the longitudinal distance of the static trace and the guardrail information of the previous frame. The formula is as follows:

[0082] y_esti = coefA + coefB * xpos + coefC * xpos * xpos;

[0083] In the formula, xpos represents the lateral distance. If the d_point_2_farest and d_point_2_nearest of the static trace are less than the set value, and the difference between the lateral distance and the lateral estimation value is less than the preset threshold, then absorb this static trace; otherwise, do not absorb it.

[0084] Specifically, traverse all static traces, calculate the distance between the static trace and the farthest distance of the guardrail in the previous frame and the distance between the static trace and the nearest distance of the guardrail in the previous frame, and calculate the lateral estimation value of this static trace based on the longitudinal distance and the guardrail information of the previous frame. If the values of d_point_2_farest and d_point_2_nearest of this point are small, and the actual lateral distance of this point differs little from the estimated lateral distance y_esti (the threshold can be set according to the longitudinal distance of this point), then absorb this static trace; otherwise, do not absorb it. When setting the preset threshold, set it according to the longitudinal distance of the static trace. The larger the longitudinal distance, the larger the corresponding preset threshold.

[0085] Among them, the specific steps for absorbing the trace of the straight guardrail are as follows: When the guardrail is a straight guardrail, correct the coefA and coefB parameters of the previous frame of the guardrail according to the ego vehicle rotation speed egoYawRate. The formula is as follows:

[0086]

[0087] coefA' = coefA / (cos(egoYawRate * 0.05) + coefB * sin(egoYawRate * 0.05));

[0088] In the formula, egoYawRate represents the ego vehicle rotation speed, coefA and coefB are the parameters of the previous frame of the guardrail, and coefA' and coefB' are the corrected parameters; calculate the lateral estimation value of the static trace according to the corrected parameters coefA' and coefB', and judge whether to absorb the corresponding static trace according to the lateral estimation value.

[0089] Specifically, since the still point trace parameters of the guardrail are different when it is straight and when it is curved, the parameters of the guardrail in the previous frame need to be corrected according to the vehicle speed of the host vehicle to obtain the corrected parameters coefA' and coefB'. The remaining point trace absorption strategy is the same as that of the curved guardrail. The lateral distance value of the still point trace is calculated according to the corrected parameters, and it is determined whether to absorb the corresponding still point trace according to the lateral estimation value.

[0090] Among them, the method for obtaining the guardrail update information includes: determining whether the number of still point traces absorbed in the current frame is within a preset range; if so, determining that the current frame is a valid guardrail frame, incrementing the number of valid guardrail frames by one, and performing polynomial fitting on the absorbed still point traces to obtain the guardrail update information; if not, determining that the current frame is an invalid frame, and determining whether the guardrail is stable according to the relationship between the number of valid guardrail frames and the valid frame threshold. If the guardrail is stable, inherit the guardrail information of the previous frame as the guardrail update information. If the guardrail is unstable, delete the guardrail.

[0091] Specifically, after point trace absorption, determine whether the number of still point traces absorbed in the current frame is within a preset range, for example, greater than 5. If it is greater than 5, determine that the current frame is a valid frame, increment the number of valid guardrail frames by one, and perform polynomial fitting (second-order fitting and / or first-order fitting) on the absorbed still point traces. The fitting method is the same as the previous one to obtain a new value and use it to update the guardrail information to obtain the guardrail update information.

[0092] If it is less than 5, determine that the current frame is an invalid frame, and determine whether the guardrail is stable according to the relationship between the number of valid frames and the valid frame threshold (for example, 6). When the number of valid guardrail frames exceeds 6, determine that the guardrail is stable, and inherit the guardrail information of the previous frame as the guardrail update information. Otherwise, determine that the guardrail is unstable and delete the guardrail.

[0093] The above steps record the determination situations of multiple frames of guardrails. When multiple consecutive frames of guardrails are stable, the guardrail will not suddenly disappear. Therefore, when the number of valid guardrail frames is sufficient, even if the radar target detection situation of a certain frame is poor, or the radar is blocked by other targets on the road and the still point traces of the current frame cannot fit the guardrail, the guardrail information of the previous frame can still be inherited, which has high robustness.

[0094] In this embodiment, the target distance, target speed, and target angle of a single-frame image are obtained by processing the received radar echo data. The coordinates of all targets are calculated and coordinate system conversion is performed to obtain multiple traces. When a guardrail is detected in the current frame, a boundary containing multiple stationary traces is obtained, and it is determined whether a guardrail has been established in the previous frame. If not, the initial value of the guardrail is set, all stationary traces in the current frame are traversed, and the stationary traces are used as the expansion center points to expand the guardrail to obtain expansion points. The trace information of the expansion center points and expansion points is recorded. When the number of recorded traces reaches the preset number, polynomial fitting is performed on the recorded stationary traces to obtain guardrail information. The accurate position of the guardrail can be obtained through polynomial fitting. If so, the guardrail information of the previous frame is obtained, trace absorption is performed, and the guardrail is updated according to the absorbed stationary traces to obtain guardrail update information. Multiple-frame information is used to maintain the guardrail, avoiding sudden changes in the shape and position of the guardrail, improving the accuracy of guardrail detection, and being applicable to the identification of straight or curved guardrails.

[0095] The above technical solution uses multiple-frame information to maintain the guardrail. Since when a vehicle target travels on a road with a guardrail, the attributes such as the length, position, curvature / straightness of the guardrail do not change suddenly. Therefore, using this attribute, the guardrail information of the previous frame is used to cluster the stationary traces of the current frame, so as to control the guardrail information of the current frame to be similar to that of the previous frame, avoiding sudden changes in the shape and position of the guardrail, and improving the accuracy of guardrail detection.

[0096] In one embodiment, when it is necessary to identify a guardrail in a complex scene, such as a multi-layer overpass, tunnel, or complex curve environment, it can be judged according to the trace disorder degree, generate a coefficient measuring the trace disorder degree, and determine whether to start guardrail identification or set the guardrail identification in a specific scene according to this coefficient.

[0097] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0098] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. Optionally, they can be implemented by program codes executable by the computing device. Thus, they can be stored in a computer storage medium (ROM / RAM, magnetic disk, optical disk) and executed by the computing device. And in some cases, the steps shown or described can be executed in a sequence different from that here, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module for implementation. Therefore, the present invention is not limited to any specific combination of hardware and software.

[0099] The above content is a further detailed description of the present invention in combination with specific implementation manners. It cannot be determined that the specific implementation of the present invention is only limited to these descriptions. For those of ordinary skill in the technical field to which the present invention pertains, without departing from the concept of the present invention, several simple deductions or substitutions can still be made, and all should be regarded as belonging to the protection scope of the present invention.

Claims

1. A millimeter wave radar guardrail detection and identification method, characterized in that: The following steps are involved: Receiving radar echo data, processing the echo data to obtain target distance, target speed and target angle of a single frame image, calculating the coordinates of all targets and performing coordinate system conversion to obtain multiple stationary point traces; When it is detected that there is a guardrail in the current frame, a boundary including multiple stationary point traces is obtained, and it is determined whether a guardrail has been established in the previous frame; If not, set the guardrail initial value, traverse all the static point traces of the current frame, and use the static point trace as the extension center point to perform guardrail extension to obtain the extension point, record the point trace information of the extension center point and the extension point, and when the number of recorded point traces reaches the preset number, perform polynomial fitting on the recorded static point traces to obtain the guardrail information; If so, the guardrail information of the previous frame is obtained, point trace absorption is performed, and the guardrail is updated according to the static point traces after absorption to obtain the guardrail update information.

2. The millimeter wave radar guardrail detection and identification method according to claim 1 is characterized in that: The receiving radar echo data, processing the echo data to obtain the target distance, target speed and target angle of a single frame image, calculating the coordinates of all targets and performing coordinate system conversion to obtain multiple stationary point traces, including: Receiving radar echo data, performing a two-dimensional fast Fourier transform on the echo data to obtain a detection matrix; Performing constant false alarm detection on the detection matrix to obtain an over-detected target, and calculating a target distance and a target speed of the over-detected target according to a distance unit and a speed unit where the over-detected target is located; Performing angle-dimensional fast Fourier transform on the inspected target to obtain the target angle; According to the target distance, target speed and target angle of the inspected target, the coordinates of all inspected targets are calculated and converted from the radar coordinate system to the vehicle's own coordinate system to obtain multiple stationary point traces.

3. The millimeter wave radar guardrail detection and identification method according to claim 1 is characterized in that: The guardrail initial values ​​include an initial lateral shortest distance, an initial lateral farthest distance, an initial longitudinal shortest distance and an initial longitudinal farthest distance; The point trace information includes the horizontal distance and the vertical distance of the point trace; The guardrail information includes the closest horizontal distance of the guardrail, the farthest horizontal distance of the guardrail, the closest vertical distance of the guardrail, the farthest vertical distance of the guardrail, guardrail parameters, the number of guardrail extension points, guardrail length and left and right direction of the guardrail.

4. The millimeter wave radar guardrail detection and identification method according to claim 3 is characterized in that: The step of using the stationary point trace as the extension center point to perform guardrail extension to obtain an extension point, and recording the point trace information of the extension center point and the extension point, comprises: Selecting a stationary point as an expansion center point, and setting a horizontal expansion threshold and a vertical expansion threshold of the expansion center point; Traversing the stationary point traces except the extension center point as candidate extension points, calculating the horizontal distance difference and the vertical distance difference between the candidate extension point and the extension center point, if the horizontal distance difference is less than the horizontal extension threshold, and the vertical distance difference is less than the vertical extension threshold, taking the corresponding candidate extension point as the extension point, and recording the point trace information of the extension point; Determine whether the lateral distance of the extension point is less than the initial lateral closest distance or greater than the initial lateral farthest distance, and whether the longitudinal distance of the extension point is less than the initial longitudinal closest distance or greater than the initial longitudinal farthest distance. If so, update the guardrail lateral closest distance, guardrail lateral farthest distance, guardrail longitudinal closest distance and guardrail longitudinal farthest distance with the longitudinal distance and lateral distance of the extension point; Traverse each extension point and set all extension points as extension center points respectively, repeat the above steps to obtain all point trace information that meets the requirements in the current frame.

5. The millimeter wave radar guardrail detection and identification method according to claim 4 is characterized in that: When the number of recorded point traces reaches a preset number, polynomial fitting is performed on the recorded static point traces to obtain guardrail information, and further includes: Counting the total number of extension points, and determining whether the total number of extension points is within a preset number range; If yes, perform a second-order fitting on the extension point to calculate the guardrail parameters, which include the first parameter, the second parameter and the third parameter. The guardrail information is obtained according to the guardrail parameters. When the third parameter is less than the set parameter threshold, the guardrail is determined to be a straight guardrail, and the first-order fitting is performed on the extension point of the guardrail to obtain the guardrail information; otherwise, the guardrail is determined to be a curved guardrail. If not, the expansion fails and returns to the guardrail expansion step.

6. The millimeter wave radar guardrail detection and identification method according to claim 5 is characterized in that: The second-order fitting includes: Construct the matrix y=bX, where y=[ypos1,ypos2,...,ypos n ] T ,b=[coefA,coefB,coefC] T ,X=[1,1,...1;xpos1,xpos2,...,xpos n ;xpos1 2 ,xpos2 2 ,...,xpos n 2 ], ypos1, ypos2, ypos n Indicates the longitudinal distance of the stationary point trace, xpos1, xpos2, xpos n Indicates the lateral distance of the stationary point trace, coefA is the first parameter, coefB is the second parameter, and coefC is the third parameter; The first parameter, the second parameter and the third parameter are calculated by the least square method as the guardrail parameters.

7. The millimeter wave radar guardrail detection and identification method according to claim 6 is characterized in that: The first-order fitting includes: Calculate the sum of the horizontal distances sumXpos and the sum of the vertical distances sumXpos of all extension points. The formula is: sumXpos=xpos1+xpos2+…+xpos n ; sumYpos=ypos1+ypos2+…+ypos n ; Calculate the sum of the squares of the horizontal distances of all extension points sumX2 and the product of the horizontal distance and the vertical distance sumXY. The formula is: sumX2=xpos1 2 +xpos2 2 +...+xpos n 2 ; sumXY=xpos1*ypos1+xpos2*ypos2+...+xpos n *ypos n ; Divide sumXpos, sumYpos, sumX2 and sumXY= by the number of extension points to obtain averXpos, averYpos, aversumX2 and averXY; The first and second parameters of the guardrail information are calculated according to averXpos, averYpos, aversumX2 and averXY. The formula is: coefA=averYpos-coefB*averXpos; coefB=(sumXY-n*averXpos*averYpos) / (sumX2-n*averXpos*averXpos).

8. The millimeter wave radar guardrail detection and identification method according to claim 7 is characterized in that: The obtaining of the guardrail information of the previous frame and performing point trace absorption includes: When the guardrail is a curved guardrail, traverse all static point traces, calculate the distance d_point_2_farest between the static point trace and the guardrail of the previous frame and the distance d_point_2_nearest between the static point trace and the guardrail of the previous frame, and the formula is; d_point_2_farest=sqrt((xpos-maxXpos_old) 2 +(ypos-maxYpos_old) 2 ) d_point_2_nearest=sqrt((xpos-minXpos_old) 2 +(ypos-minYpos_old) 2 ) Where xpos is the horizontal distance of the stationary point trace, ypos is the vertical distance of the stationary point trace, maxXpos_old is the farthest horizontal distance of the guardrail in the previous frame, maxYpos_old is the farthest vertical distance of the guardrail in the previous frame, minXpos_old is the shortest horizontal distance of the guardrail in the previous frame, and minYpos_old is the shortest vertical distance of the guardrail in the previous frame; The lateral estimated value y_esti is calculated based on the longitudinal distance of the stationary point trace and the guardrail information of the previous frame. The formula is: y_esti=coefA+coefB*xpos+coefC*xpos*xpos; Where xpos represents the lateral distance. If d_point_2_farest and d_point_2_nearest of the stationary point are less than the set value, and the difference between the lateral distance and the lateral estimated value is less than the preset threshold, the stationary point is absorbed, otherwise it is not absorbed.

9. The millimeter wave radar guardrail detection and identification method according to claim 8, characterized in that: The obtaining of the guardrail information of the previous frame and performing point trace absorption includes: When the guardrail is a straight guardrail, the coefA and coefB parameters of the guardrail in the previous frame are corrected according to the egoYawRate of the vehicle. The formula is: coefA′=coefA / (cos(egoYawRate*0.05)+coefB*sin(egoYawRate*0.05)); Where egoYawRate represents the ego vehicle speed, coefA and coefB are the guardrail parameters of the previous frame, and coefA′ and coefB′ are the corrected parameters; The lateral estimation value of the stationary point track is calculated according to the corrected parameters coefA′ and coefB′, and it is determined whether to absorb the corresponding stationary point track according to the lateral estimation value.

10. The millimeter wave radar guardrail detection and identification method according to claim 8, characterized in that: The step of updating the guardrail according to the absorbed static point traces to obtain guardrail update information includes: Determine whether the number of static traces absorbed by the current frame is within a preset range; If yes, the current frame is considered as a guardrail valid frame, the number of guardrail valid frames is increased by one, and the absorbed static point trace is fitted with a polynomial to obtain the guardrail update information; If not, the current frame is deemed to be an invalid frame, and whether the guardrail is stable is determined based on the relationship between the guardrail valid frame number and the valid frame threshold. If the guardrail is stable, the guardrail information of the previous frame is inherited as the guardrail update information. If the guardrail is unstable, the guardrail is deleted.

Images