Method for rejecting static target based on RANSAC (Random Sample Consensus) algorithm

Abstract

The invention relates to a method for eliminating a static target based on an RANSAC (Random Sample Consensus) algorithm, which comprises the following steps that: a millimeter-wave radar transmits electromagnetic waves and receives echo signals, and a point cloud data set is obtained through preprocessing; adopting an RANSAC algorithm to calculate binomial model parameters, namely the number of corresponding inner points; taking the binomial model parameter corresponding to the maximum number of the inner points as a final binomial model parameter; calculating inner points in the target point cloud data set P through the final binomial model parameters, and outputting the finally reserved inner points as a filtered point cloud set; according to the method, the binomial model of the target speed and angle is fitted through the RANSAC algorithm, and the point clouds which do not meet the binomial model in the target point cloud set are eliminated, so that the dynamic target caused by the motion of the vehicle can be identified and eliminated more accurately, the interference of the environmental dynamic target on the target tracking is effectively avoided, and the target tracking accuracy is improved. And the detection precision and reliability of the real moving target by the vehicle-mounted radar are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of millimeter wave radar, and in particular to a method for eliminating static targets based on a RANSAC algorithm. Background Art

[0002] With the continuous development of automobile intelligence and autonomous driving technology, on-board radar, as an important environmental perception sensor, can monitor the environment around the vehicle in real time and provide accurate data support for the vehicle's autonomous driving system to ensure driving safety. However, in actual application scenarios, due to the movement of the vehicle itself, the on-board radar will identify its own static target as a dynamic point cloud. This misidentification will interfere with the radar's detection of real moving targets, thereby affecting the target tracking process behind the point cloud, resulting in target tracking abnormalities and reducing the performance and safety of the vehicle's autonomous driving system. At present, the commonly used methods for eliminating static targets include: calculating the speed component based on the vehicle speed and angle, and judging whether the target is stationary by comparing the speed measured by the radar with the theoretically calculated speed. This method directly uses the vehicle's own speed and target angle information, and can quickly judge whether the target is stationary through simple calculations. However, this method has high requirements for the accuracy of speed and angle measurement, requires accurate speed information, and usually requires additional hardware devices or communication interfaces to obtain speed data. In addition, there is a zero-speed channel zeroing method, which reduces the influence of static targets by identifying and eliminating the reflected signal in the zero-speed area in the range-Doppler map. Similarly, this method has high requirements for the measurement accuracy of Doppler frequency shift. Summary of the invention

[0003] In view of this, the present invention provides a method for eliminating static targets based on the RANSAC algorithm. The binomial model of target speed and angle is fitted by the RANSAC algorithm to eliminate dynamic targets caused by the vehicle's own motion. Under the premise of effectively solving the target tracking anomaly caused by environmental static targets when tracking point clouds, the complexity and cost of the system are reduced and the adaptability and reliability are enhanced.

[0004] To achieve the above object, the present invention provides a method for eliminating static targets based on the RANSAC algorithm, comprising the following steps: S1, the millimeter wave radar transmitting antenna emits electromagnetic waves outward, the receiving antenna receives the echo signal, and pre-processes the echo signal to obtain the target point cloud data set P; S2, using the random sampling consensus RANSAC algorithm to obtain the binomial model of the target's speed and angle; S3, obtaining the inliers in the target point cloud data set P according to the binomial model, and recording the values ​​of the inliers in the current iteration and the corresponding binomial model parameters through the inlier counter numPtsInlier; S4, repeating steps S2 and S3 until the number of iterations reaches a preset number, comparing the values ​​of the inliers recorded by the inlier counter numPtsInlier, and taking the binomial model parameter corresponding to the maximum position of the inlier value as the final binomial model parameter ; S5. According to the binomial model parameters obtained in step S4 Calculate the internal points in the target point cloud data set P and output the final target point cloud set; S501, traversing the target point cloud dataset P; S502, calculation point The predicted value is calculated as follows: ; in, represents the i-th point cloud, Indicates The horizontal angle of the point cloud; S503, calculation point Predicted values ​​and points Speed ​​Difference , the expression is: ; S504, the difference Compare with the preset threshold, if the difference If it is less than the preset threshold, it is judged as an inlier, and the inlier counter numPtsInlier is incremented by 1, that is: ; Record the value of numPtsInlier; S505: Output the internal points obtained in step S504 as the filtered target point cloud data set.

[0005] Preferably, the pretreatment comprises the following steps: S101, after the echo signal is mixed, the module on-chip chip performs AD sampling on the mixed signal; S102, using a distance processing algorithm 1D-FFT to obtain a one-dimensional range image of the target, multiple radars transmit electromagnetic wave chirps to obtain multiple one-dimensional range images; S103, obtaining target velocity data through a Doppler processing algorithm 2D-FFT of multiple one-dimensional range images; S104, synthesizing all acquired one-dimensional range images and velocity data to generate a range Doppler detection matrix; S105 , performing constant false alarm processing on the range Doppler detection matrix, and estimating the data angles of different receiving channels at each distance to obtain a target point cloud data set P.

[0006] Preferably, the target point cloud data set P is expressed as: ; ; in, Indicates the number of point clouds in the current frame, Indicates distance, Indicates the horizontal angle, Indicates the pitch angle, represents the Doppler velocity, Indicates signal strength.

[0007] Preferably, the step of obtaining the binomial model comprises: S201: Randomly extract three points from the point cloud data set P , , , and generate the matrix , the expression is: ; in, Indicates the horizontal angle; S202, calculation matrix The inverse matrix , the expression is: ; S203, obtaining binomial model parameters.

[0008] Preferably, the expression of the binomial model parameter is: ; in, represents the binomial model parameters, .

[0009] Preferably, obtaining the interior points in the target point cloud data set P comprises the following steps: S301, traversing the target point cloud dataset P; S302, calculation point The predicted value is calculated as follows: ; in, represents the i-th point cloud, Indicates The horizontal angle of the point cloud; S303, calculation point Predicted values ​​and points Speed ​​Difference , the expression is: ; S304, the difference Compare with the preset threshold, if the difference If it is less than the preset threshold, it is judged as an inlier, and the inlier counter numPtsInlier is incremented by 1, that is: ; Record the value of numPtsInlier and the corresponding binomial model parameters.

[0010] Preferably, the final binomial model parameters The expression is: ; in, Represents the binomial model parameters when numPtsInlier is the largest during the iteration.

[0011] Compared with the prior art, the present invention has the following beneficial effects: The present invention fits a binomial model of target speed and angle through the RANSAC algorithm, and eliminates point clouds that do not satisfy the binomial model in the target point cloud set. It can more accurately identify and eliminate dynamic targets caused by the vehicle's own motion, effectively avoid the interference of environmental dynamic targets on target tracking, improve the detection accuracy and reliability of vehicle-mounted radar for real moving targets, and improve the performance and safety of the autonomous driving system. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 This is a flow chart of eliminating static targets in the present invention. DETAILED DESCRIPTION

[0013] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments.

[0014] In order to eliminate dynamic targets caused by the operation of the vehicle itself and avoid target tracking anomalies caused by identifying environmental static targets as dynamic point clouds when tracking point cloud sets, the present invention provides the following technical solutions: a method for eliminating static targets based on the RANSAC algorithm, comprising the following steps: S1, the millimeter wave radar transmitting antenna emits electromagnetic waves outward, the receiving antenna receives the echo signal, and pre-processes the echo signal to obtain the target point cloud data set P; The preprocessing includes the following steps: S101, after the echo signal is mixed, the module on-chip chip performs AD sampling on the mixed signal; S102, using a distance processing algorithm 1D-FFT to obtain a one-dimensional range image of the target, multiple radars transmit electromagnetic wave chirps to obtain multiple one-dimensional range images; S103, obtaining target velocity data through a Doppler processing algorithm 2D-FFT of multiple one-dimensional range images; S104, synthesizing all acquired one-dimensional range images and velocity data to generate a range Doppler detection matrix; S105, performing constant false alarm processing on the range Doppler detection matrix, and estimating the data angles of different receiving channels at each distance to obtain a target point cloud data set P, which is expressed as: ; ; in, Indicates the number of point clouds in the current frame, Indicates distance, Indicates the horizontal angle, Indicates the pitch angle, represents the Doppler velocity, Indicates signal strength.

[0015] S2, using the random sampling consensus RANSAC algorithm to obtain the binomial model of the target's speed and angle; S201, three points are randomly selected from the point cloud dataset P , , , and generate the matrix , the expression is: ; in, Indicates the horizontal angle; S202, calculation matrix The inverse matrix , the expression is: ; S203, obtain binomial model parameters, the expression is: ; in, represents the binomial model parameters, .

[0016] S3, obtaining the inliers in the target point cloud data set P according to the binomial model, and recording the values ​​of the inliers in the current iteration and the corresponding binomial model parameters through the inlier counter numPtsInlier; S301, traversing the target point cloud dataset P; S302, calculation point The predicted value is calculated as follows: ; in, represents the i-th point cloud, Indicates The horizontal angle of the point cloud; S303, calculation point Predicted values ​​and points Speed ​​Difference , the expression is: ; S304, Difference Compare with the preset threshold, if the difference If it is less than the preset threshold, it is judged as an inlier, and the inlier counter numPtsInlier is incremented by 1, that is: ; The value of the inlier counter for this iteration is recorded in numPtsInlier[j]; the binomial model parameters are recorded in , , In , j represents the iteration number, j={1,2,......,k}, and k is the preset number of iterations.

[0017] S4, repeating steps S2 and S3 until the number of iterations reaches a preset number, comparing the values ​​of the inliers recorded by the inlier counter numPtsInlier, and taking the binomial model parameter corresponding to the maximum position of the inlier value as the final binomial model parameter , the expression is; ; in, Indicates numPtsInlier during iteration Parameters of the binomial model at their maximum value.

[0018] S5. According to the binomial model parameters obtained in step S4 Calculate the internal points in the target point cloud data set P and output the final target point cloud set; S501, traversing the target point cloud dataset P; S502, calculation point The predicted value is calculated as follows: ; in, represents the i-th point cloud, Indicates The horizontal angle of the point cloud; S503, calculation point Predicted values ​​and points Speed ​​Difference , the expression is: ; S504, Difference Compare with the preset threshold, if the difference If it is less than the preset threshold, it is judged as an inlier, and the inlier counter numPtsInlier is incremented by 1, that is: ; Record the value of numPtsInlier, which is the number of point clouds after filtering; S505: Output the internal points obtained in step S504 as the filtered target point cloud data set.

[0019] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for eliminating static targets based on the RANSAC algorithm, characterized in that: The following steps are involved: S1, the millimeter wave radar transmitting antenna emits electromagnetic waves outward, the receiving antenna receives the echo signal, and pre-processes the echo signal to obtain the target point cloud data set P; S2, using the random sampling consensus RANSAC algorithm to obtain the binomial model of the target's speed and angle; S3, obtaining the inliers in the target point cloud data set P according to the binomial model, and recording the values ​​of the inliers in the current iteration and the corresponding binomial model parameters through the inlier counter numPtsInlier; S4, repeating steps S2 and S3 until the number of iterations reaches a preset number, comparing the values ​​of the inliers recorded by the inlier counter numPtsInlier, and taking the binomial model parameter corresponding to the maximum position of the inlier value as the final binomial model parameter ; S5. According to the binomial model parameters obtained in step S4 Calculate the internal points in the target point cloud data set P and output the final target point cloud set; S501, traversing the target point cloud dataset P; S502, calculation point The predicted value is calculated as follows: ； in, represents the i-th point cloud, Indicates The horizontal angle of the point cloud; S503, calculation point Predicted values ​​and points Speed ​​Difference , the expression is: ； S504, the difference Compare with the preset threshold, if the difference If it is less than the preset threshold, it is judged as an inlier, and the inlier counter numPtsInlier is incremented by 1, that is: ； Record the value of numPtsInlier; S505: Output the internal points obtained in step S504 as the filtered target point cloud data set.

2. The method for eliminating static targets based on the RANSAC algorithm according to claim 1, characterized in that: The pre-processing comprises the following steps: S101, after the echo signal is mixed, the module on-chip chip performs AD sampling on the mixed signal; S102, using a distance processing algorithm 1D-FFT to obtain a one-dimensional range image of the target, multiple radars transmit electromagnetic wave chirps to obtain multiple one-dimensional range images; S103, obtaining target velocity data through a Doppler processing algorithm 2D-FFT of multiple one-dimensional range images; S104, synthesizing all acquired one-dimensional range images and velocity data to generate a range Doppler detection matrix; S105 , performing constant false alarm processing on the range Doppler detection matrix, and estimating the data angles of different receiving channels at each distance to obtain a target point cloud data set P.

3. The method for eliminating static targets based on the RANSAC algorithm according to claim 2, characterized in that: The expression of the target point cloud dataset P is: ； ； in, Indicates the number of point clouds in the current frame, Indicates distance, Indicates the horizontal angle, Indicates the pitch angle, represents the Doppler velocity, Indicates signal strength.

4. The method for eliminating static targets based on the RANSAC algorithm according to claim 1, characterized in that: The steps of obtaining the binomial model include: S201: Randomly extract three points from the point cloud data set P , , , and generate the matrix , the expression is: ； in, Indicates the horizontal angle; S202, calculation matrix The inverse matrix , the expression is: ； S203, obtaining binomial model parameters.

5. The method for eliminating static targets based on the RANSAC algorithm according to claim 4, characterized in that: The expression of the binomial model parameters is: ； in, represents the binomial model parameters, .

6. The method for eliminating static targets based on the RANSAC algorithm according to claim 1, characterized in that: Obtaining the internal points and inliers in the target point cloud dataset P includes the following steps: S301, traversing the target point cloud dataset P; S302, calculation point The predicted value is calculated as follows: ； in, represents the i-th point cloud, Indicates The horizontal angle of the point cloud; S303, calculation point Predicted values ​​and points Speed ​​Difference , the expression is: ； S304, the difference Compare with the preset threshold, if the difference If it is less than the preset threshold, it is judged as an inlier, and the inlier counter numPtsInlier is incremented by 1, that is: 。 7. The method for eliminating static targets based on the RANSAC algorithm according to claim 1, characterized in that: Final binomial model parameters The expression is: ； in, Represents the binomial model parameters when numPtsInlier is the largest during the iteration.

Images