Vehicle interference suppression method and system for lightweight synthetic aperture traffic slope monitoring radar

Abstract

The invention discloses a vehicle interference suppression method and system for a lightweight synthetic aperture traffic slope monitoring radar, and the method comprises the steps: collecting a monitoring image through a radar, calculating the pixel-level similarity of corresponding sub-regions of two adjacent images through an NCC normalization cross-correlation function, extracting the motion vector distribution in the sub-regions through an optical flow method, and carrying out the detection of the motion vector distribution. Quantifying dynamic characteristics of vehicle interference to obtain region-level motion characteristics; constructing a multi-scale interference discrimination model based on a random forest algorithm, inputting the pixel-level correlation degree and the region-level motion features into the multi-scale interference discrimination model, and adjusting an interference determination threshold value; when the relevancy is lower than a judgment threshold value, the sub-image area is judged to be an available sub-image area, and the dynamic amplitude deviation of potential scattering points is calculated; and eliminating unstable potential scattering points through the dynamic amplitude deviation, and generating a target slope deformation monitoring result. The reliability of traffic slope monitoring is improved, and the requirement of real-time monitoring can be met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of traffic slope monitoring, and in particular to a vehicle interference suppression method and system for a lightweight synthetic aperture traffic slope monitoring radar. BACKGROUND

[0002] In the field of traffic slope monitoring, lightweight synthetic aperture radar is widely used due to its high resolution, all-day, all-weather advantages. However, there are many vehicles on the traffic slope, which will cause serious interference to the radar measurement results, mainly in the form of shielding and saturation, which will lead to fluctuations in the monitoring scene data and greatly reduce the accuracy and reliability of the monitoring results. Currently, there is a lack of effective means to suppress vehicle interference in traffic slope monitoring radar, which cannot meet the needs of high-precision traffic slope monitoring. SUMMARY

[0003] The present application aims to solve the above problems and provides a vehicle interference suppression method and system for a lightweight synthetic aperture traffic slope monitoring radar.

[0004] To achieve the above-mentioned purpose, the technical solution of the present application is as follows: further, in the vehicle interference suppression method for a lightweight synthetic aperture traffic slope monitoring radar, the vehicle interference suppression method comprises the following steps:

[0005] The radar collects monitoring images, and after preprocessing, the monitoring images are divided into multiple sub-image regions. The NCC normalized cross-correlation function is used to calculate the pixel-level similarity of the corresponding sub-regions of two adjacent images to obtain the pixel-level correlation degree.

[0006] The motion vector distribution in the sub-region is extracted by the optical flow method to quantify the dynamic characteristics of vehicle interference and obtain the regional motion feature.

[0007] A multi-scale interference discrimination model is constructed based on the random forest algorithm, the pixel-level correlation degree and the regional motion feature are input into the multi-scale interference discrimination model, and the interference discrimination threshold is adjusted.

[0008] When the correlation degree is lower than the discrimination threshold, it is determined that the sub-image region is usable, and the dynamic amplitude dispersion of the potential scattering points is calculated.

[0009] The unstable potential scattering points are removed by the dynamic amplitude dispersion to generate the target slope deformation monitoring result.

[0010] Further, in the vehicle interference suppression method for the lightweight synthetic aperture traffic slope monitoring radar, the monitoring image is collected by the radar, preprocessed, and divided into multiple sub-image regions, the pixel-level similarity of corresponding sub-regions of adjacent two images is calculated by using the NCC normalized cross-correlation function, and pixel-level correlation is obtained, including:

[0011] A plurality of monitoring images collected by the lightweight synthetic aperture traffic slope monitoring radar are obtained, and each image is divided into a plurality of sub-image regions;

[0012] The pixel-level similarity of corresponding sub-regions of adjacent two images is calculated by using the NCC normalized cross-correlation function, and pixel-level correlation is obtained, and the calculation formula is as follows:

[0013] NCC(i,j) = ∑x,y(I 1(x,y)-I 1)2∑x,y(I 2(x,y)-I 2)2∑x,y(I 1(x,y)-I 1)(I 2(x,y)-I 2);

[0014] Wherein, I 1(x,y) and I 2(x,y) are pixel values of corresponding sub-image regions at coordinates (x,y) in adjacent two images, and I 1 and I 2 are pixel mean values of corresponding sub-image regions.

[0015] Further, in the vehicle interference suppression method for the lightweight synthetic aperture traffic slope monitoring radar, the motion vector distribution in the sub-region is extracted by the optical flow method, the dynamic characteristics of the vehicle interference are quantified, and the regional motion feature is obtained, including:

[0016] The plurality of images collected by the lightweight synthetic aperture radar are processed to obtain each sub-image region, and the coverage range of the motion vector in the sub-image region is counted;

[0017] The motion vector field of each pixel point in the sub-image region is calculated by using the Horn-Schunck optical flow algorithm, and the motion trajectory and displacement size are composed;

[0018] If the motion vector direction in the sub-region tends to be consistent, it indicates that the region exists vehicle interference; if the vector direction is chaotic, it indicates that it is a natural scene;

[0019] The abnormal motion mode is identified by analyzing the time sequence of the motion vector, and the regional motion feature is obtained.

[0020] Further, in the vehicle interference suppression method for the lightweight synthetic aperture traffic slope monitoring radar, the pixel-level correlation and the regional motion feature are input into the multi-scale interference discrimination model based on the random forest algorithm, the interference discrimination threshold is adjusted, and the multi-scale interference discrimination model is constructed, including:

[0021] A multi-scale interference discrimination model is constructed based on a random forest algorithm, the pixel-level correlation degree and the region-level motion feature are input into the multi-scale interference discrimination model, and the nonlinear classification capability of the model is used to identify the boundary between the interference region and the non-interference region.

[0022] The discrimination rule of the interference region is learned through the training data, the interference probability of each sub-region is output, and the interference determination threshold is dynamically adjusted according to the interference probability.

[0023] Further, in the vehicle interference suppression method of the above-mentioned lightweight synthetic aperture traffic side slope monitoring radar, when the correlation degree is lower than the determination threshold, the sub-image region is determined to be usable, and the dynamic amplitude deviation of the potential scattering point is calculated, comprising:

[0024] The determination threshold is set, when the correlation degree of a certain sub-image region is lower than the determination threshold, it is determined that the sub-image region is seriously interfered by the vehicle and has low usability; when the correlation degree is higher than the determination threshold, it is determined that the sub-image region has high usability.

[0025] Further, in the vehicle interference suppression method of the above-mentioned lightweight synthetic aperture traffic side slope monitoring radar, when the correlation degree is lower than the determination threshold, the sub-image region is determined to be usable, and the dynamic amplitude deviation of the potential scattering point is calculated, further comprising:

[0026] The potential scattering point is amplitude compensated according to the overall image amplitude, and the potential scattering point is estimated through the amplitude deviation: for the sub-image region with high usability, the overall image amplitude is calculated, and the overall image amplitude is the sum of the amplitude values of all pixel points in the sub-image region;

[0027] The potential scattering point is determined, the potential scattering point is amplitude compensated according to the overall image amplitude, and the compensation formula is: Acompensated=Aoriginal+α×(Atotal-Aavg);

[0028] Wherein, Acompensate is the compensated amplitude value of the potential scattering point, Aoriginal is the original amplitude value of the potential scattering point, α is the amplitude compensation coefficient, Atotal is the overall image amplitude of the sub-image region, and Aavg is the average value of the amplitude values of all pixel points in the sub-image region.

[0029] Further, in the vehicle interference suppression method of the above-mentioned lightweight synthetic aperture traffic side slope monitoring radar, the unstable potential scattering point is removed through the dynamic amplitude deviation, and the target side slope deformation monitoring result is generated, comprising:

[0030] The stability of the potential scattering point is estimated by calculating the dispersion of the amplitude value of the potential scattering point, and the amplitude dispersion calculation formula is S = n1 i = 1 ∑n (Ai-A) 2;

[0031] Wherein, S is the amplitude dispersion, n is the number of potential scattering points, Ai is the amplitude value of the i th potential scattering point, and A is the average value of the amplitude values of all potential scattering points.

[0032] When the amplitude dispersion is less than the set dispersion threshold, it is determined that the potential scattering point is stable, and is regarded as a reliable monitoring point; when the amplitude dispersion is greater than the dispersion threshold, the potential scattering point is generated or excluded, and the target slope deformation monitoring result is generated.

[0033] Further, in a vehicle interference suppression system of a light-weight synthetic aperture traffic slope monitoring radar, the vehicle interference suppression system comprises the following modules:

[0034] A monitoring image acquisition module is configured to acquire a monitoring image by a radar, divide the monitoring image into a plurality of sub-image regions after pre-processing, calculate the pixel-level similarity of the corresponding sub-regions of two adjacent images by using an NCC normalized cross-correlation function, and obtain a pixel-level correlation degree.

[0035] A motion feature calculation module is configured to extract the motion vector distribution in the sub-region by using an optical flow method, quantify the dynamic characteristics of vehicle interference, and obtain a region-level motion feature.

[0036] A determination threshold adjustment module is configured to construct a multi-scale interference discrimination model based on a random forest algorithm, input the pixel-level correlation degree and the region-level motion feature into the multi-scale interference discrimination model, and adjust an interference determination threshold.

[0037] An amplitude dispersion calculation module is configured to determine a usable sub-image region when the correlation degree is lower than the determination threshold, and calculate the dynamic amplitude dispersion of the potential scattering point.

[0038] A monitoring result generation module is configured to exclude unstable potential scattering points by using the dynamic amplitude dispersion, and generate a target slope deformation monitoring result.

[0039] Further, in a vehicle interference suppression system of a light-weight synthetic aperture traffic slope monitoring radar, the amplitude dispersion calculation module comprises the following sub-modules:

[0040] A compensation sub-module is configured to perform amplitude compensation on the potential scattering point according to the overall image amplitude, and estimate the potential scattering point by using the amplitude dispersion: for a sub-image region with high availability, the overall image amplitude is calculated, and the overall image amplitude is the sum of the amplitude values of all pixel points in the sub-image region.

[0041] Acompensate=Aori g i na l+α×(Atotal-Aavg);

[0042] Acompensate=Aori g i na l+α×(Atotal-Aavg);

[0043] Wherein, Acompensate is the amplitude value of the potential scattering point after compensation, Aori g i na l is the original amplitude value of the potential scattering point, alpha is the amplitude compensation coefficient, Atota l is the overall image amplitude of the sub-image region, and Aavg is the average value of the amplitude values of all pixel points in the sub-image region.

[0044] Further, in a vehicle interference suppression system of a light-weight synthetic aperture traffic side slope monitoring radar, the voice data encryption module comprises the following sub-modules:

[0045] The transmission sub-module is configured to send single-photon signals with four different polarization states at the sending end based on the BB84 protocol.

[0046] The calculation sub-module is configured to estimate the stability of the potential scattering point by calculating the dispersion of the amplitude value of the potential scattering point, and the amplitude dispersion calculation formula is S=n1 i=1∑n(Ai-A)2.

[0047] Wherein, S is the amplitude dispersion, n is the number of potential scattering points, Ai is the amplitude value of the i-th potential scattering point, and A is the average value of the amplitude values of all potential scattering points.

[0048] The generation sub-module is configured to determine that the potential scattering point is stable when the amplitude dispersion is less than a set dispersion threshold value, and to generate a target side slope deformation monitoring result when the amplitude dispersion is greater than the dispersion threshold value.

[0049] The beneficial effects of the present application are that, through image correlation analysis, the present application can accurately identify the area severely disturbed by vehicles, avoid using the data of the disturbed area for monitoring analysis, and improve the usability and accuracy of the monitoring data. The present application compensates the amplitude of the PS (potential scattering point) based on the overall image amplitude, estimates the PS point through the amplitude dispersion, effectively eliminates the influence of vehicle interference on the amplitude of the PS point, ensures the stability of the PS point, and thus improves the reliability of traffic side slope monitoring, and can more accurately reflect the actual state of the side slope. The present application is suitable for a light-weight synthetic aperture traffic side slope monitoring radar, has low computational complexity, can meet the demand for real-time monitoring, and has good practicability and promotional value. BRIEF DESCRIPTION OF DRAWINGS

[0050] Various other advantages and benefits will become apparent to those of ordinary skill in the art, upon reading the following detailed description of the preferred embodiment. The accompanying drawings are included to provide a description of preferred embodiments, and are not intended to limit the present application.

[0051] Figure 1 A first embodiment schematic view of a vehicle interference suppression method of a lightweight synthetic aperture traffic side slope monitoring radar according to an embodiment of the present application is shown in FIG. 1.

[0052] Figure 2 A first embodiment schematic view of a vehicle interference suppression method of a lightweight synthetic aperture traffic side slope monitoring radar according to an embodiment of the present application is shown in FIG. 1.

[0053] Figure 3 A first embodiment schematic view of a vehicle interference suppression system of a lightweight synthetic aperture traffic side slope monitoring radar according to an embodiment of the present application is shown in FIG. 1. DETAILED DESCRIPTION

[0054] In order to make the objectives, technical solutions and advantages of the present application clearer, further detailed description will be given below in combination with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not intended to limit the present application.

[0055] Those skilled in the art can understand that the singular forms "a", "an" and "the" used herein include plural forms, unless specifically stated otherwise. It should be further understood that the use of the term "including" in the specification of the present application means that the features, integers, steps, operations, elements and / or components described herein are present, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0056] The present application will be described in detail below in combination with the accompanying drawings, as shown in Figure 1 A vehicle interference suppression method of a lightweight synthetic aperture traffic side slope monitoring radar according to an embodiment of the present application is shown in FIG. 1, which includes the following steps:

[0057] Step 101, acquire monitoring images by radar, divide the monitoring images into multiple sub-image regions after pre-processing, calculate the pixel-level similarity of the corresponding sub-regions of the adjacent two images by using NCC normalized cross-correlation function, and obtain the pixel-level correlation degree;

[0058] Specifically, in the present embodiment, multiple monitoring images acquired by the lightweight synthetic aperture traffic side slope monitoring radar are obtained, and each image is divided into multiple sub-image regions;

[0059] The pixel-level similarity of the corresponding sub-regions of the two adjacent images is calculated by using the NCC normalized cross-correlation function to obtain the pixel-level correlation degree, and the calculation formula is as follows:

[0060] NCC(i,j) = ∑x,y(I 1(x,y)-I 1)2∑x,y(I 2(x,y)-I 2)2∑x,y(I 1(x,y)-I 1)(I 2(x,y)-I 2);

[0061] Wherein, I 1(x,y) and I 2(x,y) are respectively the pixel values of the corresponding sub-image regions in the adjacent two images at coordinates (x,y), and I 1 and I 2 are respectively the pixel mean values of the corresponding sub-image regions.

[0062] Step 102, the motion vector distribution in the sub-region is extracted by the optical flow method, the dynamic characteristics of the vehicle interference are quantified, and the regional motion features are obtained;

[0063] Specifically, in the embodiment, a plurality of image frames collected by a lightweight synthetic aperture radar are processed to obtain each sub-image region, and the coverage range of the motion vector in the sub-image region is counted.

[0064] The motion vector field of each sub-image region is calculated by using the Horn-Schunck optical flow algorithm, and the motion trajectory and displacement size are composed.

[0065] If the motion vector direction in the sub-region tends to be consistent, it indicates that the region exists vehicle interference; if the vector direction is chaotic, it indicates that it is a natural scene.

[0066] The time sequence of the motion vector is analyzed to identify the abnormal motion mode, and the regional motion features are obtained.

[0067] Specifically,

[0068] Objective: The motion vector distribution in the sub-image region is extracted by the optical flow method, the dynamic characteristics of the vehicle interference are quantified, so as to assist in identifying the interference region and improving the suppression effect.

[0069] 1. Application principle of optical flow method

[0070] The optical flow method is a motion analysis technology based on continuous image frames, which calculates the displacement change of the pixel points in the adjacent frames to infer the motion trajectory and speed of the object. In the traffic slope monitoring scene, the movement of the vehicle will cause the pixel change in the local region of the radar image, and the optical flow method can capture such dynamic change.

[0071] Specific steps:

[0072] Input data: Process multiple frames of images collected by a lightweight synthetic aperture radar (SAR), such as a continuous time series of radar images, focusing on each sub-image region.

[0073] Motion vector calculation: For each sub-region, use an optical flow algorithm (Horn-Schunck or Lucas-Kanade) to calculate the motion vector field of the pixels within the region. The motion vector consists of direction (motion trajectory) and speed (displacement size), which can reflect the dynamic behavior of the vehicle in that region.

[0074] 2. Quantifying the dynamic characteristics of vehicle interference

[0075] Motion vector distribution analysis:

[0076] Direction consistency: If the motion vector directions in a sub-region tend to be consistent (e.g., the vehicle moves in the same direction), it may indicate that there is vehicle interference in that region; if the vector directions are chaotic, it may be a natural scene (vegetation swaying) or random noise.

[0077] Speed distribution: Vehicles usually have high motion speed (relative to the static background of the slope), so you can judge whether there is a high-speed moving interference source by counting the average speed or maximum speed of the motion vectors in the sub-region.

[0078] Motion pattern recognition: By analyzing the time series of motion vectors, identify abnormal motion patterns (sudden acceleration, deceleration, or direction mutation), which may correspond to the start, brake, or turn behavior of the vehicle.

[0079] Interference intensity evaluation:

[0080] Motion vector density: Count the coverage of motion vectors in the sub-region (i.e., the proportion of pixels participating in motion). High-density motion vectors may indicate that the vehicle has a large range of occlusion and serious interference.

[0081] Dynamic change amplitude: Compare the differences in motion vectors between consecutive frames. If the difference is significant (e.g., the vector direction or speed changes dramatically), it may indicate that the vehicle is moving quickly or entering / leaving the monitoring area.

[0082] 3. Combination with interference suppression

[0083] Interference region determination: Combine the motion features extracted by RLMF with pixel-level correlation (PLC) to form a multi-scale criterion. For example:

[0084] If the PLC value of a sub-region is low (pixel similarity is poor), but RLMF shows high-density, high-speed motion vectors, it can be determined that the region is interfered by a vehicle.

[0085] If the PLC value is high (pixel similarity is good), but the RLMF shows that the motion vector direction is chaotic or the speed is abnormal, it may be a misjudgment caused by occlusion or saturation.

[0086] Adaptive adjustment: dynamically adjust the interference judgment threshold according to the dynamic characteristics of RLMF. For example:

[0087] In areas with high vehicle density (peak hours), increase the motion vector density threshold to avoid misjudgment;

[0088] In low-speed vehicle scenarios (slow-moving road sections), lower the speed threshold to enhance the sensitivity to slow-moving interference.

[0089] Step 103, based on the random forest algorithm, a multi-scale interference discrimination model is constructed, and the pixel-level correlation and regional-level motion features are input into the multi-scale interference discrimination model, and the interference judgment threshold is adjusted;

[0090] Specifically, in this embodiment, a multi-scale interference discrimination model is constructed based on the random forest algorithm, and pixel-level correlation and regional-level motion features are input into the multi-scale interference discrimination model; the nonlinear classification ability of the model is used to identify the boundary between the interference region and the non-interference region;

[0091] Through training data, the discrimination rules of the interference region are learned, and the interference probability of each sub-region is output, and the interference judgment threshold is dynamically adjusted according to the interference probability.

[0092] Specifically,

[0093] Objective: Through the construction of a multi-scale interference discrimination model, the pixel-level correlation (PLC) and the regional-level motion feature (RLMF) are used as input features, and the random forest algorithm is used to dynamically adjust the interference judgment threshold, so as to realize the efficient identification and suppression of vehicle interference.

[0094] 1. Construction of multi-scale interference discrimination model

[0095] Model selection:

[0096] Random forest (RandomForest) algorithm is used to construct the discrimination model. Random forest is a classification method based on ensemble learning, which can effectively process high-dimensional nonlinear data through the voting mechanism of multiple decision trees, and has strong anti-noise ability and generalization performance.

[0097] Advantages: Compared with traditional single threshold judgment methods (fixed threshold or simple statistical methods), random forest can automatically learn the complex correlation of multi-scale features and adapt to different interference patterns in different scenarios.

[0098] Feature input:

[0099] The following two types of features are used as input of the model:

[0100] Pixel-level correlation (PLC):

[0101] The pixel similarity of corresponding sub-regions in adjacent radar images is calculated by the normalized cross-correlation function (NCC), reflecting the stability of static or slowly changing regions.

[0102] For example, if the PLC value of a certain sub-region is low, it indicates that the region may have occlusion or saturation phenomenon (vehicle passing).

[0103] Region-level motion feature (RLMF):

[0104] The distribution characteristics of motion vectors (direction consistency, speed distribution, motion pattern) within the sub-region are extracted by the optical flow method, quantifying the dynamic characteristics of vehicle interference.

[0105] For example, if the RLMF of a certain sub-region shows high-density and high-speed motion vectors, it may indicate the presence of fast-moving vehicle interference.

[0106] Feature fusion:

[0107] PLC and RLMF are input as multi-dimensional feature vectors into the random forest model, using its non-linear classification ability to identify the boundaries between interference regions and non-interference regions.

[0108] For example, the model can learn that the feature combination of "low PLC value and high RLMF speed" usually corresponds to vehicle interference, while the feature combination of "high PLC value and low RLMF density" usually corresponds to stable background (slope fixed structure).

[0109] 2. Adjustment of adaptive interference judgment threshold

[0110] Limitations of traditional methods:

[0111] In the prior art, interference judgment usually relies on a fixed threshold (fixed lower limit value of PLC), but the fixed threshold is difficult to adapt to different environmental conditions (light changes, vehicle density fluctuations).

[0112] Adaptive ability of random forest:

[0113] The model learns the discrimination rules of interference regions through training data (containing labeled interference / non-interference samples), and outputs the interference probability (0-1 interval value) of each sub-region.

[0114] Adjust the judgment threshold dynamically according to the interference probability:

[0115] Threshold = μ + λ·σ, where μ is the mean of the interference probability in the current scene, σ is the standard deviation, and λ is a adjustable parameter (set according to actual requirements).

[0116] Examples:

[0117] In a dense vehicle scene, lambda increases, the threshold increases, and the misjudgment is reduced;

[0118] In a sparse vehicle scene, lambda decreases, the threshold decreases, and the sensitivity to weak interference is enhanced.

[0119] Real-time optimization:

[0120] The random forest model completes complex calculations in the training stage, and only needs to perform efficient feature extraction and classification operations in runtime, meeting the real-time requirements of lightweight radar systems.

[0121] Step 104, when the correlation is lower than the determination threshold, determine that the sub-image region is usable, and calculate the dynamic amplitude dispersion of the potential scattering points;

[0122] Specifically, in the present embodiment, the determination threshold is set, when the correlation of a certain sub-image region is lower than the determination threshold, it is determined that the sub-image region is severely interfered by vehicles and has low usability; when the correlation is higher than the determination threshold, it is determined that the sub-image region has high usability.

[0123] According to the image overall amplitude, the amplitude of the potential scattering points is compensated, and the potential scattering points are estimated by the amplitude dispersion: for the sub-image region with high usability, the image overall amplitude is calculated, which is the sum of the amplitude values of all pixel points in the sub-image region;

[0124] Determine the potential scattering points, compensate the amplitude of the potential scattering points according to the image overall amplitude, and the compensation formula is: Acompensated=Aoriginal+α×(Atotal-Aavg);

[0125] Wherein, Acompensate is the compensated amplitude value of the potential scattering points, Aoriginal is the original amplitude value of the potential scattering points, alpha is the amplitude compensation coefficient, Atotal is the image overall amplitude of the sub-image region, and Aavg is the average value of the amplitude values of all pixel points in the sub-image region.

[0126] Step 105, remove unstable potential scattering points by dynamic amplitude dispersion to generate target slope deformation monitoring results.

[0127] Specifically, in the present embodiment, the stability of the potential scattering points is estimated by calculating the dispersion of the amplitude values of the potential scattering points, and the amplitude dispersion calculation formula is: S=n1 i=1∑n(Ai-A)2;

[0128] Wherein, S is the amplitude dispersion, n is the number of potential scattering points, Ai is the amplitude value of the ith potential scattering point, and A is the average value of all potential scattering point amplitude values.

[0129] When the amplitude deviation is less than the set deviation threshold, it is determined that the potential scattering point is stable and is a reliable monitoring point; when the amplitude deviation is greater than the deviation threshold, the potential scattering point is removed or rejected, and a target slope deformation monitoring result is generated.

[0130] The beneficial effects are that the image correlation degree analysis can accurately identify the area seriously disturbed by vehicles, avoid using the data of the disturbed area for monitoring analysis, and improve the availability and accuracy of the monitoring data. The PS (potential scattering point) point is compensated based on the overall amplitude of the image, and the PS point is estimated through the amplitude deviation, effectively eliminating the influence of vehicle disturbance on the amplitude of the PS point, ensuring the stability of the PS point, and improving the reliability of the traffic slope monitoring, which can more accurately reflect the actual state of the slope. The method is suitable for a lightweight synthetic aperture traffic slope monitoring radar, has low computational complexity, can meet the demand of real-time monitoring, and has good practicability and popularization value.

[0131] As shown in Figure 2 A vehicle disturbance suppression method of a lightweight synthetic aperture traffic slope monitoring radar, the vehicle disturbance suppression method comprises the following steps of:

[0132] Step 201, processing multiple frames of images collected by a lightweight synthetic aperture radar to obtain each sub-image region, and counting the coverage range of the motion vector in the sub-image region;

[0133] Step 202, calculating the motion vector field of each pixel point in each sub-image region by using a Horn-Schunck optical flow algorithm, and obtaining a motion vector motion trajectory and a displacement size group;

[0134] Step 203, if the motion vector direction in the sub-region tends to be consistent, it indicates that the region is disturbed by vehicles; if the vector direction is chaotic, it indicates that it is a natural scene;

[0135] Step 204, identifying an abnormal motion mode by analyzing the time sequence of the motion vector to obtain a regional motion feature.

[0136] The above describes an embodiment of the vehicle disturbance suppression method of the lightweight synthetic aperture traffic slope monitoring radar, please refer to Figure 3 In a vehicle disturbance suppression system of a lightweight synthetic aperture traffic slope monitoring radar, the vehicle disturbance suppression system comprises the following modules:

[0137] The monitoring image acquisition module is configured to acquire monitoring images by using a radar, and divide the monitoring images into a plurality of sub-image regions after pre-processing the monitoring images; the pixel-level similarity of corresponding sub-regions of adjacent two images is calculated by using an NCC normalized cross-correlation function to obtain a pixel-level correlation degree;

[0138] The motion feature calculation module is configured to extract a motion vector distribution in the sub-region by using an optical flow method, quantize a dynamic characteristic of vehicle interference, and obtain a region-level motion feature.

[0139] The determination threshold adjustment module is configured to construct a multi-scale interference discrimination model based on a random forest algorithm, input the pixel-level correlation degree and the region-level motion feature into the multi-scale interference discrimination model, and adjust an interference determination threshold.

[0140] The amplitude dispersion calculation module is configured to determine a usable sub-image region when the correlation degree is lower than the determination threshold, and calculate a dynamic amplitude dispersion of a potential scattering point.

[0141] The monitoring result generation module is configured to remove unstable potential scattering points by using the dynamic amplitude dispersion, and generate a target side slope deformation monitoring result.

[0142] Specifically, the embodiment further includes

[0143] A vehicle interference suppression method for a light-weight synthetic aperture traffic side slope monitoring radar, and specific steps are as follows: analyzing the usability according to an image correlation degree:

[0144] A plurality of monitoring images acquired by the light-weight synthetic aperture traffic side slope monitoring radar are obtained, and each image is divided into a plurality of sub-image regions.

[0145] The correlation degree of corresponding sub-image regions in adjacent two images is calculated, and the correlation degree is calculated by using a normalized cross-correlation function, and the formula is as follows:

[0146] NCC(i,j)=∑x,y(I 1(x,y)-I 1)2∑x,y(I 2(x,y)-I 2)2∑x,y(I 1(x,y)-I 1)(I 2(x,y)-I 2) where I 1(x,y) and I 2(x,y) are pixel values of corresponding sub-image regions at coordinates (x, y) in adjacent two images, and I 1 and I 2 are pixel mean values of the corresponding sub-image regions.

[0147] A correlation degree threshold is set, when the correlation degree of a certain sub-image region is lower than the threshold, it is determined that the sub-image region is seriously interfered by a vehicle and has low usability; when the correlation degree is higher than the threshold, it is considered that the sub-image region has high usability and can be used for subsequent monitoring analysis.

[0148] The PS points are compensated according to the overall amplitude of the image, and the PS points are estimated through amplitude deviation: for a sub-image area with high availability, the overall amplitude of the image is calculated, which is the sum of the amplitude values of all pixel points in the sub-image area;

[0149] The potential scattering points (PS points) are determined, and the PS points are compensated according to the overall amplitude of the image. The specific compensation formula is:

[0150] Acompensated=Aori g i na l+α×(Atotal-Aavg) wherein, Acompensate is the amplitude value of the compensated PS point, Aori g i nal is the original amplitude value of the PS point, α is the amplitude compensation coefficient, Atota l is the overall amplitude of the image of the sub-image area, and Aavg is the average value of the amplitude values of all pixel points in the sub-image area.

[0151] The stability of the PS points is estimated by calculating the deviation of the amplitude values of the PS points, and the amplitude deviation calculation formula is: S = n1 i = 1 ∑n(Ai-A)2 wherein, S is the amplitude deviation, n is the number of PS points, Ai is the amplitude value of the i th PS point, and A is the average value of all PS point amplitude values.

[0152] When the amplitude deviation is less than the set deviation threshold, the PS point is considered stable and can be used as a reliable monitoring point; when the amplitude deviation is greater than the deviation threshold, the PS point is further processed or removed to ensure the accuracy of the monitoring result.

[0153] The basic principles, main features and advantages of the present application are shown and described above. It should be understood by those skilled in the art that the present application is not limited by the above examples, and the above examples and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A vehicle interference suppression method for a light-weight synthetic aperture traffic-side slope monitoring radar, characterized by, The vehicle interference suppression method comprises the following steps: The vehicle interference suppression method comprises the following steps: The radar is used to collect monitoring images, and the monitoring images are divided into multiple sub-image regions after pretreatment, and the pixel-level similarity of the corresponding sub-regions of adjacent two images is calculated by using the NCC normalized cross-correlation function to obtain the pixel-level correlation degree; The motion vector distribution in the sub-region is extracted by using the optical flow method, the dynamic characteristics of the vehicle interference are quantified, and the regional motion features are obtained; A multi-scale interference discrimination model is constructed based on the random forest algorithm, the pixel-level correlation degree and the regional motion features are input into the multi-scale interference discrimination model, and the interference judgment threshold is adjusted; When the correlation degree is lower than the judgment threshold, the sub-image region is determined to be usable, and the dynamic amplitude deviation of the potential scattering points is calculated; 2. The vehicle clutter suppression method for a light-weight synthetic aperture traffic side slope monitoring radar of claim 1, wherein The unstable potential scattering points are removed by the dynamic amplitude deviation, and the target slope deformation monitoring result is generated. The radar is used to collect monitoring images, and the monitoring images are divided into multiple sub-image regions after pretreatment, and the pixel-level similarity of the corresponding sub-regions of adjacent two images is calculated by using the NCC normalized cross-correlation function to obtain the pixel-level correlation degree, comprising: A plurality of monitoring images collected by a lightweight synthetic aperture traffic slope monitoring radar are obtained, and each image is divided into multiple sub-image regions; 3. The method of claim 1, wherein the vehicle clutter is suppressed by The pixel-level similarity of the corresponding sub-regions of adjacent two images is calculated by using the NCC normalized cross-correlation function to obtain the pixel-level correlation degree. The motion vector distribution in the sub-region is extracted by using the optical flow method, the dynamic characteristics of the vehicle interference are quantified, and the regional motion features are obtained, comprising: A plurality of images collected by a lightweight synthetic aperture radar are processed to obtain each sub-image region, and the coverage range of the motion vector in the sub-image region is counted; The motion vector field, motion trajectory and displacement size of each pixel point in each sub-image region are calculated by using the Horn-Schunck optical flow algorithm; If the motion vector direction in the sub-region tends to be consistent, it indicates that the region is interfered by a vehicle; if the vector direction is chaotic, it indicates a natural scene; 4. The method of claim 1, wherein the vehicle clutter is suppressed by a light weight synthetic aperture traffic side slope monitoring radar. The time sequence of the motion vector is analyzed to identify the abnormal motion mode, and the regional motion features are obtained. A multi-scale interference discrimination model is constructed based on the random forest algorithm, the pixel-level correlation degree and the regional motion features are input into the multi-scale interference discrimination model, and the interference judgment threshold is adjusted, comprising: A multi-scale interference discrimination model is constructed based on the random forest algorithm, the pixel-level correlation degree and the regional motion features are input into the multi-scale interference discrimination model; the nonlinear classification ability of the model is used to identify the boundary between the interference region and the non-interference region; 5. The method of claim 1, wherein the vehicle clutter is suppressed by a light weight synthetic aperture traffic side slope monitoring radar. The discrimination rules of the interference region are learned by using the training data, the interference probability of each sub-region is output, and the interference judgment threshold is dynamically adjusted according to the interference probability. The pixel-level correlation degree and the regional motion features are input into the multi-scale interference discrimination model, and the interference judgment threshold is adjusted, comprising: The judgment threshold is set, when the correlation degree of a sub-image region is lower than the judgment threshold, it is determined that the sub-image region is seriously interfered by a vehicle and has low usability; when the correlation degree is higher than the judgment threshold, it is determined that the sub-image region has high usability.

6. The method of claim 1, wherein the vehicle clutter is suppressed by a light weight synthetic aperture traffic side slope monitoring radar. The sub-image region is determined as available when the correlation is lower than the determination threshold, and the dynamic amplitude deviation of the potential scattering point is calculated, and the method further comprises: The potential scattering point is amplitude-compensated according to the image overall amplitude, and the potential scattering point is estimated through the amplitude deviation: for the sub-image region with higher availability, the image overall amplitude is calculated, and the image overall amplitude is the sum of amplitude values of all pixel points in the sub-image region; The potential scattering point is determined, and the potential scattering point is amplitude-compensated according to the image overall amplitude, and the compensation formula is: Acompensated=Aoriginal+α×(Atotal-Aavg); Wherein, Acompensate is the compensated amplitude value of the potential scattering point, Aoriginal is the original amplitude value of the potential scattering point, α is the amplitude compensation coefficient, Atotal is the image overall amplitude of the sub-image region, and Aavg is the average value of the amplitude values of all pixel points in the sub-image region.

7. The method of claim 1, wherein the vehicle clutter is suppressed by a light weight synthetic aperture traffic side slope monitoring radar. The unstable potential scattering point is removed through the dynamic amplitude deviation, and a target slope deformation monitoring result is generated, and the method comprises: The stability of the potential scattering point is estimated by calculating the deviation of the amplitude value of the potential scattering point; When the amplitude deviation is less than a set deviation threshold, the potential scattering point is determined as stable, and is used as a reliable monitoring point; when the amplitude deviation is greater than the deviation threshold, the potential scattering point is removed, and a target slope deformation monitoring result is generated.

8. A vehicle interference suppression system for a lightweight synthetic aperture traffic side slope monitoring radar, characterized by The vehicle interference suppression system comprises the following modules: A monitoring image acquisition module is configured to acquire monitoring images by a radar, divide the monitoring images into a plurality of sub-image regions after pre-processing the monitoring images, calculate pixel-level similarity of corresponding sub-regions of adjacent two images by using an NCC normalized cross-correlation function, and obtain pixel-level correlation; A motion feature calculation module is configured to extract motion vector distribution in the sub-region by using an optical flow method, quantify dynamic characteristics of vehicle interference, and obtain region-level motion features; A determination threshold adjustment module is configured to construct a multi-scale interference discrimination model based on a random forest algorithm, input the pixel-level correlation and the region-level motion features into the multi-scale interference discrimination model, and adjust an interference determination threshold; An amplitude deviation calculation module is configured to determine a usable sub-image region when the correlation is lower than the determination threshold, and calculate dynamic amplitude deviation of a potential scattering point; A monitoring result generation module is configured to remove unstable potential scattering points through the dynamic amplitude deviation, and generate a target slope deformation monitoring result.

9. A vehicle clutter suppression system for a light synthetic aperture traffic edge monitoring radar as claimed in claim 8, characterized in that The amplitude deviation calculation module comprises the following sub-modules: A compensation sub-module is configured to amplitude-compensate the potential scattering point according to the image overall amplitude, and estimate the potential scattering point through the amplitude deviation: for the sub-image region with higher availability, the image overall amplitude is calculated, and the image overall amplitude is the sum of amplitude values of all pixel points in the sub-image region; A determination sub-module is configured to determine the potential scattering point, and amplitude-compensate the potential scattering point according to the image overall amplitude, and the compensation formula is: Acompensated=Aoriginal+α×(Atotal-Aavg); Wherein, Acompensate is the compensated potential scattering point amplitude value, Aoriginal is the original potential scattering point amplitude value, a is the amplitude compensation coefficient, Atotal is the image overall amplitude of the sub-image region, Aavg is the average value of the amplitude values of all pixel points in the sub-image region.

10. The vehicle clutter suppression system of a light weight synthetic aperture roadside monitoring radar of claim 8, wherein, The monitoring result generation module comprises the following sub-modules: A calculation sub-module is configured to estimate the stability of the potential scattering point by calculating the dispersion of the amplitude value of the potential scattering point. A generation sub-module is configured to determine that the potential scattering point is stable when the amplitude dispersion is less than a set dispersion threshold value, and to generate a target slope deformation monitoring result by taking the potential scattering point as a reliable monitoring point, or to generate the target slope deformation monitoring result by eliminating the potential scattering point when the amplitude dispersion is greater than the dispersion threshold value.

Images