MIMO millimeter wave imaging radar point cloud image interference point elimination method

Abstract

The invention discloses an MIMO millimeter wave imaging radar point cloud image interference point elimination method, and belongs to the technical field of radar point cloud data processing, and the method comprises the steps: obtaining radar point cloud data, carrying out the preprocessing, converting the speed into the distance through the dynamic adjustment of clustering parameters and the addition of a new dimension, and improving the clear degree of a dynamic target point cloud image and a static target point cloud image. The method effectively overcomes the problems that a traditional noise point elimination method based on density noise space clustering cannot correctly eliminate noise points with far sparse and near dense characteristics, and abnormal points which are located near a target point cloud and have different speed characteristics cannot be eliminated due to the fact that speed information is ignored. The accuracy and reliability of the radar point cloud imaging system are improved, the interference point elimination requirement of the radar point cloud imaging system is met, and the point cloud image optimization effect in radar imaging is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of radar point cloud data processing, and in particular to a method for eliminating interference points of a MIMO millimeter wave imaging radar point cloud image. Background Art

[0002] MIMO millimeter wave imaging radar has broad application prospects in the fields of autonomous driving, drone navigation, security monitoring, etc., and it can provide high-resolution environmental perception capabilities. However, there are often a large number of interference points or noise in radar point cloud data. These interference points may be caused by environmental complexity, multipath effects, scatterers or other irrelevant targets, which seriously affect the accuracy of target detection and tracking. In addition, due to factors such as the non-ideality of manufacturing process, signal processing and algorithm operation, the actual MIMO radar imaging system will have some outliers or non-outliers but inconsistent interference points when outputting point cloud data.

[0003] The existing technology has the following deficiencies in removing noise from radar point cloud data: First, the K-means algorithm needs to know the number of targets in advance when processing radar point cloud data, and the classification effect of data sets based on distance distribution is poor. Secondly, although the DBSCAN algorithm does not require prior information on the number of targets, has fewer parameters and is simple to implement, due to the "close-up and far-away" characteristics of radar point clouds, the DBSCAN algorithm with fixed parameters does not work well when processing targets and noise at different distances. It is easy for small targets at a distance to be mistaken for noise, or for nearby noise to be mistakenly retained. These deficiencies make it difficult for existing methods to effectively eliminate interference points in complex radar point cloud data, affecting the accuracy and reliability of radar imaging systems. Summary of the invention

[0004] The purpose of the present invention is to provide a method for eliminating interference points of MIMO millimeter-wave imaging radar point cloud images, dynamically adjust clustering parameters through prior information such as radar system parameters and target size, and add a new dimension to convert speed into distance, so as to improve the clarity of moving and static target point cloud images, effectively overcome the problem that the traditional density-based noise point elimination method with noisy spatial clustering cannot correctly eliminate noise points with the characteristics of far sparse and near dense, and the problem that abnormal points located near the target point cloud but with different speed characteristics cannot be eliminated due to ignoring speed information, and can meet the needs of eliminating interference points of radar point cloud imaging system and realize the point cloud image optimization needs in radar imaging.

[0005] The purpose of the present invention is achieved by adopting the following technical solutions:

[0006] A method for eliminating interference points of a MIMO millimeter-wave imaging radar point cloud image, the method comprising the following steps:

[0007] Step 1: Data input and coordinate conversion. First, input the radar data and convert the data into coordinates in the new distance dimension to separate targets with close distance but large speed difference.

[0008] Step 2: Radar parameter input, input the radar system parameter prior information into the system;

[0009] Step 3: DBSCAN algorithm parameter configuration, dynamically adjust the algorithm's search radius and minimum number of points. Set more reasonable DBSCAN algorithm parameters based on prior information related to the radar system to meet the needs of eliminating interference points;

[0010] Step 4: Clustering implementation and clustering result output, remove the points of non-clustered clusters in the sample.

[0011] Furthermore, the step 1 comprises the following steps:

[0012] Step 1.1: Radar data input: Use MIMO millimeter wave imaging radar to collect point cloud data in the environment, and input the point cloud data collected by the MIMO millimeter wave imaging radar into the system.

[0013] Step 12: Add a time factor Tv to all targets based on the relationship between target distance and speed.

[0014] Step 1.3: Multiply the target's speed by the time factor Tv. Different speeds will present different distances in the new F dimension under the same time factor Tv, successfully separating targets that are close in distance but have a large speed difference.

[0015] Furthermore, the radar system parameter prior information includes distance resolution, angle resolution, velocity resolution, size of the detected target, and radar imaging distance.

[0016] Furthermore, the step 3 comprises the following steps:

[0017] Step 3.1: Radar point cloud density analysis. Calculate the target point cloud density in each local range segment area, analyze the distribution of local target point cloud density, and draw a density map or density heat map to visualize the density changes in the data set.

[0018] Step 3.2: Divide the radar detection range into multiple distances. According to the change law of target point cloud density with distance range and radar angular resolution, the distance intervals are reasonably divided. Following the rule of short to long, the radar's effective distance range is divided into non-equally spaced distance intervals. When the lateral size of the target within the detection distance range is smaller than the beam width, the subsequent distance range can be regarded as a distance segment and no further division is required.

[0019] Step 3.3: Dynamically set the search radius of the DBSCAN algorithm. For high-density areas, select a smaller search radius value to capture the details and small clusters in these areas. For low-density areas, select a larger search radius value to ensure that the points in these areas can form reasonable clusters instead of being marked as noise. After each clustering, adjust the search radius according to the clustering effect until the ideal clustering effect is achieved.

[0020] Step 3.4: Dynamically set the minimum point value of the DBSCAN algorithm. For each core point, dynamically adjust the minimum point value to match the point density of the area according to the different target point cloud densities in the distance area. After each clustering, evaluate the quality of the results and adjust the minimum point value according to the false detection rate or missed detection rate in the clustering results until the ideal clustering effect is achieved.

[0021] Furthermore, step 4 includes the following steps:

[0022] Step 4.1: Core point determination and temporary cluster formation. Starting from any sample point, record the number of sample points that fall within the circular range with it as the center and the radius as the search radius. When the number of points is greater than or equal to the minimum parameter, the set of sample points is determined to be a temporary cluster and the sample point at the center is the core point. All sample points are traversed in turn to obtain all core points and temporary clusters in the sample data; usually, the distance formula is used to determine whether the sample point falls within the range of the core point:

[0023]

[0024] Among them, dist represents the direct distance between the current sample point and the core point, which is judged by the search radius value. c ,y c ,z c are the three distance dimension coordinates of the core point, and x i ,y i ,z i are the three-dimensional distance dimension coordinates of the sample points respectively.

[0025] Step 4.2: Combine temporary clusters to obtain the final cluster. Starting from any core point, find the core point that can be reached by its density, and merge the temporary clusters to the same cluster. Traverse all core points in turn to complete the discrimination of target points and non-target points of the final sample data.

[0026] Step 4.3: Remove the points of non-clustered clusters in the sample and output the clustering results.

[0027] The advantages of the present invention are:

[0028] (1) Dynamic parameter optimization: The present invention dynamically adjusts the clustering parameters of the DBSCAN algorithm based on the prior information of the radar system, so that it can better adapt to the "close-to-dense and far-to-sparse" characteristics of radar point cloud data. This optimization method effectively avoids the problem of small targets at a long distance being misidentified as noise points or close-to-distance noise points being mistakenly retained, and improves the accuracy of noise elimination.

[0029] (2) Improving radar imaging quality: By effectively removing interference points, the present invention improves the point cloud data quality of the MIMO millimeter-wave imaging radar, improves the overall performance of the imaging system, and meets the requirements of high-resolution environmental perception.

[0030] (3) Easy implementation: Although the present invention introduces dynamic parameter adjustment and multi-dimensional analysis, the overall algorithm structure still maintains the simplicity of the DBSCAN algorithm and is easy to integrate and implement in existing radar imaging systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Schematic diagram of the interference point elimination method provided by the embodiment of the present invention

[0032] Figure 2 The spatial position change diagram of targets with different speeds in the new F dimension

[0033] Figure 3 Schematic diagram of sample point-distance relationship provided by an embodiment of the present invention

[0034] Figure 4 Schematic diagram of the change in the number of target point clouds of different sizes at different distances provided by an embodiment of the present invention DETAILED DESCRIPTION

[0035] The specific technical implementation scheme of the present invention is further described below in conjunction with the accompanying drawings to help those skilled in the art further understand the present invention without limiting their rights.

[0036] A method for eliminating interference points from MIMO millimeter-wave imaging radar point cloud images, see Figure 1 , the method comprises the following steps:

[0037] Step 1: Data input and coordinate conversion. First, input the radar data and convert the data into coordinates in the new distance dimension to separate targets with close distance but large speed difference.

[0038] Step 2: radar parameter input, inputting radar system parameter prior information into the system. The radar system parameter prior information includes range resolution, angle resolution, velocity resolution, size of the detected target, and radar imaging distance.

[0039] Step 3: DBSCAN algorithm parameter configuration: set more reasonable DBSCAN algorithm parameters according to the prior information related to the radar system to meet the requirements of interference point elimination.

[0040] Step 4: clustering implementation and clustering result output, removing the points of non-clustered clusters in the sample. Further, the step 1 includes the following steps:

[0041] Furthermore, the step 1 comprises the following steps:

[0042] Step 1.1: Radar data input. Use MIMO millimeter wave imaging radar to collect point cloud data in the environment. Point cloud data is a three-dimensional data set that contains multiple data points, each of which is composed of its spatial coordinates. These data points represent the surface features of the object reflected by the radar, including the points of the target object and the interference points. The point cloud data collected by the MIMO millimeter wave imaging radar is input into the system.

[0043] Step 1.2: Add a time factor Tv to all targets based on the relationship between target distance and speed.

[0044] Step 1.3: Multiply the target's velocity by the time factor Tv. Figure 2 , different speeds will present different distance positions in the new F dimension under the action of the same time factor Tv, successfully separating targets with close distances but large speed differences.

[0045] Furthermore, the step 2 comprises the following steps:

[0046] Step 2.1: Radar parameter input, including prior information such as range resolution, angle resolution, velocity resolution, size of detected targets, and radar imaging distance.

[0047] Furthermore, the step 3 comprises the following steps:

[0048] Step 3.1: Radar point cloud density analysis. Figure 3 , calculate the target point cloud density of each local distance segment area, analyze the distribution of local target point cloud density, and draw a density map or density heat map to visualize the density changes in the data set.

[0049] Step 3.2: Divide the radar detection range into multiple distances. Figure 4 According to the change law of target point cloud density with distance range and radar angular resolution, the distance interval is reasonably divided. Following the rule from short to long, the radar's effective distance range is divided into non-equally spaced distance intervals. When the lateral size of the target within the detection distance range is smaller than the beam width, the subsequent distance range can be regarded as a distance segment and no further division is required.

[0050] Step 3.3: Dynamically set the search radius of the DBSCAN algorithm. For high-density areas, select a smaller search radius value to capture the details and small clusters in these areas. For low-density areas, select a larger search radius value to ensure that the points in these areas can form reasonable clusters instead of being marked as noise. After each clustering, adjust the search radius according to the clustering effect until the ideal clustering effect is achieved.

[0051] Step 3.4: Dynamically set the minimum point value of the DBSCAN algorithm. For each core point, dynamically adjust the minimum point value to match the point density of the area according to the different target point cloud densities in the distance area. After each clustering, evaluate the quality of the results and adjust the minimum point value according to the false detection rate or missed detection rate in the clustering results until the ideal clustering effect is achieved.

[0052] Furthermore, step 4 includes the following steps:

[0053] Step 4.1: Core point determination and temporary cluster formation. Starting from any sample point, record the number of sample points that fall within the circular range with it as the center and the radius as the search radius. When the number of points is greater than or equal to the minimum parameter, the set of sample points is determined to be a temporary cluster and the sample point at the center is the core point. All sample points are traversed in turn to obtain all core points and temporary clusters in the sample data; usually, the distance formula is used to determine whether the sample point falls within the range of the core point:

[0054]

[0055] Where dist represents the direct distance between the current sample point and the core point, which is judged by the search radius value. c ,y c ,z c are the three distance dimension coordinates of the core point, x i ,y i ,z i are the three-dimensional distance dimension coordinates of the sample points respectively.

[0056] Step 4.2: Combine temporary clusters to obtain the final cluster. Starting from any core point, find the core point that can be reached by its density, and merge the temporary clusters to the same cluster. Traverse all core points in turn to complete the discrimination of target points and non-target points of the final sample data.

[0057] Step 4.3: Remove the points of non-clustered clusters in the sample and output the clustering results.

[0058] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for eliminating interference points of MIMO millimeter wave imaging radar point cloud image, characterized in that: The following steps are involved: Step 1: Data input and coordinate conversion. First, input radar data and convert the data into coordinates in the new distance dimension to separate targets with close distance but large speed difference. Step 2: Radar parameter input, input the radar system parameter prior information into the system; Step 3: Configure the DBSCAN algorithm parameters and dynamically adjust the algorithm’s search radius and minimum number of points; Step 4: Clustering implementation and clustering result output, remove the points of non-clustered clusters in the sample.

2. A method for eliminating interference points of MIMO millimeter wave imaging radar point cloud images according to claim 1, characterized in that: The step 1 comprises the following steps: Step 1.1: Radar data input: Use MIMO millimeter wave imaging radar to collect point cloud data in the environment; input the collected point cloud data into the system. Step 1.2: Add a time factor Tv to all targets based on the relationship between target distance and speed; Step 1.3: Multiply the target's velocity by the time factor Tv.

3. The method for eliminating interference points of a MIMO millimeter wave imaging radar point cloud image according to claim 1, characterized in that: The radar system parameter prior information includes distance resolution, angle resolution, velocity resolution, size of the detected target, and radar imaging distance.

4. The method for eliminating interference points of a MIMO millimeter wave imaging radar point cloud image according to claim 1, characterized in that: The step 3 comprises the following steps: Step 3.1: Radar point cloud density analysis: calculate the target point cloud density in each local range segment area, analyze the distribution of local target point cloud density, and draw a density map or density heat map; Step 3.2: Divide the radar detection range into multiple distances; Step 3.3: Dynamically set the search radius of the DBSCAN algorithm; Step 3.4: Dynamically set the minimum point value of the DBSCAN algorithm.

5. The method for eliminating interference points of a MIMO millimeter wave imaging radar point cloud image according to claim 1, characterized in that: The step 4 comprises the following steps: Step 4.1: Core point determination and temporary cluster formation. Starting from any sample point, record the number of sample points that fall within the circular range with it as the center and the radius as the search radius. When the number of points is greater than or equal to the minimum parameter, the sample point set is determined to be a temporary cluster and the sample point at the center is the core point. All sample points are traversed in turn to obtain all core points and temporary clusters in the sample data; determine whether the sample point falls within the range of the core point based on the distance formula: Where dist represents the direct distance between the current sample point and the core point, x c ,y c ,z c are the three distance dimension coordinates of the core point, x i ,y i ,z i They are the three-dimensional distance dimension coordinates of the sample points respectively; Step 4.2: Merge temporary clusters to get the final cluster. Starting from any core point, find the core point whose density can be reached, and merge the temporary clusters to the same cluster. Traverse all core points in turn to complete the discrimination of target points and non-target points of the final sample data. Step 4.3: Remove the points of non-clustered clusters in the sample and output the clustering results.

Images