A method for resolving and suppressing point trail clutter based on echo multi-features

Abstract

This invention discloses a clutter discrimination and suppression method based on multiple echo features, comprising: acquiring radar front-end clutter and target echo detection video data and dividing it into several connected regions; extracting multi-dimensional features and labeling prior information for each connected region; using the Relief feature selection algorithm to calculate and filter the weights of the multi-dimensional features, removing redundant features to obtain the optimal feature vector; using this feature vector to train a support vector machine classifier, and obtaining the optimal parameter model through cross-validation; acquiring measured clutter data, extracting corresponding features and inputting them into the model; and determining whether to remove clutter and retain targets based on the output. This invention effectively filters out dynamic clutter spots, avoids false deletion of weak targets, significantly reduces the false alarm rate of the system, and alleviates the computational burden of subsequent track processing.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing and target detection technology, and in particular to a method for resolving and suppressing point clutter based on multiple echo features. Background Technology

[0002] In the field of radar detection, clutter suppression is a key technology for improving the reliability of target detection and the quality of subsequent track processing. With the increasing complexity of radar operating environments, actual echo signals are often mixed with internal noise, ground clutter, sea clutter, and various electromagnetic interferences. Traditional methods mainly rely on clutter removal or Doppler filtering techniques based on specific statistical models, such as Moving Target Indication (MTI) and Moving Target Detection (MTD). Although these methods can effectively suppress static clutter through frequency domain filtering, in maritime detection scenarios, sea clutter itself has strong spatiotemporal correlation, and maritime targets are mostly slow-moving. This causes the weak Doppler characteristics of targets to often severely overlap with the Doppler range of slow-moving clutter, resulting in a sharp drop in the effectiveness of traditional methods. In recent years, with the leap in signal processing computing power, clutter suppression methods based on the fusion of single or multiple features such as echo amplitude, phase, and time-frequency characteristics have gradually emerged, achieving certain clutter filtering effects in uniform clutter environments.

[0003] However, existing clutter suppression techniques still have significant limitations when facing complex and ever-changing real-world sea and meteorological environments. First, current methods do not fully utilize echo characteristics, often limiting themselves to amplitude or single Doppler information. This makes them ineffective in dealing with non-uniform dynamic clutter caused by complex sea conditions, and they are prone to mistakenly deleting real targets with complex motion characteristics or weak signals as clutter. Second, traditional statistical models have weak scene adaptability and generalization capabilities, and limited ability to represent dynamically changing clutter. In unknown clutter environments, this often leads to a large amount of clutter residue (such as clutter spots that happen to pass through the Doppler filter), resulting in a high false alarm rate and significantly increasing the computational burden of subsequent track initiation. Furthermore, existing multi-feature high-dimensional fusion algorithms generally face high time complexity and computational bottlenecks, making it difficult to meet the high-speed, real-time processing requirements of radar systems for massive detection points. These shortcomings severely restrict the reliable resolution and stable detection capabilities of maritime radar in low signal-to-clutter and complex electromagnetic environments. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0005] In view of the aforementioned existing problems, this invention is proposed. Therefore, this invention provides a point clutter resolution and suppression method based on multiple echo features to solve the problems mentioned in the background art.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a method for resolving and suppressing point clutter based on multiple echo features, comprising: This section is divided into several connected regions; Multidimensional feature extraction is performed on each of the connected regions, and prior information is classified and labeled. The Relief feature selection algorithm is used to calculate and filter the weights of the extracted multidimensional features, remove redundant features, and obtain the optimal feature vector. The optimal feature vector is used to train the support vector machine classifier, and the optimal classifier parameters are obtained through cross-validation to obtain the trained support vector machine model. Acquire radar measured point data, extract features corresponding to the optimal feature vector and input them into the trained support vector machine model, determine the point attributes based on the output sign of the model, remove points judged as clutter and retain target points.

[0007] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By introducing the concept of connected component partitioning in image processing, multidimensional features of clutter and targets (such as amplitude standard deviation, azimuth / range broadening, etc.) are extracted. The Relief feature selection algorithm is then used for weight calculation and filtering to remove redundant features. This method not only solves the problems of high time complexity and massive computational load in traditional multi-feature high-dimensional fusion algorithms, but also fully explores the optimal robust features for distinguishing targets from clutter, effectively avoiding the misidentification and deletion of real targets with complex motion characteristics or weak signals as clutter in complex and non-uniform sea conditions.

[0008] 2. A Support Vector Machine (SVM) classifier is trained using the selected optimal feature vectors. Leveraging the significant advantages of SVM (especially the radial basis function kernel) in handling nonlinear and high-dimensional pattern recognition, this overcomes the shortcomings of traditional statistical models, such as weak scene adaptability and poor generalization ability. This model can effectively identify and remove clutter residue spots that could just pass through a traditional Doppler filter. It maintains extremely high resolution even in complex electromagnetic environments with low signal-to-noise ratios and low signal-to-clutter ratios. Verification by experimental data shows that its true target point retention rate and false clutter point suppression rate can both reach over 98%.

[0009] 3. Furthermore, the output of this invention is a target track with high confidence that has been effectively filtered. The significantly reduced number of residual tracks greatly reduces the false alarm rate of the radar system, thereby alleviating the data processing load of subsequent track initiation and multi-target automatic tracking algorithms from the source, and improving the efficiency of high-speed real-time processing of the radar system as well as the accuracy and stability of multi-target tracking. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart illustrating the overall process of a point clutter resolution and suppression method based on multiple echo features according to an embodiment of the present invention. Figure 2 This is a schematic diagram of clutter and target echo data for a clutter resolution and suppression method based on multiple echo features according to an embodiment of the present invention. Figure 3 This is a connected region numbering diagram of the point clutter resolution and suppression method based on multiple echo features according to an embodiment of the present invention. Figure 4 This is a schematic diagram of the amplitude standard deviation and distance broadening variable classifier results for the point clutter resolution and suppression method based on multiple echo features according to an embodiment of the present invention. Figure 5 This is a schematic diagram of the amplitude standard deviation and azimuth broadening of the variable classifier results for the point clutter resolution and suppression method based on multiple echo features according to an embodiment of the present invention. Figure 6 This is a schematic diagram of the azimuth broadening and range broadening variable classifier results for the point clutter resolution and suppression method based on multiple echo features described in an embodiment of the present invention. Figure 7 This is a schematic diagram of the multi-feature clutter discrimination and suppression method based on echo multi-features according to an embodiment of the present invention before multi-feature clutter filtering; Figure 8 This is a schematic diagram of the multi-feature trace filtering of the trace clutter resolution and suppression method based on echo multi-features according to an embodiment of the present invention. Detailed Implementation

[0011] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0012] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0013] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0014] This invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of this invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not adhering to the usual scale. Furthermore, the schematic diagrams are merely examples and should not be construed as limiting the scope of protection of this invention. In actual fabrication, the three-dimensional spatial dimensions of length, width, and depth should be included.

[0015] Furthermore, in the description of this invention, it should be noted that the terms "upper," "lower," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used solely for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. In addition, the terms "first," "second," or "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0016] Unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" in this invention should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; similarly, they can refer to mechanical connections, electrical connections, or direct connections, or indirect connections through an intermediate medium, or internal connections between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. Example 1

[0017] Reference Figure 1This is the first embodiment of the present invention, which provides a method for resolving and suppressing point clutter based on multiple echo features, including: S1. Acquire clutter and target echo data and delineate connected regions.

[0018] Furthermore, clutter and target echo detection video data samples are acquired from the radar front end. At the same time, the concept of connected component partitioning in image processing is introduced to divide spatially adjacent echo points in the radar detection video data into the same connected region.

[0019] Specifically, based on the principle of connected component partitioning in image processing, the entire detection video data is traversed, and each defined connected region is assigned an incrementally unique number to facilitate feature extraction for each independent block.

[0020] S2. Extract multidimensional features and label prior information.

[0021] Furthermore, multi-dimensional features are extracted from each numbered connected region block, and prior information (target or clutter) is classified and labeled based on theoretical and practical considerations. In this embodiment, the eight extracted features and the calculation methods for each feature are as follows: (1) Number of EPs (number of connected points): Defined as the total number of all numbered echo points contained in the corresponding connected region. The number of EPs is obtained by counting the points in the region.

[0022] (2) Azimuth widening: defined as the maximum azimuth difference among all echo points in the connected region. Assume the maximum value of the azimuth coordinates of all echo points in the connected region is... The minimum value is Then the directional expansion The calculation formula is: (3) Distance widening: Defined as the maximum distance difference between all echo points in the connected region. Assume the maximum distance difference between all echo points in the connected region on the distance coordinate is... The minimum value is Then the distance widens The calculation formula is: (4) Inclination angle: Defined as the angle between the major axis of the connected region and the x-axis of the coordinate system. Widening based on orientation. and distance widening Inclination angle The calculation formula is: (5) Average distance: Defined as the average distance between all points within a connected region and the centroid of that region. Assume there are a total of [number missing] points within the connected region. The point, the first The coordinates of the points are The amplitude value at that point is First, calculate the centroid coordinates of the connected region. : Next, calculate the first [unit] within the connected region. Euclidean distance from each point to the centroid : Then the average distance The calculation formula is: (6) Amplitude Standard Deviation: Defined as the standard deviation of the amplitude values ​​of all detection points within a connected region relative to the mean amplitude of all points in that region. Let the mean amplitude of the points within the region be... Then the standard deviation of the amplitude The calculation formula is: (7) Azimuth-distance widening ratio: defined as the ratio R of the azimuth widening to the distance widening of the connected region. (8) Standard deviation of boundary distance: defined as the ratio of the standard deviation of the distance between the boundary point and the centroid of the connected region to the maximum distance from the boundary point to the centroid. First, the boundary point set is obtained by contour extraction of the connected region. ,in This represents the total number of boundary points. Next, calculate the value for each boundary point. With center of mass Euclidean distance And calculate the mean distance from all boundary points to the centroid. : Then, the standard deviation of the distance from the boundary point to the centroid is calculated. : Finally, let the maximum distance from the boundary point to the centroid be... Then the standard deviation of the boundary distance for: S3, Multiple Feature Selection and Redundancy Removal.

[0023] It should be noted that, in order to save computing resources and improve classifier performance, the present invention uses the classic Relief feature selection algorithm to calculate and filter the weights of the eight features extracted by S2.

[0024] It's important to explain that the classic Relief algorithm is a commonly used feature selection method, primarily designed for binary classification problems (such as target / clutter). It assigns different weights to each feature based on its relevance to the class; the higher the weight, the more beneficial the feature is for classification. Effective features are selected based on their weights. A core rule of the Relief algorithm for updating the weights of each feature is: if the distance between a sample and its nearest neighbor of the same class on a certain feature is less than the distance between the sample and its nearest neighbor of a different class on that feature, then that feature is helpful in distinguishing between the same and different classes, and its weight should be increased; conversely, its weight should be decreased. This process is repeated multiple times, ultimately yielding the average weight of each feature. A higher weight indicates a stronger classification ability for that feature.

[0025] Specifically, based on the above rules, let the sample set be... For each sample Find its nearest neighbor samples of the same type and neighboring samples of different types For the first The formula for calculating the weight update of each feature is as follows: in, Indicates the first The weights of each feature; Indicates sample In the The values ​​that can be taken on each feature; and These represent the nearest neighbors of the same class and the nearest neighbors of different classes in the first position. The function takes values ​​on each feature. Used to calculate distance difference, for continuous feature attributes, it is defined as the absolute value of the difference between two values, i.e.: Furthermore, if the distance between a sample and its nearest neighbor of the same class on a given feature is less than the distance between it and its nearest neighbor of a different class, the weight of that feature increases; conversely, it decreases. After multiple iterations with all samples, the average weight of each feature is calculated. The features are then sorted from largest to smallest according to their average weight, and the three features with the largest average weights (amplitude standard deviation, azimuth broadening, and distance broadening) are selected to form the optimal feature vector.

[0026] S4. Classifier training.

[0027] It should be noted that commonly used classifiers include traditional neural network classifiers and support vector machines. Since support vector machines are classifiers based on minimizing structural risk, they have the largest margin in the feature space and have outstanding advantages in solving pattern class recognition problems with small sample sizes, nonlinearity, and high dimensionality. In this embodiment, support vector machines (SVM) are used for classifier training.

[0028] It's important to explain that Support Vector Machine (SVM) is a classic supervised learning algorithm primarily used for classification tasks. Its core idea is to find an optimal hyperplane that maximizes the margin between positive and negative samples, thereby obtaining a classifier with strong generalization ability and high robustness. Geometrically, the goal of SVM is to find an optimal hyperplane that separates samples of different classes. This optimality criterion is not simply separating all samples, but rather requiring the hyperplane to be as far away from the nearest sample points as possible; this distance is called the margin. These sample points closest to the hyperplane are called support vectors, and they directly determine the hyperplane's position and orientation, while other sample points far from the boundary have no impact on the model. Therefore, in practice, SVM achieves the maximum margin idea through a convex quadratic programming problem.

[0029] Furthermore, based on the above, the optimization objective of using SVM for classifier training is to minimize the norm of the weight vector and maximize the classification margin. Additionally, for linearly inseparable or noisy data, we introduce slack variables. and penalty coefficient Processing nonlinear and noisy data. The convex quadratic optimization problem is equivalent to: in, Let be the normal vector of the hyperplane. For bias terms, Controlling the cost of misclassification As slack variables, For sample labels, This is a mapping function.

[0030] It should be noted that transforming the primal constrained optimization problem into a dual problem using the Lagrange multiplier method is key to solving SVM. A significant advantage of the dual form is that both the objective and decision functions involve only the inner product between samples. To handle nonlinear classification problems, SVM maps the original features to a high-dimensional space using kernel functions, making the data linearly separable in that space. Kernel functions can directly compute the inner product in the mapped space, thus avoiding complex explicit computations in the high-dimensional space. Commonly used kernel functions include linear kernels, polynomial kernels, and Gaussian radial basis function (RBF) kernels.

[0031] Furthermore, in this embodiment, we preferably use the radial basis function (RBF kernel) to map the features to a high-dimensional space, and its formula is as follows: in, For feature vectors, For regularization hyperparameters (equivalent to) ), which controls the width and scope of the kernel function.

[0032] Furthermore, the data samples are divided into training and validation sets.

[0033] Specifically, on the validation set, with classification accuracy as the optimization objective (accuracy = number of correctly classified samples / total number of validation samples), cross-validation is used to adjust the penalty coefficient. and kernel parameters By performing combination optimization, the optimal classifier parameters are obtained.

[0034] S5. Classification of measured data and clutter suppression.

[0035] Furthermore, radar measured point data is acquired, and the same connected component partitioning as in S1 is performed. The optimal 3D feature vector selected in S3 is then extracted. The feature data is input into the trained SVM model, and the model's decision function is: in, The number of support vectors; For the first Lagrange multipliers corresponding to each support vector; The class label (+1 or -1) for the support vectors; The output value of the kernel function; For bias terms; It is a symbolic function.

[0036] Furthermore, the attribute of the dot is determined based on the output symbol: if (With a positive sign), determine that the points in the current connected region are the target points and save them; if (With a negative sign) it is identified as clutter and removed. Example 2

[0037] Reference Figures 2 to 8 This is the second embodiment of the present invention, which is the experimental verification part of the present invention. This experimental verification part uses measured data from the Army's border and coastal defense X-band radar in the Yantai sea area to simulate and verify the above-mentioned point clutter resolution and suppression method based on multiple echo features. The specific process is as follows: First, the detection video data is acquired from the signal processing output of this type of radar, and the data containing clutter and targets is extracted. Its original distribution pattern is as follows: Figure 2 As shown, the video data is divided into several connected regions using the principle of connected component partitioning and then numbered incrementally. For example... Figure 3 The image shows a visualization of the local data connectivity number (different gray levels / colors represent unique numbers for different connected regions, effectively distinguishing spatially adjacent echo point clusters).

[0038] Secondly, according to the formula in Example 1, for Figure 3 For each connected region divided in the middle, eight characteristic values ​​are calculated: number of EPs, azimuth expansion, distance expansion, tilt angle, average distance, amplitude standard deviation, azimuth-distance expansion ratio, and boundary distance standard deviation.

[0039] Specifically, the eight extracted feature values ​​were processed using the classic Relief algorithm, and the average weight distribution of each feature was calculated after multiple iterations, as shown in Table 1.

[0040] Table 1. Distribution of Feature Weights As shown in Table 1, the amplitude standard deviation has the largest weight (0.344616), followed by azimuth broadening (0.327069) and range broadening (0.301078). Other features, such as the azimuth-range broadening ratio, have extremely small weights due to their low ability to distinguish between targets and clutter. Therefore, redundant features are removed, and the three features with the largest average weights are selected to form the optimal feature vector, which is then input into the classifier.

[0041] In addition, radar data was collected three times in this experiment. After manual cleaning and target / clutter prior information labeling, a total of 3000 valid data samples were formed. These 3000 samples were divided into two datasets: 1500 for model training and 1500 for model validation. Based on the determined optimal feature vectors, to intuitively demonstrate the separability of features to the data, amplitude standard deviation, range broadening, and azimuth broadening were combined pairwise. The classification boundary results under the SVM model are as follows: Figure 4 (Amplitude standard deviation and distance broadening) Figure 5 (Amplitude standard deviation and azimuth broadening) Figure 6 (As shown in the diagram, azimuth broadening and distance broadening are also shown.) The figure clearly illustrates the nonlinear classification hyperplane defined by the RBF kernel function in high-dimensional space. The trained SVM model was tested using a validation dataset, and through cross-validation optimization, the optimal parameter combination (penalty coefficient) that maximized classification accuracy was finally determined. With kernel parameters ).

[0042] Finally, the optimal feature vector selection rules and the optimal SVM parameter model were directly deployed to another batch of unlabeled, measured complex sea surface radar point data for clutter filtering. Verification results comparison: Figure 7 The radar P-display image before multi-feature point filtering is shown. The system maintains [its functionality] even in environments with strong sea clutter. The constant false alarm rate (CFAR) was used for detection, initially detecting over 1 million points. After routine preliminary processing, approximately 100,000 points remained. At this point, the image was still filled with a massive amount of clutter, except for a small number of real targets. When the multi-dimensional feature point filtering processing of this invention was enabled, the radar P-display image appeared as follows: Figure 8 As shown, the SVM model accurately stripped all points based on the output sign of the decision function. The vast majority of clutter and noise points were successfully identified and removed, reducing the remaining points to approximately 2000, with only high-confidence target points retained.

[0043] In summary, after filtering using the method of this invention, the retention rate of real target points (real points after filtering / real points before filtering) reaches 98%, while the suppression rate of false clutter points (false points removed / false points before filtering) is ≥98%. This experiment fully demonstrates that this invention effectively ensures high-probability target detection while significantly reducing the false alarm rate of the system and alleviating the computational load on subsequent track initiation and tracking algorithms.

[0044] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0045] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1A device that provides the functions specified in one or more boxes.

[0046] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0047] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0048] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0049] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for resolving and suppressing point clutter based on multiple echo features, characterized in that, include: Acquire clutter and target echo detection video data samples from the radar front end, and divide the detection video data samples into several connected regions; Multidimensional feature extraction is performed on each of the connected regions, and prior information is classified and labeled. The Relief feature selection algorithm is used to calculate and filter the weights of the extracted multidimensional features, remove redundant features, and obtain the optimal feature vector. The optimal feature vector is used to train the support vector machine classifier, and the optimal classifier parameters are obtained through cross-validation to obtain the trained support vector machine model. Acquire radar measured point data, extract features corresponding to the optimal feature vector and input them into the trained support vector machine model, determine the point attributes based on the output sign of the model, remove points judged as clutter and retain target points.

2. The method for resolving and suppressing point clutter based on multiple echo features as described in claim 1, characterized in that, The detected video data samples are divided into several connected regions, including: Based on the principle of connected component partitioning in image processing, echo points that are spatially adjacent in radar detection video data are divided into the same connected region, and all connected regions obtained by traversal are assigned an incrementally unique number.

3. The method for resolving and suppressing point clutter based on multiple echo features as described in claim 1, characterized in that, The extracted multidimensional features include: The number of points in the connected region, azimuth expansion, distance expansion, tilt angle, amplitude standard deviation, and azimuth-distance expansion ratio.

4. The method for resolving and suppressing point clutter based on multiple echo features as described in claim 3, characterized in that, The extracted multidimensional features also include the average distance, and the extraction process of the average distance includes: Calculate the centroid coordinates of the corresponding connected region. The centroid coordinates are the sum of the products of the coordinates of all points in the region and their corresponding amplitude values, and the ratio of the sum of the amplitude values ​​of all points. Calculate the Euclidean distance from each point within the connected region to the centroid coordinates; Calculate the average Euclidean distance from all points to the centroid coordinates, and use this average distance as the mean distance.

5. The method for resolving and suppressing point clutter based on multiple echo features as described in claim 4, characterized in that, The extracted multidimensional features also include the boundary distance standard deviation, and the extraction process of the boundary distance standard deviation includes: Contour extraction is performed on the connected regions to obtain the set of boundary points; Calculate the distance between each boundary point and the centroid coordinates, and obtain the mean value of the distances from all boundary points to the centroid; Calculate the standard deviation of the distance from all boundary points to the centroid relative to this mean; The standard deviation of the boundary distance is calculated by dividing the standard deviation by the maximum distance from the boundary point to the centroid.

6. The method for resolving and suppressing point clutter based on multiple echo features as described in claim 1, characterized in that, The weights of the extracted multidimensional features are calculated using the Relief feature selection algorithm, including: In the sample set, for each sample, find its nearest neighbors of the same class and its nearest neighbors of different classes; For any feature, calculate the distance between a sample and its nearest neighbors of the same class on that feature, and the distance between a sample and its nearest neighbors of different classes on that feature; If the distance between a sample and its nearest neighbor samples of the same class on this feature is less than the distance between it and its nearest neighbor samples of different classes, then the weight update amount of this feature is increased; otherwise, the weight update amount of this feature is decreased. After multiple iterations, the average weight of each feature is calculated.

7. The method for resolving and suppressing point clutter based on multiple echo features as described in claim 6, characterized in that, After the weight calculation is completed, all extracted features are sorted from largest to smallest according to their average weight, and the three features with the largest average weight are selected to form the optimal feature vector.

8. The method for resolving and suppressing point clutter based on multiple echo features as described in claim 1, characterized in that, The training process of a support vector machine classifier includes: The data samples are divided into training and validation sets; The radial basis function kernel is used as the kernel function of the support vector machine to map the original features to a high-dimensional space; On the validation set, with classification accuracy as the optimization objective, cross-validation is used to combine and optimize the penalty coefficient and kernel function parameters. The set of parameters that maximizes the ratio of the number of correctly classified samples to the total number of validation samples is selected as the optimal classifier parameters.

9. The method for resolving and suppressing point clutter based on multiple echo features as described in claim 8, characterized in that, The radial basis function kernel is a similarity metric function based on Euclidean distance. Its output value depends on the radial distance between the feature vectors of two input samples, and the width of the kernel function and the complexity of the model are controlled by a regularization hyperparameter.

10. The method for resolving and suppressing point clutter based on multiple echo features as described in claim 1, characterized in that, Acquire radar-measured point data, extract features corresponding to the optimal feature vector and input them into the trained support vector machine model, determine the point attributes based on the model's output sign, remove points identified as clutter and retain target points, including: Perform the same division on the radar measured point data as on the connected region; Extract feature data with the exact same dimensions as the optimal feature vector obtained after the filtering process; The extracted feature data is input into the trained support vector machine model to obtain the output sign of the decision function; If the model output sign is positive, the points in the current connected region are determined to be the target points; if the model output sign is negative, the points in the current connected region are determined to be clutter points and are removed.

Images