CN-121995333-A - Radar clutter region division method based on multi-stage optimization

CN121995333ACN 121995333 ACN121995333 ACN 121995333ACN-121995333-A

Abstract

The invention discloses a radar clutter region dividing method based on multistage optimization, which is used for acquiring an initial radar clutter classification map of a pixel level, abstracting a macroscopic clutter region in the initial radar clutter classification map into nodes, abstracting an overlapping relation between the macroscopic clutter regions into edges, constructing a scene map in a map structure form, determining an optimal label of the nodes by taking a global energy function as an optimal target, and obtaining a final radar clutter classification map by taking the optimal label as a final label of a corresponding macroscopic clutter region.

Inventors

YANG YANBO
YANG HAOZHEN
WANG DEWU
Han Piqiang
TIAN BO
HE RUIXI

Assignees

西北工业大学
北京无线电测量研究所

Dates

Publication Date: 20260508
Application Date: 20251231

Claims (10)

1. The radar clutter area dividing method based on the multi-stage optimization is characterized by comprising the following steps of: Acquiring an initial radar clutter classification diagram of a pixel level; Abstracting a macro clutter region in the initial radar clutter classification map into nodes, abstracting an overlapping relation between the macro clutter regions into edges, and constructing a scene map in a map structure form, wherein feature vectors of the nodes comprise geometric features, physical statistical features and texture features of the macro clutter regions; constructing a global energy function based on nodes and clutter labels corresponding to the nodes, and determining an optimal label of the nodes by taking the minimum global energy function as an optimization target; And taking the optimal label as a final label of a corresponding macro clutter region to obtain a final radar clutter classification map.
2. The multi-stage optimization-based radar clutter region division method of claim 1, wherein the global energy function is: Where E (L) represents a global energy function based on a label distribution scheme L, V represents a node set, L v represents a label of the node V, U (V, L v ) represents a cost of assigning the node V to the label L v , λ represents a weight coefficient, edges represents a set of edges in the scene graph, (U, V) represents an edge between the node U and the node V, and P (L u ,l v ) represents a cost of assigning neighboring nodes U and the node V to the label L u and the label L v , respectively.
3. The method for dividing radar clutter regions based on multi-stage optimization according to claim 2, wherein the calculation method of U (v, l v ) is as follows: Training a classifier by using the feature vectors and the initial labels of all nodes in the scene graph; And acquiring the posterior probability that the node v belongs to each label by using a trained classifier, and taking the negative log likelihood of the posterior probability as U (v, l v ).
4. The method for radar clutter area division based on multi-stage optimization according to claim 3, wherein the calculation method of P (l u ,l v ) is as follows: When tag l u and tag l v are the same, P (l u ,l v ) is 0; When tag l u and tag l v are different, P (l u ,l v ) is a first threshold; When the label l u and the label l v are the preset combinations, P (l u ,l v ) is the second threshold and the second threshold is smaller than the first threshold.
5. The multi-stage optimization-based radar clutter region division method according to claim 3 or 4, wherein the geometric features include area, perimeter, solidity and eccentricity of the macro clutter region; The physical statistical characteristics comprise the mean value, standard deviation and variance of a key physical channel in a macroscopic clutter region, wherein the key physical channel comprises an amplitude channel, a phase channel, a power channel and an SNR channel; the texture features include contrast, dissimilarity, homogeneity and energy of the gray co-occurrence matrix of the amplitude channel in the macro clutter region.
6. The method for radar clutter region division based on multi-stage optimization according to claim 5, wherein when the first condition is satisfied, an overlapping relationship between macro clutter regions is abstracted to edges; The first condition is: wherein C i represents the ith macro clutter region, C j represents the jth macro clutter region, Represents morphological dilation operation, B adj represents structural elements when morphological dilation operation is performed on C i , and phi represents empty set.
7. The multi-stage optimization-based radar clutter region division method according to claim 6, further comprising, before constructing the scene graph in the form of a graph structure: Noise filtering based on connected domain analysis is carried out on the initial radar clutter classification map; and executing morphological closing operation on the macroscopic region in the initial radar clutter classification diagram after noise filtering to obtain the macroscopic clutter region.
8. The method for dividing radar clutter areas based on multi-stage optimization according to claim 6 or 7, wherein the step of using the optimal tag as a final tag of the corresponding macro clutter area further comprises: Constructing a local region of interest based on boundaries to be refined in the radar clutter classification map; constructing an edge stopping function according to the edge information of the amplitude channel of the original radar feature map to optimize the local region of interest; and iterating evolution boundaries in the optimized local region of interest based on the geodesic active contour model.
9. The multi-stage optimization-based radar clutter region division method of claim 2, wherein the optimal label for the node is determined based on U (v, l v ) when the initial radar clutter classification map contains only a single clutter label or the set of edges in the scene map is empty.
10. A method of radar clutter area division based on multistage optimization, comprising a memory, a processor and a computer program stored in the memory and running on the processor, characterized in that the processor implements the method according to any of claims 1-9 when executing the computer program.

Description

Radar clutter region division method based on multi-stage optimization Technical Field The invention belongs to the technical field of radar signal/data processing, and particularly relates to a radar clutter region dividing method based on multistage optimization. Background Radar is used as a core of a modern detection system, and is increasingly widely applied to the fields of aerospace, navigation, meteorological monitoring and the like. The performance of radar systems depends largely on the accurate interpretation of complex environmental echoes. The radar clutter serves as a non-target echo signal and is a key interference factor affecting target detection and environment perception. Therefore, the data received by the radar is processed to generate a clutter classification map which can accurately reflect the spatial distribution of different clutter types, and the method has a vital role in subsequent tasks such as clutter suppression, target tracking, scene understanding and the like. With the development of deep learning, a method represented by a convolutional neural network or a graph neural network has been successfully applied to generate a radar clutter classification map at the pixel level. The method can automatically learn complex features from the original radar echo data, and can accurately classify various categories such as ground clutter, sea clutter and the like. However, the conventional pixel-level clutter classification method has fundamental limitations that firstly, due to lack of macroscopic structure cognition, a large number of noise points and broken areas exist in an output image, a coherent representation of large-scale environments such as a ground clutter area, a sea clutter area and the like cannot be formed, secondly, the real topological relation of serious distortion of regional breaking is difficult to accurately maintain global consistency of key geographic structures such as a shore-sea boundary, inland, a water area and the like, thirdly, the output only stays at a low-level clutter type discrimination level, and composite high-level semantic information with higher decision value such as a shore-sea boundary transition zone, a clean detectable area and the like cannot be refined. Therefore, how to start from the pixel classification results of noise, fragmentation and topology distortion, realize macroscopic structure perception, global topology recovery and reconstruction of advanced composite semantics, thereby generating a radar environment partition map with complete structure, accurate topology and rich semantics, and still being a core technical problem to be solved in the current radar environment perception field. Disclosure of Invention The invention aims to provide a radar clutter region dividing method based on multi-stage optimization, which is used for optimizing an initial radar clutter classification map and improving the classification precision of the radar clutter classification map. The technical scheme adopted by the invention is that the radar clutter region dividing method based on multi-stage optimization comprises the following steps: Acquiring an initial radar clutter classification diagram of a pixel level; Abstracting a macro clutter region in an initial radar clutter classification map into nodes, abstracting an overlapping relation between the macro clutter regions into edges, and constructing a scene map in a map structure form, wherein feature vectors of the nodes comprise geometric features, physical statistical features and texture features of the macro clutter regions; Constructing a global energy function based on the node and the clutter label corresponding to the node, and determining an optimal label of the node by taking the minimum global energy function as an optimal target; And taking the optimal label as a final label of the corresponding macro clutter region to obtain a final radar clutter classification map. The method has the advantages that the method can reconstruct the original classification result graph full of noise into the radar environment partition graph which has correct macroscopic structure, accurate key boundary and contains advanced semantic information such as a coastal sea boundary region and the like by carrying out the multi-stage processing of graph structuring and global topological optimization on the original radar clutter classification graph at the pixel level, thereby greatly improving the accuracy, the robustness and the interpretability of radar environment perception. Drawings FIG. 1 is a flow chart showing the overall steps of the method of the present invention; FIG. 2 is a schematic diagram showing the comparison of the processing effects of the initial preprocessing module according to the embodiment of the present invention; FIG. 3 is a schematic diagram of a scene graph construction module according to an embodiment of the present invention; FIG. 4 is a schematic diagram of the pro