CN-121978635-A - Radar signal sorting method and system based on rough set theory and adaptive weighting

CN121978635ACN 121978635 ACN121978635 ACN 121978635ACN-121978635-A

Abstract

The invention discloses a radar signal sorting method and system based on a rough set theory and self-adaptive weighting, and relates to the technical field of baseband communication. The radar signal sorting method and system based on the rough set theory and the self-adaptive weighting comprise the following steps of S1, carrying out overall process dynamic collection on an aliasing full-pulse signal to obtain a pulse description word sequence, S2, dividing the full-pulse data stream to obtain granularity combination entropy, constructing a global weighted distance matrix, S3, carrying out local density estimation, constructing a decision diagram, screening pulse points, S4, constructing basic probability distribution and uncertainty function values, and outputting a signal sorting result. The method effectively improves the stability, accuracy and interpretability of radar signal sorting under the conditions of rapid parameter change and strong noise, and solves the problem that the existing unsupervised sorting cannot self-adaptively describe the pulse parameter stability difference under the conditions of fixed distance measurement and fixed weight.

Inventors

CHEN LINSHU
Ye Xueqi
LI LIANG
LIU YUANHUI
DIAO ZULONG
XIONG NAIXUE

Assignees

湖南科技大学

Dates

Publication Date: 20260505
Application Date: 20260409

Claims (10)

1. The radar signal sorting method based on the rough set theory and the self-adaptive weighting is characterized by comprising the following steps: s1, carrying out overall process dynamic acquisition on an aliasing full pulse signal to obtain a pulse description word sequence, and carrying out normalization, abnormal marking and isolation on the pulse description word sequence to obtain a full pulse data stream; s2, dividing the full-pulse data stream based on the pulse description word sequence to obtain granularity combination entropy, and carrying out equivalence class division to construct a global weighted distance matrix; s3, based on the global weighted distance matrix, carrying out local density estimation, constructing a decision graph, screening pulse points and obtaining a shared nearest neighbor number; S4, constructing basic probability distribution and uncertainty function values, generating global confidence coefficient, marking noise pulses and outputting signal sorting results.
2. The method for sorting radar signals based on rough set theory and adaptive weighting according to claim 1, wherein the method is characterized in that the method comprises the following steps of dynamically collecting the aliasing full pulse signals in a whole process to obtain a pulse descriptor sequence, normalizing the pulse descriptor sequence, and marking and isolating abnormality, and the specific process of obtaining the full pulse data stream is as follows: Carrying out overall process dynamic acquisition on an aliased full-pulse signal intercepted by a radar reconnaissance receiver to obtain a pulse description word sequence, wherein the pulse description word sequence comprises carrier frequency, pulse width, arrival angle and amplitude parameters, and recording an arrival time stamp for each pulse; The pulse description word sequence is written into the full pulse data stream through structured encapsulation, the field caliber and the recording granularity are unified, the full pulse data stream is subjected to dimension alignment by adopting a maximum and minimum value normalization method, and missing pulses, abnormal pulses and outlier pulses in the full pulse data stream are marked and isolated, so that errors of a dirty data amplification sorting link are avoided.
3. The method for sorting radar signals based on rough set theory and adaptive weighting according to claim 1, wherein the specific process of dividing the full pulse data stream based on the pulse descriptor sequence to obtain the granularity combination entropy is as follows: Based on the pulse description word sequence, calculating the unrecognizable relation of the pulses on the pulse description word sequence by adopting a threshold approximate discrimination method in the full pulse data stream, wherein the values on carrier frequency, pulse width, arrival angle and amplitude parameter dimension are respectively taken for any two pulses; if the absolute value of the difference between the two pulses in each parameter dimension does not exceed the discrimination threshold corresponding to the dimension, judging that the two pulses are unrecognizable in the pulse description word sequence and marking as meeting the unrecognizable relation; dividing the full pulse data stream based on the unrecognizable relation to obtain a pulse equivalence class division result, wherein each equivalence class represents a group of pulse samples which are mutually unrecognizable in carrier frequency, pulse width, arrival angle and amplitude parameter dimension; Counting the number of pulses in a full pulse data stream, carrying out non-overlapping full coverage division on the full pulse data stream through an unrecognizable relation, counting the number of pulses of each equivalence class to obtain the number of pulses of each equivalence class, counting the number of the equivalence classes obtained through the division to obtain the total number of radar pulse equivalence classes, numbering the equivalence classes one by one in sequence, squaring the number of pulses contained in each equivalence class, sequentially adding and summing to obtain the total number of equivalents, dividing the square of the number of pulses in the full pulse data stream to obtain a granularity item, dividing the number of pulses contained in each equivalence class by the number of pulses in the full pulse data stream to obtain the equivalence class occupation ratio, multiplying the equivalence class occupation ratio by the equivalence class occupation ratio, summing the equivalence class occupation ratio, obtaining an entropy item by subtracting the entropy item, multiplying the entropy item by the granularity item, and obtaining the granularity item combination.
4. The radar signal sorting method based on rough set theory and adaptive weighting according to claim 1, wherein the specific process of performing equivalence class division and constructing a global weighted distance matrix is as follows: The method comprises the steps of constructing a pulse descriptor sequence with parameters removed according to any parameter in the pulse descriptor sequence, recalculating unrecognizable relations and dividing equivalence classes based on the pulse descriptor sequence with parameters removed to obtain granularity combination entropy with parameters removed; The method comprises the steps of mapping a parameter distinguishing contribution degree into sorting weights through a segmentation mapping method, constructing weighted Euclidean distance to measure the difference between two radar pulses based on the sorting weights of all parameters, extracting the values of carrier frequency, pulse width, arrival angle and amplitude parameter dimensions of any two radar pulses, calculating the difference in each dimension and carrying out weighted summarization according to the sorting weights of the dimensions to obtain weighted distances between the two pulses, and carrying out weighted distance calculation on all pulses in a full pulse data stream to form a global weighted distance matrix.
5. The method for sorting radar signals based on rough set theory and adaptive weighting according to claim 1, wherein the specific process for estimating local density based on the global weighted distance matrix is as follows: Based on the global weighted distance matrix, k neighbor pulses with the nearest weighted distance of each radar pulse are used as nearest neighbor sets, variable bandwidth Keuchy kernel is adopted for local density estimation, weighted Euclidean distance between the current pulse and the neighbor pulses is used as a distance item, distance from the neighbor pulse to the nearest neighbor sets is used as self-adaptive bandwidth, and Keuchy kernel density values of the current pulse are obtained by accumulating Keuchy kernel responses of all the neighbor pulses in the nearest neighbor sets.
6. The radar signal sorting method based on rough set theory and adaptive weighting according to claim 1, wherein the specific process of constructing a decision diagram and screening pulse points is as follows: Dividing each neighbor pulse in the nearest neighbor set by the self-adaptive bandwidth to obtain a scaled distance, squaring the scaled distance, adding one to obtain the contribution value of the neighbor pulse, sequentially accumulating the contribution values of all neighbor pulses in the nearest neighbor set to obtain the accumulated value of the Cauchy kernel density of the current pulse, finding out the maximum value of the Cauchy kernel density value of each neighbor pulse corresponding to the variable bandwidth in the same nearest neighbor set as a normalization reference, dividing the accumulated value of the Cauchy kernel density of the current pulse by the normalization reference to obtain the relative Cauchy kernel density value; And constructing a decision diagram according to the relative Cauchy kernel density value and the pulse sample, and judging a pulse point with the relative Cauchy kernel density value not smaller than a density threshold as a radiation source center.
7. The method for sorting radar signals based on rough set theory and adaptive weighting according to claim 1, wherein the specific process of obtaining the shared nearest neighbor number is as follows: The method comprises the steps of respectively obtaining nearest neighbor sets of two radar pulses, calculating the number of pulses contained in an intersection set of the two nearest neighbor sets, defining the number of pulses as a shared nearest neighbor number of the two pulses, carrying out core pulse judgment and association based on the shared nearest neighbor number, calculating the shared nearest neighbor number of the pulses and a radiation source center for any pulse and the radiation source center, judging the pulses as core pulses of the radiation source when the shared nearest neighbor number is larger than a shared number threshold value, and directly associating the core pulses to the corresponding radiation source center.
8. The method for sorting radar signals based on rough set theory and adaptive weighting according to claim 1, wherein the specific process of constructing basic probability distribution and uncertainty function values and generating global confidence is as follows: The method comprises the steps of taking the pulses which fail to judge the core pulse and associate the core pulse as overlapping pulses, searching neighbor pulses in a nearest neighbor set, constructing basic probability distribution and uncertainty function values of each neighbor pulse to each radiation source center by utilizing evidence theory, taking weighted Euclidean distance between the overlapping pulses and the neighbor pulses as distance attenuation items, and normalizing the number of samples falling into the radiation source center in the nearest neighbor set of the neighbor pulses according to the nearest neighbor number as neighborhood support items; The method comprises the steps of integrating all basic probability distribution by utilizing an evidence theory combination rule, adding basic probability distribution provided by each neighbor pulse of overlapped pulses and an uncertainty function value to obtain a support synthesis quantity, sequentially multiplying all support synthesis quantities to obtain a comprehensive support item, sequentially multiplying the uncertainty function values to obtain a comprehensive uncertainty item, subtracting the comprehensive uncertainty item from the comprehensive support item to obtain a net support quantity, multiplying the comprehensive uncertainty item by one minus the number of radiation source centers to serve as a correction item, adding the correction item to the comprehensive support item to obtain a normalization item, and dividing the net support quantity by the normalization item to obtain a global confidence value of the overlapped pulses belonging to the radiation source centers.
9. The method for sorting radar signals based on rough set theory and adaptive weighting according to claim 1, wherein the specific process of marking noise pulses and outputting signal sorting results is as follows: Comparing the global confidence value of each radiation source center by the overlapped pulse, selecting the radiation source with the largest global confidence value as the candidate attribution, simultaneously comparing the largest global confidence value with a confidence judgment threshold value, classifying the overlapped pulse to the corresponding radiation source when the largest global confidence value is not smaller than the confidence judgment threshold value, judging the overlapped pulse as an uncertain pulse and marking the same as a noise pulse when the largest global confidence value is smaller than the confidence judgment threshold value; And summarizing the core pulse association result and the overlapping pulse judgment result, distributing a radiation source number label for each pulse in the full pulse data stream, identifying and marking the noise pulse, and outputting a signal sorting result and a global confidence value corresponding to each pulse.
10. A radar signal sorting system based on rough set theory and adaptive weighting, applying a radar signal sorting method based on rough set theory and adaptive weighting according to any one of claims 1-9, comprising: The full pulse data acquisition module is used for carrying out full process dynamic acquisition on the aliasing full pulse signals, obtaining pulse description word sequences, and carrying out normalization, abnormal marking and isolation on the pulse description word sequences to obtain full pulse data streams; the rough set granularity entropy weight module is used for dividing the full-pulse data stream based on the pulse description word sequence to obtain granularity combination entropy, carrying out equivalence class division and constructing a global weighted distance matrix; The core pulse rapid association module is used for carrying out local density estimation based on the global weighted distance matrix, constructing a decision diagram and screening pulse points to obtain a shared nearest neighbor number; And the overlapping pulse evidence fusion module is used for constructing basic probability distribution and uncertainty function values, generating global confidence coefficient, marking noise pulses and outputting signal sorting results.

Description

Radar signal sorting method and system based on rough set theory and adaptive weighting Technical Field The invention relates to the technical field of baseband communication, in particular to a radar signal sorting method and system based on a rough set theory and self-adaptive weighting. Background The conventional non-supervision sorting method generally adopts a fixed distance measure to correlate with a single clustering criterion after the multidimensional parameters of pulse descriptors are normalized along with the rapid change of a radar system and the aggravation of multi-source signals in a complex electromagnetic environment, so that the stability difference of different parameter dimensions in different scenes is difficult to be described, and when carrier frequency and amplitude parameters fluctuate strongly or are influenced by shielding and measuring noise obviously, the fixed measure can easily misjudge the fluctuation as the radiation source difference, so that the problems that homologous pulses are split, heterologous pulses are mixed in an overlapping area, center point drift is further caused, the continuity of core pulses is damaged, the attribution uncertainty of the overlapping pulses cannot be quantified and the like are caused. For example, the invention patent with the bulletin number of CN118444275B discloses an intelligent sorting method of parameter agile radar signals, which comprises a feature extraction module, a feature fusion module, a radiation source mapping module and an intelligent sorting network construction and training module, wherein the feature extraction module is used for constructing a network based on multi-branch cavity convolution to extract pulse parameter features, the feature fusion module is used for constructing a network based on an attention mechanism to fuse the extracted features, the radiation source mapping module is used for realizing the mapping of the features and the radiation sources based on the transposed convolution, the intelligent sorting network construction and training module is used for constructing a deep sorting network in a cascading mode, and performing supervised training by utilizing a large number of marked radar signal samples, and finally sorting the parameter agile radar signals by utilizing the trained network. For example, the invention patent with the publication number of CN119596244B discloses a radar signal spectral clustering sorting method based on SOM anchor point extraction and graph fusion, which comprises a data preprocessing module, an anchor point extraction module, a graph fusion module and a spectral clustering module, wherein the data preprocessing module is used for carrying out normalization processing on radar pulse parameters and constructing a KNN graph, the anchor point extraction module is used for iteratively training and extracting topological structure anchor points of data by utilizing a self-organizing map (SOM), the graph fusion module is used for calculating the similarity between the anchor points and nodes to construct an adaptive anchor graph and carrying out weighted fusion with the KNN graph, and the spectral clustering module is used for constructing a Laplace matrix based on the fusion graph and realizing final signal sorting by calculating feature vectors and combining a K-means algorithm. In the prior art, under the scene that parameters are rapidly changed along with time and system, partial fields are in strong fluctuation or are obviously influenced by channels and shielding, so that homologous pulses are discretely increased in the dimensions, and meanwhile, multiple sources are naturally close in a plurality of dimensions to form inter-class overlapping. The fixed metric may misinterpret the fluctuations and measurement noise as source differences, causing the homology to be broken down and the overlap region to be blended. Therefore, in view of the above problems, there is a need for a method and system for radar signal sorting based on rough set theory and adaptive weighting. Disclosure of Invention Technical problem to be solved Aiming at the defects of the prior art, the invention provides a radar signal sorting method and a radar signal sorting system based on a rough set theory and self-adaptive weighting, which solve the problem that the stability difference of pulse parameters cannot be self-adaptively described under the fixed distance measurement and the fixed weight in the existing unsupervised sorting. Technical proposal The radar signal sorting method based on the rough set theory and the self-adaptive weighting comprises the following steps of S1, carrying out overall process dynamic collection on an aliasing whole pulse signal, obtaining a pulse description word sequence, normalizing, abnormally marking and isolating the pulse description word sequence to obtain a whole pulse data stream, S2, dividing the whole pulse data stream based on the pulse description wor