CN-121978672-A - Cluster target classification and identification method and device for multi-dimensional feature extraction

CN121978672ACN 121978672 ACN121978672 ACN 121978672ACN-121978672-A

Abstract

The invention discloses a cluster target classification and identification method and equipment for multi-dimensional feature extraction, and relates to the technical field of radar target identification. The method comprises the steps of receiving echo signals of a cluster target, carrying out pulse compression on the echo signals to obtain pulse compressed data, carrying out bispectrum analysis on the pulse compressed data based on a preset distance unit to obtain bispectrum estimation, carrying out feature extraction on the bispectrum estimation to construct a three-dimensional feature vector, and inputting the three-dimensional feature vector into a classification recognition model to classify and recognize the cluster target. The invention improves the recognition capability of the cluster targets and the universality of radar target cluster recognition.

Inventors

LI YUNLI
TANG XIANHUI
JIANG WEN
LU XUANYU

Assignees

四川九洲电器集团有限责任公司

Dates

Publication Date: 20260505
Application Date: 20260203

Claims (10)

1. A cluster target classification and identification method for multi-dimensional feature extraction, the method comprising: Receiving echo signals of a cluster target, and performing pulse compression on the echo signals to obtain pulse compressed data; based on a preset distance unit, carrying out bispectrum analysis on the data after pulse compression to obtain bispectrum estimation; extracting features of the bispectrum estimation to construct a three-dimensional feature vector; And inputting the three-dimensional feature vector into a classification recognition model to classify and recognize the cluster target.
2. The method of claim 1, wherein receiving echo signals of a clustered target and pulse compressing the echo signals comprises: Performing Fourier transform on the received echo signals in a fast time domain to obtain frequency domain echo signals; Substituting the frequency domain echo signals into a distance compression formula, and calculating to obtain pulse compressed data.
3. The method according to claim 1 or 2, wherein performing a bispectral analysis on the pulse-compressed data based on predetermined distance units to obtain a bispectral estimate comprises: selecting a preset distance unit according to the pulse compressed data; sampling the pulse compressed data in each preset distance unit in a fast time dimension and a slow time dimension; segmenting the sampled data, and performing discrete Fourier transform on each segmented data respectively; calculating a triple correlation result of the discrete Fourier transform result of each piece of segmented data; and averaging the triple correlation results of the discrete Fourier transform results of the segmented data to obtain bispectral estimation of the segmented data.
4. The method of claim 1, wherein feature extracting the bispectral estimate to construct a three-dimensional feature vector comprises: Based on the bispectral estimation, bispectral entropy, a first moment and a second moment are calculated respectively; The bispectral entropy, the first moment and the second moment are combined to construct the three-dimensional feature vector.
5. The method of claim 1, further comprising inputting the three-dimensional feature vector into a classification recognition model to train the classification recognition model prior to classifying and recognizing the clustered objects, wherein training the classification recognition model comprises: performing bispectrum analysis on the pulse compressed data based on a plurality of preset distance units to obtain bispectrum estimation; Feature extracting the bispectrum estimates of the plurality of predetermined distance units to construct a three-dimensional feature vector for each predetermined distance unit; constructing a multi-sample feature set based on the three-dimensional feature vectors for the plurality of predetermined distance units; Training the classification recognition model based on the multi-sample feature set, wherein the classification recognition model is a classification recognition model based on a feedforward neural network.
6. A clustered object classification recognition device for multi-dimensional feature extraction, the device comprising: the pulse compression module is used for receiving echo signals of the cluster targets and carrying out pulse compression on the echo signals so as to obtain pulse compressed data; the dual spectrum estimation module is used for carrying out dual spectrum analysis on the data after pulse compression based on a preset distance unit so as to obtain dual spectrum estimation; the three-dimensional feature vector construction module is used for carrying out feature extraction on the bispectrum estimation to construct a three-dimensional feature vector; and the classification recognition module is used for inputting the three-dimensional feature vector into a classification recognition model so as to classify and recognize the cluster targets.
7. The apparatus of claim 6, wherein the pulse compression module comprises: the frequency domain echo signal acquisition unit is used for carrying out Fourier transformation on the received echo signal in a fast time domain so as to obtain a frequency domain echo signal; and the pulse compression data calculation unit is used for substituting the frequency domain echo signals into a distance compression formula to calculate and obtain pulse compressed data.
8. The apparatus according to claim 6 or 7, wherein the bispectrum estimation module comprises: The distance unit determining unit is used for selecting a preset distance unit according to the pulse compressed data; the data sampling unit is used for sampling the pulse compressed data in each preset distance unit in a fast time dimension and a slow time dimension; The data segmentation unit is used for segmenting the sampled data and performing discrete Fourier transform on each segmented data respectively; a triple correlation calculation unit for calculating a triple correlation result of the discrete Fourier transform result of each segment data; and the bispectrum estimation calculation unit is used for averaging the triple correlation results of the discrete Fourier transform results of the segmented data to obtain bispectrum estimation of the segmented data.
9. The apparatus of claim 6, wherein the three-dimensional feature vector construction module comprises: the three-dimensional feature calculation unit is used for calculating bispectral entropy, first moment and second moment based on the bispectral estimation; And the three-dimensional feature vector combination unit is used for combining the bispectral entropy, the first moment and the second moment to construct the three-dimensional feature vector.
10. A clustered object classification recognition device for multi-dimensional feature extraction, the device comprising: At least one processor; At least one memory coupled to the at least one processor and storing instructions for execution by the at least one processor, the instructions when executed by the at least one processor implementing the method of any one of claims 1 to 5.

Description

Cluster target classification and identification method and device for multi-dimensional feature extraction Technical Field The invention relates to the technical field of radar target identification, in particular to a cluster target classification identification method and equipment for multi-dimensional feature extraction. Background The cluster battle is an important threat to the future battle field by using the subversion innovative battle form, the rapid development situation and the huge battle potential, and the conventional radar identification technology forms challenges. The cluster targets are more in variety, dense in distribution, complex in topological distribution, mutually coupled in electromagnetic scattering and the like, and challenges are presented to the anti-interference capability, the recognition accuracy and the processing capability of the radar. Thus, the radar is required to have a more intelligent and accurate identification capability for complex types of clustered targets. Existing radar target classification recognition technologies focus on single targets, and low-resolution radars are limited to extracting one-dimensional range profile features, or low-repetition frequency radars are limited to extracting micro-doppler features. Disclosure of Invention Aiming at the problem that the existing radar target classification and identification technology is concentrated on single targets or solves the problem of difficult identification of complex types of group targets, the invention provides a cluster target classification and identification method and equipment for multi-dimensional feature extraction, and improves the identification capability of the cluster targets and the universality of radar target group identification. The invention is realized by the following technical scheme. In a first aspect, a method for classifying and identifying a cluster target by using multidimensional feature extraction is provided, and the method comprises the following steps: Receiving echo signals of a cluster target, and performing pulse compression on the echo signals to obtain pulse compressed data; based on a preset distance unit, carrying out bispectrum analysis on the data after pulse compression to obtain bispectrum estimation; extracting features of the bispectrum estimation to construct a three-dimensional feature vector; And inputting the three-dimensional feature vector into a classification recognition model to classify and recognize the cluster target. In some embodiments, receiving echo signals of a clustered target and pulse compressing the echo signals, includes: Performing Fourier transform on the received echo signals in a fast time domain to obtain frequency domain echo signals; Substituting the frequency domain echo signals into a distance compression formula, and calculating to obtain pulse compressed data. In some embodiments, performing a bispectral analysis on the pulse-compressed data based on a predetermined distance unit to obtain a bispectral estimate comprises: selecting a preset distance unit according to the pulse compressed data; sampling the pulse compressed data in each preset distance unit in a fast time dimension and a slow time dimension; segmenting the sampled data, and performing discrete Fourier transform on each segmented data respectively; calculating a triple correlation result of the discrete Fourier transform result of each piece of segmented data; and averaging the triple correlation results of the discrete Fourier transform results of the segmented data to obtain bispectral estimation of the segmented data. In some embodiments, feature extraction is performed on the bispectral estimate to construct a three-dimensional feature vector, comprising: Based on the bispectral estimation, bispectral entropy, a first moment and a second moment are calculated respectively; The bispectral entropy, the first moment and the second moment are combined to construct the three-dimensional feature vector. In some embodiments, the method further comprises inputting the three-dimensional feature vector into a classification recognition model to train the classification recognition model prior to classifying and recognizing the clustered objects, wherein training the classification recognition model comprises: performing bispectrum analysis on the pulse compressed data based on a plurality of preset distance units to obtain bispectrum estimation; Feature extracting the bispectrum estimates of the plurality of predetermined distance units to construct a three-dimensional feature vector for each predetermined distance unit; constructing a multi-sample feature set based on the three-dimensional feature vectors for the plurality of predetermined distance units; Training the classification recognition model based on the multi-sample feature set, wherein the classification recognition model is a classification recognition model based on a feedforward neural network. I