A target identification method and system based on a radar one-dimensional distance image

Abstract

The invention discloses a target identification method and system based on a radar one-dimensional distance image. The echo data of a training sample pair is subjected to Fourier Mein transform processing, and then the echo data is input into a one-dimensional convolutional neural network model for training. The Fourier Meilin transformation has time shift invariance and scale invariance, and thescale invariance can enable the distance image to be insensitive to changes of azimuth angles and attitude angles in a Meilin transformation domain; Time shift invariance enables Fourier melin transformation to adapt to time translation and scale change of signals; Therefore, a stable feature vector can be obtained after the one-dimensional distance image is subjected to Fourier Mellin transformation, the attitude sensitivity, the translation sensitivity and the amplitude sensitivity of the distance image are reduced, the robustness of the distance image is enhanced, and the accuracy of targetidentification is improved.

Description

technical field [0001] The invention relates to the technical field of radar, in particular to a target recognition method and system based on a radar one-dimensional range image. Background technique [0002] In the current field of target recognition, the recognition of radar one-dimensional range images uses the classic process of pattern recognition, and the preprocessing of one-dimensional range images mainly uses envelope alignment to reduce its attitude and translation sensitivity. Among them, envelope alignment does not work well when the local noise of the range image is relatively strong, and methods such as non-coherent averaging to improve the signal-to-noise ratio of range image data are based on envelope alignment. Therefore, the translation sensitivity and attitude sensitivity of the one-dimensional range image will reduce the accuracy of target recognition. Contents of the invention [0003] The purpose of the present invention is to provide a target recog...

AI Technical Summary

Problems solved by technology

[0002] In the current field of target recognition, the recognition of radar one-dimensional range images uses the classic process of pattern recognition, in which the preprocessing of one-dimensional range images mainly uses envelope alignment to reduce its attitude and translation sensitivity, but in actual situations Among them, envelope alignment does not work well when the local noise of the range image is relatively strong, and methods such as non-coherent averaging to improve the signal-to-noise ratio of range image data are based on envelope alignment.

Therefore, the translation sensitivity and attitude sensitivity of the one-dimensional range image will reduce the accuracy of target recognition

Images