Radar one-dimensional range profile target recognition method based on depth convolution neural network

Abstract

The invention discloses a radar one-dimensional range profile target recognition method based on a depth convolution neural network, includes the following steps: a data set is collected, the collected data is preprocessed, features are extracted from the preprocessed data, the HRRP signal is divided into two parts: low SNR and high SNR, A feature enhancement algorithm based on robust Boltzmann isconstructed, and a HRRP target recognition model based on convolution neural network and bidirectional loop neural network based on LSTM is constructed. The parameters of the network model are fine-tuned by using gradient descent algorithm, and an effective target recognition model is obtained. A radar HRRP automatic target recognition technology with small sample robustness and noise robustnessconstructed by the invention has strong engineering practicability, and a radar one-dimensional range profile target recognition model based on a convolution neural network and a cyclic neural networkis proposed from the aspects of feature extraction and the design of a classifier.

Description

technical field [0001] The invention belongs to the field of radar target recognition, and relates to a radar one-dimensional range image target recognition method based on a deep convolutional neural network. Background technique [0002] Radar automatic target recognition is based on the theory of electromagnetic scattering, by extracting the features in the radar echo signal and processing the features to determine the type, model and other attributes of the radar target. Since the 1960s, the United States has specially established a "ballistic missile early warning system". Since then, radar identification has been included in an important development plan, and HRRP (high-resolution range profile) automatic target recognition has been included in an important research and development plan. Moreover, a radar identification platform has been established in universities in the United States. On this basis, many universities in the United States use radar simulation software...

AI Technical Summary

Problems solved by technology

Compared with cooperative targets and radar HRRP target recognition with a complete HRRP database, there are two major problems in the recognition of high-speed non-cooperative targets: 1. Small sample recognition; 2. Low signal-to-noise ratio HRRP robust recognition. When the number of samples is close to or smaller than the sample dimension, a series of problems such as inaccurate parameter estimation, sharp decline in classifier recognition performance and generalization performance, etc.

The easiest way to solve these problems is to increase the number of training samples. However, usually for high-speed non-cooperative enemy targets (such as fighter jets, etc.), it is difficult for radar to detect and continuously track a large number of HRRP samples, which cannot meet the requirements of many classifiers to estimate model parameters. Basic requirements, causing the algorithm to fail

The reason for the robust recognition of HRRP with low SNR is that in actual engineering, the HRRP data set used in the training sample library is usually obtained from experiments under cooperation or directly generated by electromagnetic simulation programs, and its SNR is high; but the test The stage is generally carried out under actual battlefield conditions. On the one hand, the electromagnetic environment is very complex at this time, and the target echo will always contain a certain amount of noise, resulting in a low signal-to-noise ratio of the HRRP obtained. Radar recognition is an unavoidable mode in the practical application of radar

Images