A Radar Unknown Target Recognition Method Based on Long Short-Term Memory Network

Abstract

The invention belongs to the technical field of unknown target recognition, and in particular relates to a radar unknown target recognition method based on a long short-term memory network. The present invention first preprocesses the broadband radar one-dimensional range profile data (HRRP); secondly performs principal component analysis (PCA) on the feature vector extracted from the one-dimensional range profile by the long-term short-term memory network, reduces the feature vector dimension, and finally passes the nearest Neighboring method (NN) processes the low-dimensional feature vector of known target sample data, obtains the discrimination threshold, and discriminates the output vector of the long short-term memory network to identify unknown targets. For the conventional convolutional network, since the long-short-term memory network can extract the relevant information between adjacent one-dimensional range image samples, it can effectively describe the change characteristics between the input one-dimensional range image sample sequences, thereby improving the recognition performance of unknown targets .

Description

technical field [0001] The invention belongs to the technical field of unknown target identification, in particular to a radar unknown target identification method based on a long short-term memory network. Background technique [0002] One-dimensional range profile (HRRP) reflects the size of the target and the distribution of the scattering center and other structural characteristics. It has the advantages of strong real-time performance, easy acquisition and storage, and is currently the main method for identifying air targets such as aircraft. [0003] With the successful application of deep learning theory in the fields of speech, image and natural language processing, some scholars have begun to introduce deep learning methods into one-dimensional range image recognition of radar targets. For example, the convolutional neural network with translation invariance has a good recognition rate for the one-dimensional range profile of the trained radar target, but cannot cor...

AI Technical Summary

Problems solved by technology

For example, the convolutional neural network with the characteristics of translation invariance has a good recognition rate for the trained radar target one-dimensional range image, but it cannot correctly identify the unknown target (that is, the target that has not participated in the training). For this, the use of deep convolution Neural network + threshold can realize the recognition of unknown targets. However, since the convolutional neural network can not use the relevant information between adjacent samples, there is still room for further improvement in the recognition performance of unknown targets.

Images