Radar target recognition algorithm based on novel MDR-Net

Abstract

The invention discloses a radar target recognition algorithm based on a novel MDR-Net. The invention comprises a radar data preprocessing module, a multi-dimensional sparse feature extraction module (MDS), a multi-dimensional radar target recognition network (MDR-Net) based on an MDS module and a classifier module; the MDS module has very strong capability of carrying out feature extraction on sparse data, the nonlinearity is specially designed and enhanced, a high-precision MDS module (HMDS) can be adopted, and a light-weight MDS module (LMDS) can also be adopted; HMDR-Net or LMDR-Net can bebuilt on the basis of the HMDS module or the LMDS module, and HMDR-Net recognition precision is higher, the LMDR-Net can save a large amount of calculation cost and parameter storage space, the requirement of the radar automatic target recognition technology for real-time operation can be better met, and the two kinds of MDR-Net can achieve higher recognition accuracy for radar data compared witha traditional method and other neural networks. The classifier module can select a single-layer or three-layer neural network full connection layer or a global average value pooling method and the like.

Description

technical field [0001] The invention belongs to the field of radar target recognition, in particular to a novel MDR-Net-based radar target recognition algorithm. Background technique [0002] Radar Automatic Target Recognition (RATR) technology can provide key features such as target attributes, categories, models, etc. It can work around the clock and is robust to changes in the radar sensor environment. In order to obtain richer target information from radar signals, RATR technology is more and more focused on the research of high-resolution radar. Synthetic Aperture Radar (SAR) image is a kind of high-resolution radar image, compared with high-resolution range image (High Range Resolution Profile, HRRP), it can provide two-dimensional resolution information of the target, including more abundant target detail features. Traditional radar target recognition algorithms, such as K-neighbor algorithm (KNN), support vector machine (SVM), multi-task relational learning (MTRL),...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a high-resolution radar target recognition system, which can achieve extremely high recognition accuracy for high-resolution radar signals compared with traditional methods, while eliminating the steps of manual feature extraction and artificial selection of classifiers, It overcomes the difficulty of feature extraction and classification in traditional methods, with lower parameters and better real-time performance

Images