SAR automatic target recognition method based on multi-view deep learning framework

Abstract

The invention discloses an SAR automatic target recognition method based on a multi-view deep learning framework. The method is applied to the field of radar target recognition and aims to solve the problem that a deep neural network is hard to be trained effectively under the condition of a small quantity of original SAR images. According to the method, a small quantity of original SAR images areutilized to generate a large quantity of multi-view combination samples by use of different views of the original SAR images in combination with data collection cost and performance recognition requirements in actual application, and effective recognition information contained in the samples is increased; and based on the deep learning theory, a multi-input parallel deep neural network is constructed, effective features of different views are automatically extracted, and a category prediction result is given, so that rapid and precise recognition of an SAR target is realized. The method has the advantages of being flexible, accurate, efficient and high in generalization capability.

Description

technical field [0001] The invention belongs to the field of radar target recognition, in particular to a synthetic aperture radar automatic target recognition technology in the field of radar target recognition. Background technique [0002] Synthetic Aperture Radar (SAR) is a high-resolution microwave imaging radar with all-weather and all-weather working capabilities. Utilization and other fields have extremely high civilian and military value. Due to the electromagnetic scattering characteristics and the coherent imaging mechanism, SAR imaging is sensitive to the target azimuth angle, and there are a large number of coherent spots in the SAR image, which further increases the difference with the optical image and increases the difficulty of manual interpretation. SAR automatic target recognition (Automatic Target Recognition, ATR) is based on the theory of modern signal processing and pattern recognition. It provides strong technical support in many aspects such as pre...

AI Technical Summary

Problems solved by technology

However, this method does not increase the effective identification information of the original SAR image, the improvement of identification performance after sample expansion is very limited, and a large amount of storage space needs to be allocated

The problem that the deep neural network is difficult to be effectively trained in the case of fewer original SAR images has not yet been reasonably solved

Images