SAR Image Target Recognition Method Based on Fractional Spectral Sparse Representation

Abstract

The invention discloses a SAR image target recognition method based on fractional spectrum sparse representation. The method of the invention includes two processes: constructing an atomic library with a training image set and classifying and identifying the image to be tested. Specifically, in terms of feature extraction, the Two types of features in the score spectrum domain and image pixel domain; in terms of feature classification, using the correlation between the image score spectrum and pixel features, a new joint sparse representation and sparse classification method is designed to classify and identify the image to be tested. By using the score domain The time-frequency transformation obtains the high-resolution spectral features of the signal, and performs joint sparse representation and sparse classification with the image space pixel domain features, which effectively improves the problems of existing methods in feature extraction and feature classification, and improves SAR target recognition. the accuracy rate.

Description

technical field [0001] The invention belongs to the technical field of image processing, and relates to an image classification and recognition method, in particular to a method for recognizing a synthetic aperture radar (Synthetic Aperture Radar, SAR) image target by using fractional spectrum sparse representation. Background technique [0002] Synthetic aperture radar (SAR) is a coherent imaging radar operating in the microwave band. With its high-resolution and all-weather, all-time, and large-area imaging detection capabilities, it has become an earth observation method that is generally valued by countries all over the world, and has a good application prospect. SAR target recognition is an important aspect of SAR image interpretation, the purpose is to realize the automatic classification and recognition of interested targets in the image. The SAR target recognition process usually includes two steps: feature extraction and feature classification. [0003] In terms of...

AI Technical Summary

Problems solved by technology

However, there are the following problems in SRC: the SRC process is to select a series of atoms in the training image set (or atom library), use them to perform sparse linear representation on the test image, and then use the class with the smallest representation error as the final category of the test image

In this process, atom selection is very critical, but current methods tend to select a group of related atoms to represent a test sample, and cannot guarantee that the selected atoms come from the same correct class, resulting in similar local features may be Atom group representations of different classes, affecting classification results

Images