Large-scene SAR target recognition method based on deep neural network

Abstract

The invention, which belongs to the field of radar remote sensing application technology, in particular relates to a large-scene SAR target recognition method based on a deep neural network. A neuralnetwork structure based on a multi-scale feature map is built, multi-scale features of an SAR image are extracted, and the output of each layer of convolutional neural network is used for prediction.Meanwhile, with combination of advantages of the a deep neural network in terms of feature extraction, layer-by-layer non-linear transformation is carried out based on the neural network structure andlow-level and high-level features of an SAR image target are extracted automatically. The main four steps, including detection, authentication, feature extraction, and identification, of the traditional SAR target identification method are integrated into one neural network; and not detector, authentication device, and classifier need to be designed independently. A cutting module is added in front of the network to complete quick explanation of the large-scene SAR image.

Description

[0001] technical field [0002] The invention belongs to the technical field of radar remote sensing applications, in particular to a large-scene SAR target recognition method based on a deep neural network. The present invention builds a deep network structure based on multi-scale feature maps for SAR images, and realizes rapid recognition of large-scene SAR image targets Background technique [0003] Synthetic Aperture Radar (hereinafter referred to as SAR) can obtain high-resolution radar images all-weather and all-weather, and is an important means of earth observation. As a branch of SAR technology, SAR target recognition is of great significance in both military and civilian fields, and has become the focus of international research. [0004] SAR images are quite different from ordinary optical images in terms of imaging mechanism, geometric features, and radiation features. The image formed by SAR is not sensitive to the strength of the ground object echo, and the lay...

AI Technical Summary

Problems solved by technology

3) The method based on compressed sensing, which is mostly aimed at the target with certain characteristics, is not universal

This step-by-step SAR target recognition method, if the previous step does not get better results, will have a great impact on the results of the next step

Images