SAR Texture Image Classification Method Based on Deep Neural Network

Abstract

The present invention proposes a SAR texture image classification method based on a deep neural network, which mainly solves the problem of low classification accuracy of SAR texture images with a large number of samples and a large number of feature dimensions in the prior art. The implementation steps are: (1) extract the low-level features of the SAR image; (2) train the low-level features of the SAR image through the first layer RBF neural network of the deep neural network to obtain the high-level features of the image; (3) obtain the high-level features of the image through the deep neural network The second layer of RBM neural network trains advanced features to obtain more advanced features of the image; (4) trains more advanced features through the third layer of RBF neural network of the deep neural network to obtain image texture classification features; The texture classification feature is compared with the test sample label, and the parameters of each layer of the deep neural network are adjusted to obtain the optimal test classification accuracy. The invention has high classification accuracy and can be used for target recognition or target tracking.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a multi-sample, multi-category, and complex-featured SAR texture image classification method based on a deep neural network, which can be used in target recognition, target tracking and other fields. Background technique [0002] Synthetic aperture radar (SAR) is widely used in the field of earth science remote sensing. SAR texture image classification is the application of pattern recognition in SAR image processing. It converts the image data from the two-dimensional gray space to the target pattern space. The result of the classification is to divide the image into multiple different categories according to different attributes. subregion. The reliable classification features of SAR images are mainly grayscale features and texture features, but the results obtained by using grayscale features for classification in practical applications are not very ideal...

AI Technical Summary

Problems solved by technology

However, traditional machine learning and signal processing methods are shallow learning structures with only a single layer of nonlinear transformation.

Images