Small sample SAR target identification method based on graph attention network

Abstract

The invention discloses an SAR target small sample identification method based on a graph attention network. The method mainly solves the problem of poor recognition rate under the condition of lack of training data in the prior art, and adopts the scheme of selecting SAR images containing a radar target, suppressing speckle noise of the SAR images, and dividing the SAR images after noise reduction into data with a label and data without a label; training an auto-encoder by using the denoised image to obtain feature vectors of all SAR images; obtaining an initial adjacency matrix by utilizingvector similarity on the premise of a small amount of label data; and setting a graph attention network, iteratively training the graph attention network by utilizing all the feature vectors until anerror function of the network is converged, and outputting a finally predicted node label matrix to realize label-free data identification. According to the method, a small number of known types of SAR targets can be utilized to predict the types of a large number of other unknown targets, the prediction accuracy is high, and the method can be used for radar target recognition under the conditionof small samples.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to a small-sample SAR target recognition method, which can be used to accurately recognize radar targets on the premise of a small number of training samples. Background technique [0002] With the rapid development and progress of synthetic aperture radar technology, the resolution of SAR images has become higher and higher, and its resolution has developed from early middle and low resolution to high resolution and super high resolution. The generation of high-resolution SAR images not only improves the shortcomings of insufficient information in traditional low-resolution SAR images, but also promotes the research on SAR image processing. Target recognition based on SAR images is one of the important applications in the field of SAR image processing, and has been widely used in ocean or land monitoring and detection. [0003] Due to the feature differences between high-...

AI Technical Summary

Problems solved by technology

Then a parallel deep convolutional network structure is used to extract the SAR image features of different viewing angles of the same target. At the same time, the features learned from each viewing angle are gradually fused. This method is based on relatively few original images. Accurate classification of targets in SAR images, but this method has two deficiencies: one is to use simple data combination to expand data, and the similarity between training data is large; The amount of data in the small sample recognition problem is still large

This method extracts the deep features of the image by stacking high-speed convolution units. The deeper network makes the classification more accurate. At the same time, it achieves a good classification effect on the limited amount of training data. To obtain better recognition results, it is easy to cause over-fitting problems in the case of small samples

Images