Sparse SAR (Synthetic Aperture Radar) target classification method and device based on transfer learning

Abstract

The invention discloses a sparse SAR target classification method and device based on transfer learning. The method comprises the following steps: (1) reconstructing a sparse SAR image based on a matched filtering SAR image by using a BiIST algorithm; (2) constructing convolutional neural networks with the same structure in a source domain and a target domain, and pre-training a neural network in the source domain by using the simulation data set; (3) migrating partial parameters in the neural network pre-trained in the step (2) to a network of a target domain, and randomly initializing other parameters; and (4) finely adjusting the network, taking the sparse SAR image obtained in the step (1) as input data for training, and outputting an obtained target classification result and accuracy. The sparse reconstruction algorithm adopted by the invention can effectively suppress the side lobes and clutters of the SAR image, improve the quality of the image, and provide guarantee for the training of a subsequent classification network; according to the sparse SAR target classification method based on transfer learning, the speed of network training convergence can be increased, and the target classification precision is further improved.

Description

technical field [0001] The invention belongs to the field of radar image processing and target classification, and in particular relates to a method and device for classifying sparse SAR targets based on migration learning. Background technique [0002] Synthetic Aperture Radar (SAR) is a high-resolution microwave remote sensing observation system, which is mainly carried on airborne and spaceborne platforms. Different from traditional radar, SAR can work under all-weather and all-weather conditions, and has a certain surface penetration capability, playing an irreplaceable and important role in the military and civilian fields. [0003] In 2012, the AlexNet deep convolutional neural network (Convolutional Neural Networks, CNN) model designed by Krizhevesky et al. won the ImageNet competition in one fell swoop, making deep learning a research hotspot in the field of image classification. However, collecting labeled SAR images is very expensive and difficult compared to larg...

AI Technical Summary

Problems solved by technology

However, collecting labeled SAR images is very expensive and difficult compared to large-scale labeled datasets in optical imagery

Images