Combination of Multi-Scale Convolutional Networks and SVM for SAR ATR

We apply a novel method to the challenge of synthetic aperture radar (SAR) automatic target recognition (ATR) by combining multi-scale Convolutional Networks (ConvNets) and Support Vector Machine (SVM). The multi-scale architecture permits the classifier to receive more features from different levels of ConvNets, and we can train the model more thoroughly. The application of SVM excels the original Softmax classifier on the nonlinear classification tasks. We conduct our experiments on the Moving and Stationary Target Acquisition and Recognition (MSTAR) database. Average classification accuracy in our experiments can achieve 99.42% on ten-class targets.