D-Atr For Sar Images Based On Deep Neural Networks
Zongyong Cui,Cui Tang,Zongjie Cao,Nengyuan Liu
DOI: https://doi.org/10.3390/rs11080906
IF: 5
2019-01-01
Remote Sensing
Abstract:Automatic target recognition (ATR) can obtain important information for target surveillance from Synthetic Aperture Radar (SAR) images. Thus, a direct automatic target recognition (D-ATR) method, based on a deep neural network (DNN), is proposed in this paper. To recognize targets in large-scene SAR images, the traditional methods of SAR ATR are comprised of four major steps: detection, discrimination, feature extraction, and classification. However, the recognition performance is sensitive to each step, as the processing result from each step will affect the following step. Meanwhile, these processes are independent, which means that there is still room for processing speed improvement. The proposed D-ATR method can integrate these steps as a whole system and directly recognize targets in large-scene SAR images, by encapsulating all of the computation in a single deep convolutional neural network (DCNN). Before the DCNN, a fast sliding method is proposed to partition the large image into sub-images, to avoid information loss when resizing the input images, and to avoid the target being divided into several parts. After the DCNN, non-maximum suppression between sub-images (NMSS) is performed on the results of the sub-images, to obtain an accurate result of the large-scene SAR image. Experiments on the MSTAR dataset and large-scene SAR images (with resolution 1478 x 1784) show that the proposed method can obtain a high accuracy and fast processing speed, and out-performs other methods, such as CFAR+SVM, Region-based CNN, and YOLOv2.