Combining Denoising with Super Resolution for Target Detection and Recognition of SAR Image Based on Deep Learning
WANG Jun1, WANG Sai1, REN Yuming1, CHEN Dehong2, CUI Shan2, WEI Shaoming1
1. School of Electronics and Information Engineering, Beihang University, Beijing 100191, China;
2. Shanghai Electro-Mechanical Engineering Institute, Shanghai 201109, China
Abstract：Combined with the development of deep learning, using deep convolution network model, this paper integrates feature extraction and target recognition into one model, so that the system can automatically extract target features and give category decisions. Denoising is usually needed for images with low signal-to-noise ratio, but the denoising ability is not proportional to the accuracy of detection and recognition. The denoising method adopted in this paper not only improves the image quality, but also effectively improves the recognition accuracy. It avoids the reduction of accuracy caused by the loss of details. At the same time, in order to further improve the recognition effect of low SNR images, a super-resolution network based on Convolutional Neural Networks (CNN) is used to provide conditions for obtaining more feature information. In addition, in order to solve the problem of incomplete data sets, some research on the sparse azimuth of MSTAR data sets is made, which can maintain a high recognition rate under the condition of fewer training samples.
王俊, 王赛, 任俞明, 陈德红, 崔闪, 魏少明. 结合深度学习去噪和超分辨的SAR检测识别[J]. 空天防御, 2020, 3(3): 24-30.
WANG Jun, WANG Sai, REN Yuming, CHEN Dehong, CUI Shan, WEI Shaoming. Combining Denoising with Super Resolution for Target Detection and Recognition of SAR Image Based on Deep Learning. Air & Space Defense, 2020, 3(3): 24-30.
