Research on Ground Object Classification Detection Method of Heterologous SAR Image Based on Transfer Learning

Abstract

In the classification detection of Synthetic Aperture Radar (SAR) images, there are SAR images acquired from different radars. For the problem of ground objects classification detection of less-label heterogeneous SAR images, this paper uses the Convolutional Neural Network (CNN) based on transfer learning to train a pre-trained network with fewer labels. To fine-tune the network, we can realize the ground object classification detection of heterogeneous SAR images. The effectiveness of the proposed method is verified by experiments.