An Improved Receptive Fields Network for Matching Remote Sensing Images

Abstract

We present a new network combining Residual Network (ResNet) and Receptive Fields Network (RF-Net) for matching remote sensing images. Firstly, a new remote sensing image datasets are setup, which consist of images and homograph matrices. The images are obtained by cropping, illumination changing and affine transforming of the original remote sensing images. The matrices are obtained by calculating the homograph between different images of one sequence. Next, a dual-channel network structure is proposed for keypoints detection. The network consists of Receptive Fields Detection (RF-Det) and ResNet for extracting receptive feature maps with detail information and the deep layer maps with semantic information. Then descriptors of these keypoints are generated using a L2-Net. Finally, the strategies of the nearest neighbor, nearest neighbor with a threshold and nearest neighbor distance ratio are used for matching descriptors. Experimental results demonstrate its superior matching performance with respect to the original RF-Net.