SAR Image Matching Based on Local Feature Detection and Description Using Convolutional Neural Network

Feature detection is a vital step for the image registration process whose target misalignment correction among images to increase convergency level. Deep learning (DL) in remote sensing has become worldwide sensation. Despite its huge potential, DL not reached intended concerning applications of Synthetic Aperture Radar (SAR) images. In this study, we focus on matching SAR using Convolutional Neural Network. The big challenge study how modify pretrained Visual Geometry Group model based multispectral dataset act as feature detector where it does require any prior knowledge about nature feature. Since have different characteristics from optical such dynamic range and imaging geometry, some problems arise should be considered during process. all these difficulties, results demonstrate robustness can provide descriptors that preserve localization data features. Also, proposed approach provides reasonable compared state-of-the-art methods outperforms correlation ORB descriptor under scaling. addition, may an end-to-end tool images, although calculations fine parameters are included.