Image Registration by Affine-ORB on SAR and UAV Images to Find Object Geolocation

Matching two different images synthetic aperture radar (SAR) and map image (Optical) are challenging tasks because of changes in scene illumination intensity value, low image resolution transparency, distinguishing the background and foreground, change in the observation angle of the object and etc. Image registration plays a crucial role in this area. Where it can find similar features between these two kinds of images that images taken at different times with different angles and perspectives. In this research, image registration is utilized based on the Affine-ORB method to extract features and key points and then find the geolocation of each random point on the real map. To do that the Homography matrix is used to transform features from Optic image to SAR image and inverse. Since we have information about the geolocation of optic center point as a JSON file, could be able to calculate the geolocation of each random point based on WGS84 and Universal Mercator projection. Whereby using map-center point could be able to calculate map coordinates shift in pixels (Δx, Δy) and then calculate shift in meters by using (Scale * shift in pixels), after that converted map center geolocation from degree to global meters and added a shift in meters to a location in meters, finally converted shift location in meters to a degree (WGS84) to obtain a new geolocation value. Experimental results show that the performance of the proposed method is quite good and the new geolocation is quite close to real geolocation, with a bit of global shift.