CGA: A New Approach to Estimate the Geolocation of a Ground Target from Drone Aerial Imagery

In this paper, we address the geolocation problem of a target on the ground from aerial images captured by the drone camera. We propose a new approach named CGA (Color Geolocation Approach) where we estimate the 3D world position of the object after detecting the 2D position of the object in the captured image. Our CGA approach is based on position encoding-decoding and uses the DEM (Digital Elevation Model) model of SRTM (with resolution of one arc second) to model the ground terrain and consists of encoding the world positions of each point with a unique color and later decoding it to recover the world position. The input data are the 2D position of the target detected in the captured image, the altitude and GNSS/GPS location of the UAV and the camera attitude using an IMU installed on top of the camera. We have implemented CGA using a 3D engine environment with the API OpenGL. One of the applications of our approach is an aid to SAR missions (Search and Rescue) to locate and help people stranded in disaster areas and immigrant paths in the great Sahara.