A UAV patrol system using panoramic stitching and object detection

Abstract

Due to their increasing popularity, drones are widely used for security purposes, such as patrols, now. There are two key issues in drone patrol: 1) showing the patrolling scene in a simple and intuitive manner; 2) detecting and highlighting objects of interest. To this end, we propose a UAV patrol system based on panoramic image stitching and object detection. This system uses the SPHP algorithm combined with the region growing algorithm based on difference images to generate a panorama image and to eliminate motion ghosts. It further adopts the popular image object detector Faster RCNN to detect objects while using knowledge about the scene categories to refine the classification scores of the objects. The proposed system is evaluated on the VisDrone video detection dataset. The experiments show that the system provides good results regarding panoramic stitching and object detection in drone videos.