Moving Target Detection Using Adaptive Background Segmentation Technique for UAV based Aerial Surveillance

Abstract

Detection of moving objects in a video plays a vital role in aerial surveillance. Here, both the target and the detecting camera remain in motion. In the videos captured by unmanned aerial vehicles the scenes are dynamic and time varying. Therefore the existing algorithms which are capable of detecting objects in static backgrounds are not efficient for videos obtained from aerial surveillance. Hence we propose an Adaptive Gaussian based updating model for background segmentation which is suitable for dynamic background variations. The proposed algorithm is simulated and results are compared with Adjacent frame differencing and running average based background segmentation algorithms for the videos captured by UAV. The input data are the HD videos captured by UAV - Dhaksha. The experimental results show that the proposed method works effectively in dynamic environment.