A vision system for multi-rotor aircraft to track moving object

This paper presents a vision method for a multi-rotor Unmanned Aerial Vehicle (UAV) to track a moving object using combined algorithm of Camshift and Kalman predictor. Because of the real-time performance demanded by the control system, vision algorithm should calculate navigation data quickly. Camshift is a tracking algorithm based on histogram matching and color probability distribution, which has a lot of advantages such as strong adaptability and precise tracking accuracy. However, the center of tracking box generated by Camshift algorithm cannot remain stable because of inevitable noise due to the change of light and camera shaking. Thusly, Kalman predictor is implemented to erase this noise as well as predict the position of moving object when it is temporarily lost by UAV. Experimental results show that through this method, UAV can continuously tracking a moving vehicle under different background environment even object is temporarily occluded or lost by UAV.