Method of Visual Tracking for Water USV Dynamic Object
DONG Huiying;ZHAO Peng;Shenyang Ligong University;
Current track Camshift algorithm has computationally intensive and poor real-time shortcoming,and a Kalman and Camshift based on a combination of methods is proposed,which can effectively improve algorithm real-time performance. Contour area of target object is gotten by object segmentation algorithm,and target centroid is estimated. Target coordinates centroid and speed is taken as Kalman filter input to predicte target position in the next frame. In searching and matching,rough and coarse position in the vicinity is adopted by using Camshift to give precise target location information,which obtaines accurate position information for Kalman filter measurement input and participates in the next round prediction,and follows by cycle to proceed. Simulation results show that the proposed method has good accuracy and timeliness under certain conditions.
【CateGory Index】： TP391.41