Target Tracking Under Complex Environment

Target tracking is one of core technologies in computer vision, and has the extremely vitalsignificance and the widespread practical value. It fuses the advanced knowledge andresearch results on Image Processing, Pattern Recognition, Artificial Intelligence, AutomaticControl and Computer Applications,etc. The essential of the targets tracking technology isfocus on recognizing moving targets automatically and getting the position information. Itsdifficulty lies in the use unreliable information to obtain the reliable track effect in thecomplex application environment.This research is based on the actual project demand. Through the study on targetidentification and the target track algorithm, it aimed at finding a reasonable method to trackthe moving target under complex environment.Focusing on the target identification under complex environment, this paper adoptauto-update background model to resolve the background perturbation and illuminationchange successfully. According to targets' gray-value change constantly in two neighboringimage sequence, a new segmentation method Adaptive threshold segmentation based onkalman filter is proposed, in which the target shape is preserved and the background influenceis decreased. In order to realize the reliability of the target identification and tracking, thenoise which has the same gray-value with targets is filtered according to the targets' shapecharacters.The target's shape in this project is similar to rectangle, so the targets not only move butalso self-rotate. During moving, the obstruction and targets are overlapped. To get the targetposition and angular velocity, Susan algorithm is used to get the corner point. Then, the bestmatching position is got with the aid of point pattern matching algorithm. To reduce thematching time, the kalman filter is introduced to predict the targets' angular velocity and addtargets scale as the constraint condition, so the result meet the need of real-time processing.Under the research of the theory and the method, the software system of the "Target trackingunder complex environment" is designed based on the VC++ platform.At last, it is proved by experimentation using actual image sequence that the algorithms arefeasible and the system software work well.