Modified Hausdorff Distance Transform Technique for Video Tracking

The recent advances in computer and DSP technology have brought new expectation as well as new challenges in real time visual tracking of moving targets from video sequences. Due to the processing time constraint and continuous change in target size, segmentation and feature extraction for matching becomes very unreliable. In this paper we have presented an automated correlation based tracking approach using edge strength and Hausdroff Distance Transform (HDT) technique for tracking of moving targets. The system works in three states, Locking, Tracking and Recovery State. Difference image processing technique or manual selection is used in locking state to identify the moving target. Tracking state tracks the identified target over subsequent video frames. The system has been successfully tested with various real video sequences. The tracking performance evaluation and comparative results using Sobel’s edge and HDT template are presented considering the image sequence where object is under high background noise condition.