Research Of Real-time Object Tracking Based On Computer Vision

Computer vision is becoming the most important way for intelligent machines to access external information and understand the world. Moving target detection and tracking are among the most important subjects in computer vision, and have been hot issues of international research in recent years. As a technology to anlyse and synthesis the image series which contains the moving object, moving target detection and tracking integrated many advanced technologies such as image processing, pattern recognition, artificial intelligence, automatic control, computer and other related fields. It can be used in many aeraes such as security surveillance, traffic control, visual navigation, video surveillance, medical image analysis and industrial detection.This paper researches the visual detection and tracking algorithms of moving object based on image processing and Kalman filter techniques after surveying the current research literatures in this domain. Camshift, which is a visual tracking method based on the color features of moving object, has been widely accepted and used for its advantages of parameter-less and fast calculation. But the conventional Camshift algorithm uses only one Region Of Interest (ROI) to search and track the target, which will inevitably result in the tracking divergence when the object passes the background with similar color to the target. Furthermore Camshift usually cannot recapture the object after the target is lost because the algorithm relies on the color distributions alone. To remedy these drawbacks of the conventional Camshift, a modified Camshif which utilizes dual ROI to enhance the robustness of the algorithm and prevent divergence is proposed in this paper. Meanwhile, the information of the velocity of the moving object is introduced into the algorithm to solve the problem of occlusion and enable the retracking of the lost object.In order to test and verify the efficiency of the proposed algorithms, different cases of experiments are conducted. The comparative experimental results between the conventional Camshift and the dual ROI Camshif with velocity information involved demonstrate that the Camshift tracking policy proposed in this paper has advantageous performance over traditional Camshift.