Dynamic Visual Tracking Based On Multi-Cue Matching

Due to attractive features such as "computation efficient" and "model free", the template matching approaches have been used in many dynamic visual tracking systems to detect moving targets. However, in practical visual tracking applications, the image of the moving target may experience shape variations due to movement or illumination changes, in which it could end up with a detection failure. To overcome these difficulties, a multi-cue matching approach with a fuzzy logic based data fusion mechanism is proposed in this study. To evaluate the performance of the proposed approach, a real-time pan-tilt dynamic visual tracking system is used as a test platform and several experiments have been conducted. Experimental results demonstrate the effectiveness of the proposed approach.