Visual tracking strategy based on fuzzy multi visual cue integration

To improve the adaptability of visual tracking system for various circumstances in space, a visual tracking strategy based on fuzzy multi visual cue integration is proposed, which synthesizes multi visual cue together to represent the features of object. The recognition of object is intelligentized by integration of each fuzzy cue. To deal with the effects of noise on segmentation threshold of moving object, a new region-based segmentation is chosen to realize stable clustering of moving region by integrating population statistics and region growing. Compared with conventional local feature recognition, this strategy is adaptive to complex and dynamic circumstances. Redundant information complements each other for ensuring feature recognition and tracking in uncertain environment. Experimental results demonstrate the robust feature recognition and tracking with this strategy, especially in part occlusion and unstable lighting.