A Study of Anti Occlusion Human Face Tracking by Improving CAHSHIFT

• Main Authors: Sian-Chang Lin, 林顯昌
• Other Authors: Ruei-Yao Wu
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/4qg68c

碩士 === 世新大學 === 資訊管理學研究所(含碩專班) === 96 === In the image tracking research, CAMSHIFT algorithm features the high computing efficiency, and the high tracking accuracy, so as to receive a lot attention in recent years. However, the CAMSHIFT algorithm is easily interference by neighboring objects with similar color called occlusion. The main reason is that the ROI adaptive mechanism of CAMSHIFT does not limit the region of adaption, and it results in tracking adjoining areas with similar color. To solve this problem, the study proposes a method named Improved Continuously Adaptive Mean Shift, (ICAMS). Otherwise, each object must be required some features which can make up the tracking process. The study’s object is human face, and the ratio of face length to width usually between 1.3 and 1.5, by this reason the study proposes the Track Window Size Limits(TWSL) technique to cooperates with ICAMS, to limit the shape and range of ROI. The proposed method successfully resists the interference by neighboring objects with similar color.