摘要
传统的CamShift算法具有很好的实时性,但无法实现自启动,且在跟踪过程中易受噪声和颜色变化的干扰,针对这些缺点,提出了一种改进的Camshift算法.首先运用统计差分算法对移动目标快速定位,得到初始前景对象的区域,然后通过统计前景区域获取精确的目标颜色直方图模型并以此区域为原型计算肤色直方图的查询表.同时结合一些后续步骤成功实现了算法的自启动.最后用实验验证了借助统计差分与目标颜色直方图模型的CamShift算法能够实现自启动并提高Camshift算法跟踪时的准确度.
The camshift tracking algorithm has the advantage of better real-time characteristics, but it is not able to start up automatically and is susceptible to noise and color change in the tracking process. In order to solve these problems, an improved Camshift algorithm was proposed. First, a statistical difference algorithm was applied for rapid localization of mobile targets, and the initial foreground object~ region was obtained. Then, an accurate target color histogram model was obtained by gathering foreground area statistics. The query table of the skin color histo- gram was also calculated taking this area as a prototype. Meanwhile, an auto-start algorithm was successfully a- chieved with several subseauent steps. Finallv. the algorithm was validated through exneriments.
出处
《智能系统学报》
2011年第4期355-359,共5页
CAAI Transactions on Intelligent Systems
基金
黑龙江省自然科学基金资助项目(F2007-13)
黑龙江省普通高等学校青年学术骨干支持计划资助项目(1155G21)
