摘要
本文提出了一种改进的CAMSHIFT算法,实现了对动态手势的实时跟踪。将图像序列通过一个肤色概率查找表转换为肤色概率分布图,结合运动信息和肤色概率分布,初始化一个搜索窗的大小和位置,然后对肤色概率分布图进行迭代运算,定位出手势的位置和大小,根据计算肤色概率分布图中手势区域的2阶矩,最终确定感兴趣区域的长度和宽度。实验表明,本文提出的算法能实时处理视频图像,避免了错误跟踪,而且跟踪准确精细。
This paper presents a modified CAMSHIFT algorithm which implements dynamic hand tracking m a real-time way. For each video frame, the raw image is converted to a flesh colour probability distribution image via a flesh colour probability table. Considering the hand motion and flesh colour probability distribution, the size and location of a search window are initialized. Then the size and location of the hand are determincd via this algorithm. The final width and height of ROI are found by computing the second moment of the hand area in the flesh colour probability distribution image. The current size and location of the tracked hand are reported and used to set the size and location of the search window in the next video image. Experiments show this method is robust to noise and distractors. And it can track the hand accurately in a real-time way.
出处
《计算机工程与科学》
CSCD
2005年第10期44-47,共4页
Computer Engineering & Science
基金
山东省自然科学基金资助项目(Y2001G04)
中国科学院沈阳自动化研究所开放课题基金资助项目
