摘要
针对目前指尖检测算法不能同时兼顾检测率和实时性方面的不足,提出一种改进的基于角度检测的实时指尖检测算法.先用背景差和肤色相结合的方法检测出目标手区域,然后提取手区域边缘,再用改进的指尖检测算法检测出候选指尖点,并用聚类算法和凹凸性方法对筛选出的候选指尖进行进一步筛选,最后用一种新的后处理方法对获得的指尖进行排误,以提高指尖的正检率.实验结果表明,该算法在保持较高正检率的基础上,同时具有良好的实时性.
An improved real-time fingertips detection algorithm based on angle detection is proposed to deal with the limitation of low detection accuracy or poor real-time of current fingertips detection algorithms. After detecting hand region by combining background subtraction and skin color, the edge of hand region can be ex- tracted. Then the candidate fingertips can be detected by improved fingertip detection method. Subsequently, the cluster method and concavity and convexity method are used to filter the candidate fingertips. At last, a new post dealing algorithm is used to filter the false detection fingertips to improve the correct detection rate of fingertips. The experimental results show that the algorithm proposed maintains a high correct detection rate, and has a good real-time performance as well.
出处
《三峡大学学报（自然科学版）》
CAS
2014年第6期103-107,共5页
Journal of China Three Gorges University：Natural Sciences
基金
国家自然科学基金(61272237
61102155)
湖北省自然科学基金(2011CBD180)
三峡大学研究生创新基金(2012CX045)
关键词
角度检测
固定k-余弦
凹凸性
指尖检测
实时性
angle detection
fixed k-cosine
concavity and convexity
fingertip detection
real-time
