摘要
为提高对手势特征点检测的精度和速度,提出一种快速、准确的复杂背景下手势特征检测方法。利用基于肤色亮度的手势分割方法将手势提取出来,获得手势轮廓,针对不同手型,利用改进的曲率算法对轮廓点进行初步筛选,通过对候选点区域内分布特点的分析,最终确定特征点位置。进一步提出了新的特征点分离算法。与已有算法相比,该算法提出基于亮度分布的肤色模型,增强了手势分割的鲁棒性;引入了手型分析尺度和特征点分布分析,对手势特征位置由粗到精进行检测,改善了算法的精度和时间开销。实验表明,该方法在复杂背景下对手势特征的检测有较高的准确率和实时性,有效地减少了传统曲率方法中干扰点的误判和特征点的冗余,对不同大小的手型具有一定的鲁棒性。最后,通过在运动三维人手跟踪中的成功运用,验证了该方法的有效性。
To improve the accuracy and speed of gesture features detection,a fast and accurate gesture features detection method under complex background was put forward.Gesture was segmented to obtain the contour by hand segmentation method based on the proposed color luminance.According to different hand types,the contour points were filtered to obtain candidate points by using improved curvature algorithm.Finally,the feature positions were determined by means of analyzing the distribution of the candidate points in the vicinity.And then,a new separation algorithm of feature points was proposed.Compared to the existing algorithms,the main innovations of proposed method lied in the fact that,a new skin model to enhance the robustness of gesture segmentation was proposed;Through introducing scale for different hand types and distribution of hand features,the feature positions were gradually extracted with high accuracy and low time consumption.The experimental results demonstrated that,the proposed method could detect feature under complicated circumstance with higher accuracy and real time performance,and it also reduced the error conclusion of interference and redundancies of feature points in traditional curavture method.It had robustness for different types gestures.The method was proved to be feasible and valid by applying successfully in the studies of 3D human hand tracking.
出处
《计算机集成制造系统》
EI
CSCD
北大核心
2011年第11期2333-2342,2340-2342,共10页
Computer Integrated Manufacturing Systems
基金
国家自然科学基金资助项目(61173079
60973093
60873089)
山东省自然科学杰出青年基金资助项目(JQ200820)
教育部新世纪优秀人才支持计划资助项目(NCET-10-0863)
国家863计划资助项目(2009AA043506-2)
山东省自然科学基金重点资助项目(ZR2011FZ003)~~
关键词
图像处理
手势特征检测
曲率算法
特征点分析
三维人手跟踪
image processing
gesture features detection
curvature algorithm
characteristic analysis
3D human hand tracking
