【摘要】 在分析肤色的信息特征和手势的运动特性基础上,构建了一种基于单目视觉的手势分割方法,可从视频图像序列中获取有意义的手势区域.针对确定的背景图像,以10帧的采样间隔采集手势图像,通过肤色信息特征来获得手势的肤色区域,同时通过差分法获得手势的运动区域,再将肤色区域与运动区域进行融合处理获得初始的手势区域,在此基础上,进一步研究了数学形态学的腐蚀和膨胀算法,实现了视频图像序列中的手势分割与定位.实验结果表明,该方法具有较好的手势分割效果和较高的定位准确度.更多还原

【Abstract】 A kind of gesture segmentation method based on monocular vision was built for the analysis of the skin color information characteristics and gesture movements features,with which a meaningful gesture region could be obtained from the video image sequence.For the identified background image,the skin color region of gestures was captured through the analysis of the skin color information characteristics with a sampling interval of 10 frames.At the same time,the movement region of gestures was captured through the difference method,and then the initial gesture region could be obtained after the fusion of the skin color region and the movement region.The erosion and dilation algorithm of mathematical morphology were further studied,and the gestures segmentation and positioning of the video image sequence were realized.The experiment results have shown that the method has a good gesture segmentation results and higher positioning accuracy.更多还原