摘要
对Kinect设备获取到的深度图像进行手部区域分割,分别比较扫描法和区域生长法的图像分割效果,并提取手部的轮廓信息实现手掌心和指尖的定位,以区域协方差作为手势的特征变量,结合指尖识别算法实现了在线手势识别。结果表明:扫描法比区域生长法所识别出的手掌点和轮廓点更多,与自适应PCA识别算法相比具有更好的识别效果,静态手势测试的指尖平均识别率达93%,在线运动手势识别的准确率均在80%以上。
The depth image obtained by the Kinect device is used to segment the hand region,and the image segmentation effects of the scanning method and the region growing method are compared respectively,and the contour information of the hand is extracted to realize the positioning of the palm and fingertips.Using the regional covariance as the feature variable of gesture,combined with the fingertip recognition algorithm,the online gesture recognition is realized.The results show that the scanning method can identify more palm points and contour points than the region growing method.Compared with the adaptive PCA algorithm,the proposed algorithm has better recognition effect,the average fingertip recognition rate of static gesture test is 93%,and the accuracy rate of online motion gesture recognition is above 80%.
作者
杨秋菊
YANG Qiuju(Department of Computer Information,Suzhou Vocational and Technical College,Suzhou 234000,China)
出处
《河南工程学院学报（自然科学版）》
2023年第2期58-62,共5页
Journal of Henan University of Engineering：Natural Science Edition
基金
校级质量工程教学研究重点项目(zy2021zlgc10)
安徽省高校自然科学重点项目(2022AH052764)。
作者简介
杨秋菊(1982-),女,安徽宿州人,讲师,主要研究方向为图形图像制作、人工智能、数据挖掘。
