基于深度图像的人体骨骼点跟踪矫正方法

A Tracking Correction Method Depth Image-Based for Human Skeleton

针对人体运动骨骼跟踪中由于遮挡经常出现的骨骼点漂移和偏离现象,提出一种部位圆限定方法来矫正和优化骨骼点坐标。该方法通过滤波和人体像素分类、部位识别,重新确定了Kinect SDK骨骼跟踪系统的20个关节点坐标。这一方法不易受外界环境因素的干扰,能较好地克服部位遮挡问题,从而获得高质量的跟踪图像。

Aiming to the questions of the skeleton points drift and deviation for tracking human motion skel- eton, a novel method of limited part of the circle is proposed to correct the keletal point coordinates. In this method, the 20 joints in Kinect SDK coordinates are rebuilt by means of filtering, human pixel classi- fication and location identification. Results of research show that the method proposed can overcome the problem of caused by ambient occlusion effectively, and to achieve excellent performance.