基于Kinect深度信息与骨骼信息的手指尖识别方法

Fingertips recognition based on Kinect depth and skeleton data

针对基于视频的手势识别技术对手掌轮廓和指尖信息要求较高的问题,提出了一种基于图像深度信息和人体骨骼信息的手指指尖识别方法和手掌轮廓检测算法。采用微软Kinect摄像头获取深度信息和人体骨骼信息,并将每个骨骼点的三维信息转换成深度图上的二维信息。根据人体骨骼信息快速找到手掌的位置,并利用基于深度阈值的轮廓检测算法将手掌轮廓和弯曲手指轮廓从背景图像中分割出来。利用k曲率算法检测到手指指尖的位置。实验结果证明,该方法可以高效地检测出伸直和弯曲手指的轮廓,识别出人体的手指,并且该方法可在黑暗的环境下进行。

Aiming at the problem that gesture recognition technology based on video requires a lot on hand contour and fingertip, a hand contour extraction method and fingertip recognition method based on depth and skeleton data is presented. It utilizes Microsoft Kinect to capture depth and skeleton data, converts 3D information of each skeleton point into 2D on depth map. According to the skeleton data of human body the position of hand is found quickly, and the contour extraction method based on depth thresholds is used to segment hand contour from the background. The k curvature algorithm is used to detect the position of the fingertip at last. Experimental results show that the method can detect the contour of straight and curved fingers successfully, and it can be used in a dark environment.