基于Kinect深度和骨架信息的指尖检测方法

A fingertip detection method based on Kinect depth and skeleton data

为克服指尖检测方法易受不同光照、复杂背景和手腕信息的影响,提出了一种基于深度和骨架信息的指尖检测方法.首先对Kinect获取的深度图像进行中值滤波和形态学闭操作处理,以消除噪声和填充空洞;接着通过骨架跟踪得到的右手关节点锁定用户并进行手势分割;然后在计算手心和最高指尖点位置的基础上,利用Freeman链码提取手势左右轮廓;最后根据指尖点之间的轮廓曲线特征提取其他指尖点.实验结果表明,该方法具有良好的指尖检测效果,且对光照、背景和手腕信息鲁棒.

To overcome the fingertip detection method which is susceptible to the different illumination, complex background and wrist data, a fingertip detection method based on skeleton and depth data was proposed. First, in order to eliminate noises and fill holes, it used median filter and morphological closing operation to process the depth image captured by Kinect sensor; subsequently, it used right hand joint point obtained by the skeleton tracking for lock user and gesture segmentation; then, on the basis of calculating the positions of palm and the highest fingertip, it used Freeman Chain Code （FCC） to extract the left and right contours of the gesture; finally, the other fingertips were extracted according to the contour curve features between fingertips. Experimental results show that the proposed method has good finger detection effect and also robust to the illumination, background and wrist information.