基于Kinect传感器的近场手势识别追踪系统的实现

Design and implementation of a near-field gesture-driven system on the kinect sensor

利用Kinect深度传感器所获取的图像深度信息实现手部从背景中的分割,并通过零均值离散高斯滤波、二值化、取最小外包矩形、欧式距离变换等一系列过程对手势目标进行识别,最后把得到的识别结果显示在电脑上,实现实时追踪。该系统相较其它类似系统具有算法简单,实时性好,实现成本低等特点。

First,hand contours are split from the background based on depth information from images got by the Kinect depth sensor.Mter a Zero-mean Dispersed Gauss Filter processing,the hand contours are binarized with a certain threshold value.And then,a minimum matrix surrounding the hand contours are given.Centroid of this matrix is considered to be the track of this hand.Skeleton information also can be obtained after an euclid distance transform. At last,recognition result are translated into a red point shown on the screen.