基于Kinect深度信息的手势识别

Gesture recognition based on Kinect depth information

设计了一种基于Kinect深度信息和双阈值分割的运动手势识别算法。结合OpenCV和OpenNI,在vs2010环境下实现了该算法。利用Kinect的深度摄像头获取深度图像;对该图像进行双阈值分割,获取手部图像;再对手部图像进行形态学处理,获取完整的手形;最后,利用OpenNI的手势生成器GestureGenerator对手势进行跟踪识别。利用深度图像进行手势识别,通过双阈值分割,不仅去除了背景干扰,也能去除一部分前景干扰。用不同颜色点、圆和线的形式表示各种手势,可以清晰地实时显示识别效果。

An algorithm of moving gesture recognition based on the depth information of Kinect and dual threshold segmentation is designed. The algorithm is realized under VS2010 by integrating OpenCV and OpenNI. Firstly, the depth camera of Kinect is used to get the depth image. Secondly, the dual thresholds are used to segment the image to get the image of just hand. Then morphology process is performed for the image to achieve more integrity hand shape. Finally, the GestureGenerator of OpenNI is utilized to track and recognize the gesture. In this method, the obstructions of background and part of foreground are removed by dual threshold segmentation. Points, circles and lines in different colors are used to describe the gestures, so the results of recognition can be clearly shown in real time.