基于Kinect深度图像信息的手势分割和指尖检测算法

为克服传统二维彩色图像处理算法易受周围环境、光照变化、背景等因素的影响，提出利用Kinect深度图像信息，实现一种快速鲁棒的手势分割与指尖检测算法。首先，根据Kinect得到的深度信息对非人体部分图像进行筛选，得到包含人手的人体图像；然后对当前得到的人体图像进行直方图分析，计算能够区分人手与非人手的阈值，并通过该阈值对人体图像进行分割得到人手图像；最后，对人手图像进行形态学处理，计算掌心位置，并提取手部轮廓，结合人手轮廓关键几何特征对指尖进行有效检测。实验表明，该方法能够实时、有效地对指尖进行检测。

Hand Gesture Segmentation and Fingertip Detection Based on Depth Image of the Kinect

To overcome the traditional two - dimensional color image processing algorithm which is suscep-tible to the surrounding environment, illumination and background, a fast and robust hand gesture seg-mentation and fingertip detection algorithm based on Kineet depth image information was presented. First,the depth information of Kinect was used to filter the non - human part of the image and keep the humanbody part which contains hands ; second, the histogram of the human body part which contains hands wasanalyzed to calculate the threshold value which is able to distinguish the torso and hand, then get the handregion by segmenting the image with the threshold value ; finally, the image of hand region was processedmorphologically to calculate the palm position and extract the contours of the hand. The fingertips was de-tected effectively by the key geometric features of the contours. Tests verified the speed and accuracy ofthis method.