基于Kinect深度信息的指尖检测  

Fingertip Detection Based on Kinect Depth Image

作　　者：徐春凤[1] 王蒙蒙 翟宏宇[1] 胡汉平 XU Chunfeng, WANG Mengmeng, ZHAI Hongyu, HU Hanping(School of Compuler Science and Technology, Changchun University of Science and Teclmology, Changchun 13002)

机构地区：[1]长春理工大学计算机科学技术学院,长春130022

出　　处：《长春理工大学学报（自然科学版）》2017年第6期115-118,104,共5页Journal of Changchun University of Science and Technology(Natural Science Edition)

基　　金：吉林省自然科学基金项目(20150101015JC);吉林省重大科技攻关项目(20170203004GX)

摘　　要：指尖检测是目前人机交互研究的热点之一,针对基于普通摄像机的指尖检测容易受到复杂背景的影响而无法准确定位的问题,提出了基于Kinect深度信息的指尖获取方法。利用Kinect获取的深度信息将手单独分离出来,从而更好的实现手或手指与虚拟物体的交互。研究了一种利用NITE库函数定位手的位置的方法,根据获取的手的位置能够迅速而准确的从复杂背景中提取手部信息。再利用道格拉斯-普克算法得到手的轮廓曲线,最后利用轮廓分析法从凸包点中识别出指尖。实验表明该方法能够精确的定位到手的各个指尖位置,识别率达到80%,该方法实现简单,实时性好,有较好的鲁棒性。Fingertip detection is one of the hot topics in human-computer interaction research. Aiming at the problem that fingertip detection based on ordinary camera is susceptible to the complex background. A fingertip acquisition method based on Kinect depth image is proposed,which separates the hand from the depth information obtained by Kinect,so as to better realize the interaction of hand or finger with virtual object. This paper studies a method to locate the hand position by using NITE library function. The hand information can be extracted from complex background quickly and accurately according to the position of the hand.Using Douglas-Peucker algorithm to obtain the contour curve of the hand,and finally identify the fingertip from the convex hull points using contour analysis. Experiments show that the proposed method can accurately locate the fingertips of each hand,and the recognition rate is 96.75%. The method is simple and has good robustness,also it is real-time.

关 键 词：KINECT 深度信息 人机交互 道格拉斯-普克算法 

分 类 号：TP311.41[自动化与计算机技术—计算机软件与理论]