摘要
以肢体语言中的手势为出发点,利用微软的Kinect体感设备,采集RGB彩色图像和深度图像,研究并提出一套手势跟踪识别的设计思路与实现方法。首先,利用Kinect所获取的色彩及深度信息,建立人体骨骼图从而更方便地定位手部位置并获取手部轮廓。其次,从HSI彩色空间对手部的深度图像进行二值化,并进行腐蚀、细化处理,去除手掌掌心部分。最后,通过对手指的计数定义相应操作。经实验测试,该系统具有对手势的识别功能,可以达到用空中手势动作进行人机交互的目的。
Based on the hand gestures of body language, a design idea and realization method of hand gestures tracking identification is proposed in this paper. The Kinect, a kind of Microsoft＇ s motion sensor, is used to acquire RGB image and depth image. Firstly, human diagram is established on basis of RGB image and depth image from Kinect to get the hand position and profile. After that, the depth image of hand is converted to binary image in HSI space. This new binary image can be eroded, refined and remove palm section. Finally, operations are defined by counting the amount of fingers in the remaining part of the binary image. Experiments show that this system can recognize hand gestures. It can be used in the field of human-computer interaction.
出处
《电视技术》
北大核心
2015年第21期91-94,共4页
Video Engineering
基金
江苏省属高校自然科学重大基础研究项目(13KJA510004)
全国大学生科技创新创业训练计划项目(SZDG2014001)
关键词
KINECT
骨骼跟踪
深度图像
手势识别
人机交互
Kinect
skeletion tracking
depth image
gesture recognition
human-computer interaction
