摘要
搭建了一个便利有效地双目视觉系统,结合深度信息,精准地提取手势信息。采用模板标定法对双目相机进行标定,精确地获取了相机的参数,标定的像素误差为0.67个像素;在立体匹配阶段,利用BM算法快速准确地实现了左右相机图像的匹配,获得被测手势的视差;结合三角测距原理,从而生成较为稠密的深度图;最后将深度信息重新映射到原彩色图像上,实现三维重建,生成三维云图,根据云图信息双目视觉系统能有效地从复杂背景中分割手势,并换一种手势进一步验证了所搭建的双目视觉系统从复杂背景中分割手势的可行性。
A convenient and effective binocular vision system is established,which can be used for accurate extraction of gesture information combined with the depth of information.The binocular camera is calibrated by the template calibration method,and the parameters of the camera are obtained accurately,pixel error of calibration is 0.67 pixels.In the stereo matching period,the image matching of the left and right cameras is achieved quickly and accurately by the BM algorithm and the parallax of gestures is obtained.Combine the principle of triangulation,a more dense depth map is produced.Finally,the depth information is remapped to the original color image for three-dimensional reconstruction,and the three-dimensional cloud map is generated.According to the cloud information,it can judge the binocular vision system and effectively split the gestures from the complex background,furthermore,the feasibility of using the binocular vision platform to realize the segmentation of hand gestures in complex background is further verified.
出处
《长江大学学报(自然科学版)》
CAS
2018年第9期40-45,67,共7页
Journal of Yangtze University(Natural Science Edition)
基金
国家自然科学基金项目(51575407)
武汉科技大学2017年国防预研基金项目(GF201705)
关键词
双目视觉
手势识别
手势分割
模板标定法
binocular vision
gesture recognition
gesture segmentation
template calibration