基于双目视觉技术的物体深度信息的提取  

The Extraction of Depth Information about the Object Based on Binocular Vision

作　　者：王平[1] 韩燮[1] 韩焱[1] 齐和平 

机构地区：[1]中北大学计算机与控制工程学院 [2]中国兵工集团207所,太原030051

出　　处：《科学技术与工程》2014年第2期56-61,共6页Science Technology and Engineering

基　　金：国家自然科学基金项目(61071193);山西省自然基金(101001011023-1)资助

摘　　要：传统的双目匹配算法得到的数据是左右视图匹配点的二维坐标;但二维坐标将丢失物体的三维深度信息。提出一种利用视差原理在世界坐标系内计算该物点在现实中所对应点的三维坐标的新方法。该方法分为三部分:第一部分为双目相机的标定,包括内参(主点、焦距、径向畸变、不垂直因子)以及两相机之间的外参(旋转矩阵和平移矩阵)。第二部分为特征点的提取与匹配,特征点的提取则是利用SIFT算子提取左右视图的特征点,匹配则是将两视图中的对应点(现实中为同一点)利用SIFT算法进行匹配。第三部分为物体深度信息的计算,利用双目相机的视差原理和已经得出的双目相机的参数进行空间还原得到相应的左右视图投影投射矩阵,结合空间几何线性关系计算出该点在世界坐标系的三维坐标,从而得出其深度信息。The traditional binocular matching algorithm is to obtain the 2-d coordinates of the matching point view, but the 2-d coordinates will lose the 3-d depth information of the object, and the innovation point is to calcu-late the 3-d coordinate of the corresponding points in the world coordinate system mathematically based on the prin-ciple of parallax. The whole measurement process can be divided into three parts ： the first part is for the calibration of binocular camera, the second part is the analysis and matching of feature points, the third part is the measure-ment of object depth information. Calibration is to measure the various parameters of the camera hardware itself, which includes the value of focus, focal length, radial distortion, lateral distortion and the relationship of the loca-tion between two binocular camera lens, the feature-points-matching is to process the two images taken by the bin-ocular camera, and match the feature point in each image （the same point in real world）, then use SIFT matching algorithm to match the corresponding point. The last step is to calculate the final result based on the parallax of the binocular camera principle and the parameters which got in calibration step, after that will obtain the 3-d coordinate of the point in the world coordinate system and get the depth information.

关 键 词：双目相机 标定 特征点 SIKF 匹配 视差 深度计算 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]