基于宽基线立体视觉的远距离三维重建

Long-range 3D reconstruction based on wide-baseline stereo vision

针对月球车视觉导航中远距离场景的三维重建问题,提出一种基于相机自标定的宽基线立体视觉方法.该方法首先提取立体图对的尺寸不变性(SIFT)特征并进行匹配,得到特征点对应关系;再使用外极线几何约束描述立体图像校正的直射变换,对其Sampson误差使用Levenberg-Marquardt(LM)算法求得最小二乘最优解,完成相机参数估计和图像校正;对校正后的图像使用种子像素视差扩张算法进行宽基线立体匹配,根据匹配得到的视差完成目标场景三维重建.实验表明:该方法解决了月球车相机的现场标定难题,并能够解决宽基线立体视觉面临的光照差异、透视畸变、遮挡等问题,对远距离山脉等自然场景的三维重建效果良好,重建精度较高.

A self-calibration based wide-baseline stereo vision technique was proposed for remote scene＇s 3D reconstruction in vision based navigation of lunar rover. Correspondences between stereo rig were obtained by extracting and matching of scale-invariant feature transform （SIFT） features. Then with the epipolar geometrical description of rectifying collineation,the least-squares solution to Sampson error was achieved by Levenberg-Marquardt（LM） algorithm,which obtained the camera parameter and the rectified stereo rig. Following,the wide-baseline dense matching was completed using seeds＇ disparity expansion algorithm. Finally,target＇s 3D reconstruction was achieved on the calculated disparity. This technique is demonstrated to have good performance on a number of remote natural mountain scenes and solve the problems of field calibration,change in illumination,perspective deformations and occlusions in wide-baseline stereo vision.