SLAM is one of the most important components in robot navigation. A SLAM algorithm based on image sequences captured by a single digital camera is proposed in this paper. By this algorithm, SIFT feature points are selected and matched between image pairs sequentially. After three images have been captured, the environment’s 3D map and the camera’s positions are initialized based on matched feature points and intrinsic parameters of the camera. A robust method is applied to estimate the position and orientation of the camera in the forthcoming images. Finally, a robust adaptive bundle adjustment algorithm is adopted to optimize the environment’s 3D map and the camera’s positions simultaneously. Results of quantitative and qualitative experiments show that our algorithm can reconstruct the environment and localize the camera accurately and efficiently.
将基于因子分解的运动估计结构(structure from motion,SFM)算法延伸至室外环境障碍物检测,提出了一种基于单相机的障碍物检测方法.通过图像序列特征点的匹配和跟踪,运用基于因子分解的运动估计结构算法得到场景的投影重建;通过满足绝对二次曲面(dual absolute quadric,DAQ)约束的自标定升级至欧式重建,同时得到相机的运动;通过将图像分割为等面积的区域,每个独立的区域通过从欧氏重建得到的深度信息来区分是障碍物还是背景.室外真实场景的实验结果表明,该方法能够在室外环境下获得比较好的障碍物检测效果.
