Abstract—A new algorithm for a three-dimensional reconstruction of the surrounding environment is proposed. The algorithm is able to create accurate three-dimensional maps in real-time with the help of a RGB-D depth camera. This algorithm can be used in autonomous mobile robotics, where the robot needs to localize itself in unknown environments by processing onboard sensors without external reference systems such as a global positioning system. We analyze various combinations of common detectors and descriptors of visual features in terms of their recognition efficiency. To match the point clouds between consecutive frames, the iterative closest point (ICP) method is used. To improve the quality of the three-dimensional reconstruction, an adaptive approach to calculating the estimate of the camera position is suggested. The proposed system is able to efficiently process complex scenes at a high rate, and its performance on available benchmark databases is comparable with that of the state-of-the-art systems.

摘要—提出了一种用于周围环境三维重建的新算法。该算法能够借助RGB-D深度相机实时创建准确的三维地图。此算法可用于自主移动机器人，其中机器人需要通过处理板载传感器来定位自己在未知环境中的位置，而无需使用诸如全球定位系统之类的外部参考系统。我们根据其识别效率分析常见检测器和视觉特征描述符的各种组合。为了匹配连续帧之间的点云，使用了迭代最近点（ICP）方法。为了提高三维重建的质量，提出了一种自适应方法来计算摄像机位置的估计值。