无人车基于双目视觉的同时定位与地图构建

Simultaneous localization and mapping for UGVs with binocular camera

研究了无人驾驶车面向校园环境的同时定位与地图构建问题.采用双目视觉系统进行立体视觉图像匹配,并以此为基础完成优化前的位姿拓扑地图构建;采用了一种基于ORB图像特征和BoW模型的闭环检测算法,并利用时间连续性约束和几何一致性约束来提升闭环匹配正确率.位姿拓扑地图的后端优化采用了高斯-牛顿优化方法,并且在迭代过程中充分考虑了系统信息矩阵的稀疏性.利用实验室自主研发的SmartCruiser无人驾驶车平台在校园环境进行了实验,结果验证了本文所提方法的有效性和实用性.

Simultaneous localization and mapping (SLAM)problem for unmanned ground vehicles (UGVs)in campus environments was investigated.A stereo image matching algorithm was deployed to perform consistent pose estimation so that an initial pose-graph model was constructed.A loop clo-sure detection algorithm based on oriented FAST and Rotated BRIEF (ORB)feature matching and bag-of-words (BOW)model was utilized in our work,which can provide the constraints of temporal consistency and of geometrical consistency to improve the accuracy of loop closure.The back-end im-plementation for graph-based SLAM was used the Gaussian-Newton optimization method and the sparse characteristics of the system information matrix was fully utilized in the iterative procedure. The experiments were conducted on our self-developed UGV in DUT campus,and the results show the validity and robust performance of the proposed approach.