采用V视差方法对视差图进行预处理，以获取地形的有效重构区域，并避免重构不可通行区域；通过融合惯性测量单元与里程计的数据来更新高程栅格图，并结合中值滤波和邻域栅格滤波来平滑高程栅格图，实现了3维地形高程栅格图的快速构建.结果表明，所用方法具有较好的实时性和准确性，仅需对有效重构区进行地形重构，重构效率较高.
Automatic vehicle must perceive the off road 3D environment when it is moving on the off road. The terrain can be rebuilt by stereo vision, but it may get poor efficiency and accuracy if it is rebuilt directly. This paper uses V-disparity to preprocess the disparity image to get effective reconstruction regions and remove the non accessible area from disparity image. The inertial measurement unit (IMU) and speedometer are fused to update the elevation grid map, which is smoothed by using medium filter and areas grid filter. Thus, a 3D terrain elevation grid map can be rebuilt rapidly. The method could increase significantly the efficiency and accuracy of reconstruction since it only rebuilds the effective reconstruction region. Test results show that this method has good real-time performance and high accuracy.
［1］韩光, 孙宁, 李晓飞, 等. 非结构环境理解综述［J］. 计算机应用研究, 2014, 31(8): 2248-2253.
HAN Guang, SUN Ning, LI Xiaofei, et al. Unstructured scene interpretation: A review［J］. Application Research of Computers, 2014, 31(8): 2248-2253.
［2］GLVEZ A, IGLESIAS A. Particle swarm optimization for non-uniform rational B-spline surface reconstruction from clouds of 3D data points［J］. Information Sciences, 2012, 192(6): 174-192.
［3］KAZHDAN M, HUGUES H. Screened poisson surface reconstruction［J］. ACM Transactions on Graphics (TOG), 2013, 32(3): 1-13.
［4］SINGH M K, VENKATESHK S, DUTTA A. Kernel based method for surface estimation using laser scanner data［C］∥International Conference on Communications and Signal Processing (ICCSP). Melmaruvathur, India: IEEE, 2015.
［5］KO B C, JUNG J H, NAM J Y. Fire detection and 3D surface reconstruction based on stereoscopic pictures and probabilistic fuzzy logic［J］. Fire Safety Journal, 2014, 68(8): 61-70.
［6］KWEON I, KANADE T. High resolution terrain map from multiple sensor data［C］∥Proceedings of the IEEE International Workshop on Intelligent Robots and Systems (IROS’90). Washington, USA: IEEE, 1990: 127-134.
［7］BELTER D, ABECKI P, SKRZYPCZYNSKI P. Adaptive motion planning for autonomous rough terrain traversal with a walking robot［J］. Journal of Field Robotics, 2016, 33(3): 337-370.
［8］FANKHAUSER P, BLOESCH M, GEHRING C, et al. Robot-centric elevation mapping with uncertainty estimates［C］∥Climbing and Walking Robots, 2014: 433-440.
［9］SOUZA A, MAIA R S, AROCA R V, et al. Probabilistic robotic grid mapping based on occupancy and elevation information［C］∥16th International Conference on Advanced Robotics (ICAR). Montevideo, Uruguay: IEEE, 2013: 1-6.
［10］SOUZA A, GONCALVES L M G. Occupancy-elevation grid: An alternative approach for robotic mapping and navigation［J］. Robotica, 2016, 34(11): 2592-2609.
［11］FANKHAUSER P, HUTTER M. A universal grid map library: Implementation and use case for rough terrain navigation［M］∥Robot Operating System. Springer International Publishing, 2016: 99-120.
［12］BERNINI N, BERTOZZIM, CASTANGIA L, et al. Real-time obstacle detection using stereo vision for autonomous ground vehicles: A survey［C］∥17th International IEEE Conference on Intelligent Transportation Systems (ITSC). Qingdao, China: IEEE, 2014: 873-878.
