面向巡视探测任务的复杂地形信息感知与场景重建

Complex Terrain Information Perception and Scene Modeling for Patrol Detection Tasks

针对复杂地形环境下巡视探测中避障问题,提出了一种基于点云的地形信息感知与场景建模方法。首先对获取的点云数据进行稀疏采样和滤波降噪;然后结合移动机器人越障能力极限与改进的随机采样一致性算法,拟合其自适应基准面作为可通行区域;其后使用基于密度的聚类算法感知地形特征信息,并采用凸包算法提取地形特征轮廓;最后结合自适应基准平面进行快速三维场景重建,为地面观测提供直观快速的巡视器周围三维环境模型。通过对复杂地形环境进行模拟实验,结果表明:该方法可以有效获取复杂地形信息,并可大幅度提高场景重建的效率。

Aiming at the obstacle avoidance problem in patrol detection in complex terrain environment,a point cloud-based terrain information perception and scene modeling method was proposed.First,the sparse sampling and filtering of the acquired point cloud data were performed to reduce noise.Then the mobile robot’s obstacle crossing ability limit was combined with an improved Random Sample Consensus algorithm(RANSAC)to fit its adaptive datum as a passable area.The Density-Based Spatial Clustering of Applications with Noise algorithm(DBSCAN)was used to perceive terrain feature information,and convex hull algorithm was used to extract terrain feature contours.Finally,the adaptive reference plane was used for fast 3D scene modeling to provide an intuitive and fast 3D patrol surrounding environment model for ground observation.Experiments were conducted in the simulated complex terrain environment and the results showed that complex terrain information could be obtained effectively by this method and the time efficiency of scene modeling could be greatly improved.