Abstract: A multi-information system based visual system with binocular and monocular was designed for environment detection,dynamic target tracking and path planning of mobile robot in indoors navigation.To effectively extract obstacle information and obtain available paths,a self-adaptive single scale Retinex algorithm incorporated color saturation information was proposed for indoors shadow elimination.A spatial-temporal context tracking algorithm was studied to employ target's key feature points to determine its rough location.Based on the rate of change of key feature points,the model was updated automatically to effectively improve the accuracy of dynamic target location in complex environment.The function of attraction velocity,repulsive velocity and tangential velocity in traditional velocity vector fields were improved which solved the problems of robot track jitter,concussion near the target and unreachable target when in the repulsive field of obstacles.The autonomous navigation experiment of mobile robot in indoor environment realized the features extraction and real-time positioning of obstacles,robots and targets,and achieved the effective tracking of dynamic targets in the shortest collision avoidance path.

Key words: mobile robot, autonomous navigation, multi-information system, environment detection, target tracking, path planning

摘要： 针对移动机器人室内自主导航中环境检测、动态目标定位跟踪和路径规划,设计了“双目+单目”多信息融合的视觉系统。提出一种自适应环境、融合色彩饱和度信息的单尺度Retinex室内阴影消除算法,以有效提取障碍物信息及可行路面区域。研究了一种特征点辅助的时空上下文目标跟踪算法,运用目标特征点进行目标粗定位;引入特征点变化率信息,自动调节时空上下文模型更新,以有效提高复杂环境下动态目标定位的准确性。改进了传统速度向量场吸引速度、排斥速度和切向速度函数,解决了路径规划中机器人轨迹抖动、目标点附近震荡和目标处于障碍物排斥场不可到达等问题。移动机器人室内环境下自主导航实验实现了障碍物、机器人和目标特征提取及其实时定位,移动机器人以最短的避碰路径完成动态目标的有效跟踪。

关键词: 移动机器人, 自主导航, 多信息融合系统, 环境检测, 目标跟踪, 路径规划