一种基于手绘地图的动态环境视觉导航方法

A Visual Navigation Method Using a Hand-Drawn-Route-Map in Dynamic Environments

提出了一种基于手绘地图和路径的移动机器人视觉导航方法.首先,根据较小偏差的原则提取运行路径中的关键引导点,以便将原始路径分成多段.然后,移动机器人在各段运行过程中,对预先绘制的手绘地图中对应的参考图像以及机器人摄像头实时采集到的图像信息进行匹配.这里提出预测估计的方法估计当前视野中最可能存在的图像,以加速图像的匹配过程,并利用SURF(speeded up robust feature)算法检测图像的特征,依靠KD-Tree方法快速求得匹配点,采用RANSAC(RANdom SAmple Consensus)算法求解参考图像与实时图像的投影变换矩阵H,进而得到参考图像在实时图像中的位置,并融合里程计数据,得到机器人的大致位置.再后,根据获得的机器人大致位置,计算下一段的运行方向,继续下一段运行.依此类推,直至运动到最后一段.最后,通过一系列的实验,验证了机器人在本文方法下不需要精确环境地图及精确运行路径就能顺利导航,并能实时有效地避开动态障碍物.

A visual navigation method of mobile robot based on hand-drawn route map and route is proposed.At first,some key boot points are picked up from the running path according to the principle of small deviation,so that the original path is divided into several segments.Then,during the mobile robot running along every segment,real-time image information from robot camera is matched with the corresponding one in the prior hand-drawn-route-map.In order to speeded up image processing,a kind of prediction estimation method is proposed to find the most potential image in the current field of vision. SURF（speed up robust feature） algorithm is used to detect image features.Matching points can be found rapidly in terms of KD-Tree.The projection transform matrix H between the referenced image and the real-time one is solved by RANSAC （RANdom SAmple Consensus） algorithm,in order to know the location of the referenced image in real-time one.With reference to odometer and real-time image information,the robot can be roughly localized.And then,for the next segment, the robot＇s running direction is computed according to its current rough localization until the last segment.At last,through a series of experiments,the advantage and efficiency of the new method in navigation and real-time dynamic obstacle avoidance are testified with the imprecise real map and route.