基于单目视频路径与激光雷达的移动机器人导航方法研究

发布时间：2018-04-30 23:08

本文选题：移动机器人 + 视频路径　；参考：《郑州大学》2017年硕士论文

【摘要】：移动机器人导航是机器人领域的一个研究热点。视觉传感器作为最接近人类的环境感知工具,具有其它传感器所无法比拟的优势。本文针对移动机器人视觉导航进行了研究,考虑到机器人在导航上的成本、性价比与实时性,以单目视觉中双视图定位为基础,提出了一种基于单目视频路径的导航方法,并结合2D激光雷达进行拓展,实现在未知环境下的避障。本文完成的导航工作主要包括:视觉定位、路径规划、自主避障三个方面。第一部分,视觉定位。本文以单目视觉中双视图定位为基础,引入参考路径的视频序列,从中提取有序的关键图像。使用Harris角点算法提取图像特征点,利用零均值归一化互相关算法进行图像匹配,根据当前图像和最近似关键图像,匹配点之间的约束关系求出基础矩阵并优化,得到它们的位姿参数,进而完成移动机器人定位。第二部分,路径规划。移动机器人在视频路径的指导下,判断当前图像与关键图像共有的特征点集的形心横坐标位置,以形心横坐标为参照,进行视觉路径跟踪,使移动机器人沿参考路径行走,实现机器人的路径规划。第三部分,自主避障。本文采用激光雷达测距,基于栅格地图进行避障控制建模,通过定义避障风险函数、碰撞函数、运动速度、移动曲率等一系列参数,将避障过程化繁为简,分为三个阶段实施,实现移动机器人在上述导航过程中的自主避障。最后,在真实的环境下,基于自主研发的科研平台,对所提出的导航方法进行了验证,证明该方法在室外环境中具有较好的实用性。
[Abstract]:Mobile robot navigation is a research hotspot in robot field. As the closest environmental sensing tool, visual sensor has the advantage that other sensors can't compare. In this paper, the vision navigation of mobile robot is studied. Considering the cost, cost performance and real time of robot navigation, a navigation method based on monocular video path is proposed based on biview positioning in monocular vision. Combined with 2D lidar, the obstacle avoidance is realized in unknown environment. The navigation work accomplished in this paper mainly includes three aspects: visual positioning, path planning and autonomous obstacle avoidance. The first part, visual positioning. Based on the biview localization in monocular vision, this paper introduces the video sequence of reference path to extract the ordered key image. Harris corner algorithm is used to extract image feature points, and zero-mean normalized cross-correlation algorithm is used to match the image. According to the constraint relation between the current image and the most recent key image, the basic matrix is obtained and optimized. The position and pose parameters are obtained, and then the mobile robot positioning is completed. The second part, path planning. Under the guidance of the video path, the mobile robot judges the position of the centroid coordinate of the feature point set shared by the current image and the key image, and makes the mobile robot walk along the reference path by using the centroid and transverse coordinate as the reference, so that the mobile robot can walk along the reference path. The robot path planning is realized. The third part, autonomous obstacle avoidance. In this paper, the model of obstacle avoidance control based on grid map is established by using lidar ranging. By defining a series of parameters, such as obstacle avoidance risk function, collision function, motion velocity and moving curvature, the process of obstacle avoidance is simplified and divided into three stages. The autonomous obstacle avoidance of the mobile robot in the above navigation process is realized. Finally, in the real environment, based on the independent research platform, the proposed navigation method is verified, and it is proved that the method has good practicability in outdoor environment.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP242

【相似文献】