Vision-based Location And Mapping For Mobile Robot In Unknown Environment

With the development of technology and the improvement of manufacturing's level,technology of mobile robot develops in an independent and intelligent way.Simultaneous Localization and Mapping(SLAM)makes mobile robots become intelligent and intellectualization.It's the basic of mobile robot's autonomous navigation and environmental exploration.The accuracy of localization and mapping determines whether mobile robots can be applied in the actual scenes.Recently,the advantages of vision sensors make researchers pay great attention to vision-based simultaneous localization and mapping.There are lots of frameworks and application systems which can be used to solve SLAM problems.However,because the process of handling visual information is complex and instable,many difficulties are needed to be solved.Especially for the problems relate to monocular vision-based SLAM in unknown environment.This paper focuses on the issue related to localization and mapping of mobile robot with monocular camera in unknown environment.The research work involves frame line segment features matching algorithm and the design of SLAM algorithm.The key point of the research is monocular visual SLAM issue,which is based on point-line-plane feature fusion of Extended Kalman Filter(EKF).The frame line segment matching algorithm is the foundation to fulfill the monocular vision SLAM algorithm of point-line-plane feature fusion.The innovations and achievements of this paper involve two aspects:Firstly,line segment matching method which is based on multiple geometric constraints and 0-1 programming was proposed in the SLAM feature matching module.Compared with the current line segment matching method,in this paper line segment matching are achieved based on multiple geometric constrains.What's more,0-1 programming is designed to optimize line matching results,which lays the foundation for SLAM which is based on point-line-plane feature fusion.Secondly,SLAM algorithm based on point-line-plane feature fusion is proposed in this paper.Compared with the current SLAM algorithm,this kind of SLAM is proposed by fusing point,line segment and plane features.In this way,scenes can be described specifically.Andmore geometric information can be provided for the localization and navigation of mobile robot.As a result,the accuracy of localization can be improved.