Research On An Efficient Simultaneous Localization And Map Extension Algorithm Based On EPnP Algorithm

Vision SLAM refers to utilize the visual sensor carried by robot for building the scene around robot and locating robot's pose.Due to it was fast and accurate,laser radar was applied to early SLAM systems as the primary sensor for obtaining depth information.Yet now the camera was subsequently gaining much attention in the field of SLAM owing to its low cost,accessibility,information richness and light weight.Most visual SLAM systems can currently generate more accurate map information,but lack of effective storage and reusable map information.Each operation of SLAM system is independent of each other and lack of effective contact.Therefor more repetitive work is be done and it hinders the development of SLAM technology.This thesis studies the visual SLAM algorithm based on the saved map information and realizes the map relocation and map extension under outdoor environment,and it has positive research significance and application value.For monocular SLAM algorithm in outdoor environment,feature-based method is applied in this thesis for algorithm research.In order to improve the running speed of the algorithm and speed up the matching time,mesh is used to the saved map.Meanwhile,the feature index of the map information and the mesh is established.In order to achieve scale consistency between the current SLAM system map and the saved map,a scale-consistent initialized map is constructed based on the EPnP algorithm and the map relocation algorithm.A saved map is eventually expanded by map expansion algorithm.In order to verify the practical effect of this theoretical study,an experimental system based on monocular camera,monocular SLAM system and the saved environment map information was constructed.The map relocation and map expansion algorithm experimentation were tested in outdoor in turn.The actual experimental results show that the outdoor map relocation scheme designed in this thesis has a good effect,can accurately locate the camera.Meanwhile,map expansion algorithm is realized by map relocation algorithm,which can be used to construct continually sparse map of environment and shows the practical applications of the saved map in the visual SLAM.