RGB-D Based Visual SLAM Algorithm Extraction And Its Implementation

The key to achieve autonomous control of robot,and also an important topic in mobile robotics field is Simultaneous localization and mapping(SLAM).A mobile robot has one big task which is to assign itself and building a map of its particular environment at the same time.Especially,a SLAM process in which the sensors are cameras is called Visual Simultaneous Localization and Mapping(vSLAM),the inputs of the systems are visual information.A visual SLAM algorithms which has as input a regular color(RGB)image and its depth obtained by using of a low-cost RGB-D sensor such as Microsoft Kinect are introduced in this thesis.The aim of RGB-D based Visual SLAM is to build a 3D map of the environment at the same time keep tracking of the camera pose.The aim of this work is to implement the RGB-D based Visual SLAM algorithms.The followed system is composed by the front end and the back end.The objective of the front end is to process RGB-D data,in this part we use the algorithms for features detection and descriptors extraction,the descriptors results are used to implement the features matching.After that,we estimate the motion transformation and optimize from matching results.Finally,the visual SLAM is implemented in the back end of the system,the pose graph optimization method is used,to obtain Camera pose and 3D map environment reconstruction.