The Research And Implementation Of Indoor Robot Monocular Vision Simultaneous Localization And Mapping Technology

The key of indoor robot autonomous mobile is able to realize self-position in the unknown indoor environment, while the premise and foundation of robot self-localization are to determine its own position build the surrounding environment map. It is easy to implement the location problem with the known environment map and the mapping problem with the known robot pose. It has been widely studied. In an unknown environment, as the environment map and the robot pose cannot be determined, the robot needs to perceive the environment information according to sensors taken with itself, and determines its own pose according to the perceived environment information. Therefore, the simultaneous localization and mapping of robot is closely related to the type of sensor. As visual sensors not only are able to obtain intuitionistic and rich environment information but also have the advantages of low cost, they have been widely studied and applied in the robot simultaneous localization and mapping algorithm.This article designs a monocular vision robot simultaneous localization and mapping system, which uses a monocular camera and odometer to perceive environment information, and carries out the following research work around the system:(1) Researching on visual SLAM, and designing a system that meets the paper’s needs. The system is divided into three modules, which is analyzed and researched respectively.(2) Studying on the extraction technology of visual image features on the basis of the SLAM system framework designed by the paper. Focusing on the SIFT algorithm analysis, and verifying some characteristics of SIFT feature points by experiments. According to the three-dimensional geometry imaging theory, this article solves the rotation and translation vector of the camera combined with epipolar geometry. Based on the matched points, the paper finds position information for spatial point corresponding to the feature points.(3) Establishing a state model and observation model of extended Kalman filter, and doing the simulation experiments about the robot monocular visual simultaneous localization and mapping system designed by the paper in the MATLAB environment. Designing rail experiments in order to simulate the robot to realize the real SLAM.In the process of researching indoor robot monocular camera SLAM, the paper proposes two innovative ideas according to the application environment and the characteristics of indoor robot.(1) Putting forward an ideology that the SIFT feature points is the obstacle feature points, and dividing the indoor area into obstacles and the walking area, which solves the difficulty of how to determine the unable walking obstacle area in the problem of indoor robot monocular camera SLAM.(2) Proposing an idea that using downgrade SLAM ideology to solve the problem of ordinary indoor robot kidnapping, and carring out the experiment. The experimental results show that "downgrade SLAM" can effectively solve the indoor robot kidnapping problem in the case of monocular camera.The experimental results show that this research and implementation of SLAM system meets the design requirements that it accurately implements the robot pose’s determination while environment map building.