Vision Simultaneous Localization And Mapping Based Two Demision Code On The Ceiling

With the development of the robot technology, more and more researchers focus on the Simultaneous Localization and Mapping(SLAM) which is one of the most important points of service robot. At present, the researchers in this field mainly focuses on feature extraction and feature matching, closed-loop detection and optimization. On the basis of previous theoretical research, this thesis designs a ceiling vision Simultaneous Localization and Mapping system based on two-dimensional code and selects the steady and robust two-dimensional code as the landmark to describe a single symmetrical factory environment. It simplifies the camera model and reduces the interference of the dynamic environment.In this thesis, the fusion of inertial information and visual information is used in mobile robot Simultaneous Localization and Mapping. Through the analysis of the platform, this thesis designs a differential velocity motion model and a pinhole camera model in the system framework of the traditional EKF-SLAM. In consideration of the fact that the fixed observed variance c an’t reflect the characteristics of the actual camera model, the adjustment factor is introduced to describe the camera model, which makes the observation model more accurate. At the same time, this thesis uses the distance message of multi-codes to filter the state message of single two-dimensional code. Taking into account the factor that micro fluctuation of the robot will introduce the transformation of scale, and affects the accuracy of the robot’s ceiling observation model, this thesis designs a strategy of the fluctuation detection and fluctuation compensation.In order to reduce the influence of environmental illumination on the information extraction of two dimensional code, this thesis designs a light compensation and multi threshold optimization method for the frame which can’t identify the two-dimensional code by ARTool Kit Plus. This method can effectively improve the accuracy of extraction. Taking into account the time consumption of searching the two-dimensional code from the image, this thesis designs the active search for the two-dimensional code, which effectively reduces the time of image processing and improves the operating speed of the system. Finally, the effectiveness of the algorithm is verified by the experiments in actual environment and Gazebo simulation environment.