| Simultaneous Localization And Mapping(SLAM)is the basic technology for robots to perform navigation obstacle avoidance tasks.It can use sensor data to establish a world coordinate system and obtain real-time pose estimation of robot motion in the world coordinate system and create a scene map.The main work of the thesis is as follows:Firstly,the basic algorithm flow and mathematical principle of the tight coupling combined positioning system based on monocular camera and inertial measurement unit are studied.The derivation process and physical meaning of the sensor measurement model and residual model are given.According to the mathematical model and pose estimation of the two sensors,the optimization objective function is established,and the optimized system pose estimation and accelerometer bias and gyroscope deviation are obtained.Secondly,the basic principle of monocular dense mapping algorithm is studied.The process of selecting pixel points through quadtree structure is given.The core formulas and the physical meanings of the formula are given.According to the research of algorithm principle,the image processing module is added,and the VIOSLAM based on monocular-inertial sensor combined positioning and mapping system is designed.The real-time positioning of the system and the construction of the octree occupation map are completed.Thirdly,this paper evaluates a variety of combined positioning algorithms and improved algorithm VIO-SLAM under the Euroc data set,gives the results of algorithm evaluation and data set mapping.The evaluation results show that the improved algorithm has improved in positioning accuracy and mapping density.Finally,the experiment under real scene is carried out by mynteye-d camera.The running track and scene mapping result of the combined positioning system are given.The experimental conditions and results are analyzed.The experiment shows that VIO-SLAM has better performance than the original algorithm. |
PDF Full Text Request