Research On Vision/IMU/Laser Multi Sensor Integrated Navigation And Positioning

With the continuous pursuit of intelligent and convenient life and the rapid development of navigation and positioning technology,indoor navigation and positioning technology gradually improves and serves people's life.However,in some indoor environments,it is difficult to achieve accurate positioning effect only by satellite positioning technology.Therefore,the research of the method of high precision multi-sensor integrated navigation has a high research significance.This paper studies the integration methods of inertial measurement unit(IMU),visual camera and lidar,and realizes the related system through design.The main research content of this paper is divided into two aspects: one is to use absolute orientation algorithm to process visual image,and then integrate the visual processing results with IMU to improve the positioning accuracy;on the other hand,it improves the feature extraction and registration algorithm of laser point cloud,and then designs a visual / IMU / laser multi-sensor integrated navigation and positioning system based on federal filter.The main research contents of this project are summarized as follows:Firstly,an image analysis and location method based on absolute orientation algorithm is proposed.Then,Kalman filter is used to integrate visual and IMU data to realize the vision inertial fusion navigation and positioning.Through the design experiment analysis,the vision inertial fusion scheme has a high precision,and the average error is 39.45% lower than that of pure vision method.Secondly,a feature extraction method based on voxel mesh is proposed.The experimental results show that the improved method has better point cloud feature extraction effect.On this basis,the improved feature extraction method is combined with the immediate close point(ICP),and then the fine-tuning point cloud data is accurately registered.The experiment shows that the method has a good point cloud registration efficiency,and the average registration time is about 81% shorter than that of the traditional ICP registration algorithm.Finally,the design experiment compares and selects DSO visual odometer with better effect as the front-end of vision.The fusion of vision /imu/ laser is carried out by federal filter method.The design experiment successfully achieves high precision positioning effect,with the average error reduced to 0.17 meters compared with single positioning method.