| Mobile robots are widely used in industry,service,logistics and other fields because of automation and intelligence.The key to the automation and intelligence of mobile robots is the localization technology.How to realize the high precision,high efficiency and good robustness of localization algorithm is the goal that domestic research is pursuing at present.The topic of this paper comes from the major science and technology project of Anhui Province,"Development of Heavy Duty Casting Robot under Complicated Operating Environment"(the project number is 16030901012).In this paper,the robot positioning technology is studied,mainly including an improved binocular vision distance calculation method and a VI localization algorithm,and the performance of the algorithm is confirmed by experiments.The main research contents of this artical can be summarized as follows:(1)For the problem of the feature point extraction and matching with large amount of computation,combined with the advantages of optical flow way to build binocular vision odometer that based on the hybrid tracking of optical flow and feature point way.Firstly,the camera model was recommended,and the binocular camera was calibrated by Kalibr toolbox to obtain the camera internal parameters..Then,according to the basic principle of the feature points and optical flow tracking,the difficulty of optical flow tracking was quantified,the appropriate threshold was set,and the polar constraint detection was performed.Based on this,the ORB-SLAM2 front-end visual oemometer was improved by switching the tracking method.Finally,the improved visual odometer and ORB-SLAM2 front-end visual odometer is test and compared on the KITTI dataset,and the positioning accuracy and operating efficiency of the improved binocular visual odometer are analyzed.(2)Aiming at the problem that the visual odometer is greatly affected by illumination and speed,a local pose optimization method based on sliding window was adopted to build a VI localization system by tightly coupling binocular camera and IMU information.Firstly,the essential framework of the VI localization system is described,and the IMU model are obtained.The orientation obtained from the visual initialization is taken as the initial value,and the IMU gyroscope deviation,accelerometer deviation and initial velocity are calculated to complete the joint initialization.Secondly,a tightly coupled sliding window visual inertial navigation fusion optimization model was established,the system state variables and objective functions were delineated,and the error terms were constructed.The multi-constraint and multi-state optimization model was solved.Finally,IMU was calibrated by IMU_utils,and the calibration results were analyzed by Allan analysis method.(3)The experiments that verify the positioning accuracy and real-time performance of the Ⅵ localization system was designed.Firstly,the localization performance of the Ⅵ localization system and VINS-Mono was tested and compared on the EUROC dataset,and the positioning accuracy of the VI localization system was evaluated.Then,using Xiaomi binocular inertial navigation camera as sensor,the camera-IMU joint calibration was performed.Finally,the experiment was carried out on an omnidirectional four-wheeled mobile vehicle.Taking CPU occupancy as the index,the positioning trajectory was analyzed,and the real-time performance and application performance of the visual inertial navigation fusion positioning system presented in this paper were verified.The relevant research in this paper is helpful for the follow-up research on mobile robot positioning technology,including visual positioning and visual inertial navigation fusion positioning technology,which has important theoretical and practical significance in the fields of robots,intelligent vehicles,unmanned driving and so on.Figure[52]Table[7]Reference[70]... |
PDF Full Text Request