Research On Technology Of Vision Sensor Aided High Precision Integrated Navigation

With the rapid development and wide popularization of the digital information,the concept of ‘Digital earth',‘smart earth' or ‘smart city' has been reloaded and becomes common knowledge,and thus higher requests have been put forward on PNT(positioning navigation and timing)service's quantity,quality and renew speed.Traditional navigation techniques could no longer satisfy the increasing demand for LBS(location based service).Integrated with different navigation sensors,adapted to complex scenarios,high-efficiency and robust seamless indoor and outdoor navigation system quickly becomes the new trend of acquiring the real-time geometric information.The research purpose of this dissertation is to achieve seamless and precise indoor and outdoor navigation and positioning.Based on the existing technology of GNSS/INS integrated navigation,efforts are focus on the theory,method and technology of vision sensor aided navigation system.Some key technologies like precise spatial-temporal information registration,high accuracy visual navigation,robust and efficient fusion algorithm and accuracy evaluation of integrated navigation will be studied in this paper.The main work and innovation of this paper are as followed:1.According to the practical application environment and accuracy requirements for indoor and outdoor seamless navigation,a high-accuracy vision sensor has been designed,which can obtain legible images with the aids of Xenon flash light,near-infrared LED light and natural light.Using the near-infrared LED light aided mode can effectively avoid optic pollution,and thus improve the practicability especially for indoor navigation.Meanwhile,the ten-parameter and finite element mixed camera calibration model is used for the precise lens distortion correction.After calibration,the measuring accuracy of the vision sensor can achieve about 1/80000,which can satisfy the needs of high-accuracy navigation and positioning.2.In view of the fact that there are few published literatures on precise registration methods of navigational sensors' mounting parameters,three feasible schemes for spatial information registration between sensors are proposed,namely,indirect measurement assisted method,direct geo-reference assisted method and state estimation assisted method.Through theoretical derivation and experimental verification,the accuracies of these three methods for calibrating the extrinsic parameters,as well as the respective applicable scope,are systematically analyzed and compared.3.High accuracy visual navigation technology based on cooperative targets is studied.In order to deal with the low efficiency problem of high-precision database construction of cooperation features in complex scene,a feature extraction and database construction method based on point cloud and image information fusion is proposed.The bundle adjustment method with additional point cloud observation information is used to further improve the positional accuracy of cooperation features.An adaptive dynamic window search method combining predicted frame information and navigation and positioning information is proposed,which optimizes the real-time processing speed of high-resolution images and improves the efficiency of feature detection.Aiming at the problem of robust camera pose parameter solving,a spatial resection optimization algorithm based on robust ridge estimation is adopted to effectively reduce the interference of measurement gross error and complex observation conditions on visual positioning,and improve the accuracy and reliability of vision localization.4.Hybrid data fusion method for Vision/GNSS/INS integrated system is studied.To cope with the inconsistence problem of sensor update rates,the method of multi-rate measurement information update is employed to implement the effective fusion of satellite/inertial/visual information.This method can adjust the observation information of the integrated navigation system according to the actual situation,and carry out measurement update to obtain optimal sensor data fusion.According to the actual situation,using the redundant image information,the sequential image bundle adjustment and the virtual image bundle adjustment methods are used to improve the post-processing precision of the navigation and positioning system.Besides,Combined with the proposed real-time positioning and post-processing optimization algorithm,the corresponding real-time vision processing module and post-processing vision processing module are designed.Distributed system provides a practical reference to the processing procedure of real-time and post-processing module.Firstly,the real-time module can store the original data and real-time processed information.Then,the real-time module sends the processed information from client to server.The server deals with the processed information using the proposed post-processing algorithm,improving the whole positioning accuracy of navigation system.5.Performance evaluation of integrated navigation system is discussed.Considering multiple disturbance sources and aggressive vehicle motions in the outdoor scenario,an outdoor dynamic positioning accuracy detection system is designed with hybrid use of total station,laser tracker and GNSS timing device.The overall performance of the detection system is improved by means of the statistic of single measurement time,smooth curve fitting,and ICP trajectory correction.For indoor scenarios,the spatial reference transmission process is complicated and severe occlusion may exist.Based on the optimization of spatial reference transmission,an indoor dynamic positioning accuracy evaluation method using OptiTrack dynamic capture system is proposed.Both indoor and outdoor experiments are carried out to verify the proposed evaluation method,in which the overall trajectory comparison and real-time comparison are applied for testing the dynamic position accuracy of navigation system.