Research On Visual Positioning Technology Under Low Illumination Environment

With the rapid development of the information age and the continuous advancement of science and technology,people's demand for location services of positioning and navigation has become greater and greater in recent years.The Global Positioning System(GPS)and Bei Dou satellite navigation system can provide more accurate positioning services in the outdoor environment.The signal quality received by the GNSS receiver is poor when the service target is located in an indoor environment,resulting in a significant reduction in its positioning accuracy and even failure to provide positioning service.In order to achieve positioning services in complex indoor environments,various indoor positioning technologies,especially visual positioning technologies based on camera sensors,have been greatly developed.This paper mainly aims at solving the problem that the visual positioning technology is inaccurate and accumulative error in low-light environment.First of all,this paper briefly introduces the system structures of visual odometers and visual inertial odometers commonly used in visual positioning technology.Then we investigated domestic and foreign vision positioning processing solutions in lowlight environments,and how to deal with the problem of cumulative error of visual positioning systems.Based on the understanding of the above background knowledge,we analyzed in detail the problems of visual observation deficiency in visual positioning system under low-light environment,such as insufficient number of feature points and uneven distribution of feature points.We analyzed and modeled each visual observation situation,and then summarizes several observation situations affected by low-light environment.In order to solve the problem that affects visual observation in low-light environments,we proposed an algorithm to improve the accuracy of feature matching that joint matching of features in feature regions based on texture information.This algorithm can effectively improve the stability and accuracy of visual observations and positioning accuracy of positioning system in low-light environments.In response to the problem of traj ectory drift caused by cumulative errors in long-term operation of the visual positioning system,we established an offline image landmark library to eliminate cumulative errors in the system positioning.which ensure that the positioning system can output global consistent positioning results for a long time.The solution proposed in this paper provides a positive boost to the development of indoor positioning services for a long time in complex environments.Finally,experiments were designed to analyzed and tested the performance of the improved algorithm and the landmark database proposed in this paper.These experiments are mainly divided into two parts.The first part uses public data-sets to evaluate the accuracy of the visual positioning system after adding the improved algorithm.The experimental results proved that our algorithm can improve the positioning accuracy of the visual positioning system under low light environment.The second part of the experiment is to verify the feasibility and effectiveness of the image landmark database.It is proved that the image landmark database can effectively eliminate the accumulated errors under long-term operation and improve the overall accuracy of the visual positioning system.The visual positioning system designed in this paper can improve the robustness and accuracy of indoor positioning in low-light environments,and can provide good location services for complex indoor environments such as large shopping malls and underground garages.