Research On Indoor Fusion Positioning Technology Based On Object Detection And Geographic Matching

With the rapid development of social modernization,large-scale population-intensive buildings in cities are becoming more and more popular in cities,and the demand for indoor location services has also increased significantly.Its significance is also extraordinary.In particular,the earthquakes,fires and other emergencies that occur each year have a huge impact on large and densely-populated buildings.Accurately understanding the location information of people in large buildings is the top priority of emergency rescue and making a rescue planning.At the same time,knowing exactly where you are is crucial for people who are trapped in a large building to carry out self-rescue activities.Therefore,achieving accurate and convenient indoor positioning has increasingly become a new challenge for public safety.At present,most of the current indoor positioning systems distribute a large number of positioning beacons indoors to establish a geographic information fingerprint database,thereby obtaining relatively high positioning accuracy.The current indoor positioning solution requires the deployment of a large number of positioning reference beacons at the beginning of construction,and lacks robustness when dealing with complex channel transmission environments.Different from the traditional indoor positioning methods,the positioning scheme proposed in this thesis considers the realization of indoor positioning from the perspective of geographic information assisted positioning.The information that can be used as auxiliary positioning in the indoor environment is very rich.With the enhancement of the computing power of the mobile terminal,through the use of mobile terminal data collection and semantic extraction technology,the mobile terminal understands the environmental information,and the idea of completing auxiliary positioning is gradually feasible.In this thesis,the topological relationship between objects can be constructed by detecting the types of objects appearing on the current mobile terminal and measuring the distance between the current mobile terminal and the object.The positioning model obtains a visual semantic positioning,and then uses the extended Kalman filter algorithm to fuse the visual odometer or pedestrian track estimation with the visual semantic positioning to obtain a more accurate positioning.The main content and innovation of this article includes the following aspects:1.The current indoor positioning method for object detection has been improved.By introducing building information modeling technology,a topology matching method based on building information modeling is proposed,which solves the problems of the current object detection and positioning method that is difficult to obtain in the data source and difficult to match the real environment with the electronic map.2.An indoor fusion positioning method based on pedestrian path estimation and object detection geographic matching is proposed,which solves the situation where the error is too large when the object topology is scarce based on the object detection and geographic matching positioning method.3.An indoor visual fusion positioning algorithm based on geographic matching of visual odometer and object detection is proposed to solve the situation where the error is too large when the object topology is scarce based on the object detection and geographic matching positioning method.In the future,as the computing power of smart mobile terminals becomes more powerful,the requirements for indoor positioning technology will become higher and higher,and the application of indoor positioning technology based on high precision will also increase.Multi-source information fusion positioning will further assist indoor positioning technology to further improve positioning accuracy,robustness and usability,making indoor positioning technology can make indoor life more convenient and intelligent.