The Research On Indoor Topology Mining Technology Based On Crowdsourcing

In recent years,with the rapid development of Internet technology,location-based services are welcomed by the general public,which has promoted a large number of indoor positioning systems.The core of location-based services is high-precision localization technology.Location technology includes indoor localization technology and outdoor localization technology.Indoor maps are a prerequisite of indoor localization technology.Due to the completion time of building,building ownership,building management and other reasons,indoor maps is not easy to obtain.Efficient and accurate indoor map construction technology has become the key to develop indoor localization technology.Topology map is a representation of indoor map.It represents the indoor environment as a topology diagram of nodes and edges.The nodes represent the positions of corners,doors,elevators,stairs and so on,while the edges represent the connections between nodes.Crowdsourcing is a prevailing Internet concept in recent years,which means to distribute tasks to many people in a free and voluntary way.Traditional indoor map construction method requires professionals to conduct on-site measurements,which not only consumes a lot of human and material resources,but also errors occur easily.In order to construct an indoor map efficiently and effortlessly,we propose an algorithm for the automatic construction of indoor map-An indoor topology-mining algorithm based on crowdsourcing.We use smartphones as the hardware and adopt crowdsourcing to obtain the indoor trajectory data efficiently.The loop detection method is used to reduce the error caused by the traditional trajectory coordinate estimation algorithm.In order to simplify research subject,we use turning detection technique to cut long trajectories into atomic trajectories.To achieve the purpose of separate trajectory data from different paths,we use a hierarchical clustering model based on directional sequence clustering and geomagnetic sequence clustering.This model involves affinity propagation clustering algorithm and dynamic time warping algorithm.In order to eliminate the inherent tracking errors originating from noisy sensors,we use trajectory fusion strategy based on DTW algorithm to fuse the trajectory data in the same cluster.Finally,we use topology calibration method based on the building trend to further improve the indoor topology mining accuracy.We design and realize indoor topology mining system based on above theory and make a performance evaluation.One advantage of our system is that additional hardware such as a laser rangefinder or wheel encoder are not required.Experimental result demonstrates that our method constructs indoor floor plan successfully with 1.7m accuracy,which greatly reduces the cost of mapping and benefits location-based services.