Study On Clustering Method For Indoor Moving Trajectory Based On RFID

In recent years,with the indoor positioning technology development of RFID,WiFi,bluetooth,etc,collecting indoor moving objects' trajectory has become possible.The survey suggests people spend almost 87% of their time in indoor activities.Indoor application service requirement has become increasingly rich,such as interior navigation,navigation and hot pat recommending.These applications need to analysis the data to get the potential information.Indoor trajectory clustering is one of the key technology of mobile trajectory analysis,this technology can identify the mobile mode of indoor object,location preferences,hot path,etc.But because of the differences in spatial positioning technology,constraint,and distance measurement between indoor and outdoor environment,the existing majority of trajectory clustering analysis does not apply to complex indoor space,can't adapt to a variety of indoor space,and at the same time the existing technology lacks of semantic description of complex indoor trajectory.In order to solve above problem,this paper focuses on both indoor mobile trajectory similarity algorithm and indoor trajectory clustering algorithm.First in this paper,based on the indoor space and semantic characteristics of indoor moving trajectory,we design a method based on RFID location semantics to measure the similarity of trajectory.This method extracts critical points of trajectory to reduce the time complexity of algorithm;Then it sets a weight parameter to measure trajectory similarity from two aspects of spatial shape and location semantics.While spatial similarity calculation is based on the definition of distance function which applies to indoor three-dimensional space.Semantic similarity calculation is based on the longest common subsequence and using the moving object's arrival time and stay duration at a track point.On the basis of this,this paper proposes an improved hierarchical clustering algorithm,using linear table structure to optimize logarithmic storage.At the same time,introduce the distributed computing framework of Spark to realize the parallelization of clustering algorithm and improve the computation speed.Experiments indicate that this method works effectively to deal with large-scale data and improves the efficiency of indoor trajectory clustering.