Research On Recognition Of Passengers Accompany Relationship In Cruise Ship

Location based service(LBS)has been developing in recent years.In the field of indoor positioning,various positioning technologies such as Wi Fi,Bluetooth,RFID and UWB are applied to a variety of indoor scenes,which leads to a surge of locationbased data in indoor space.Data mining and analysis help to obtain potential information and knowledge of data.Cruise ship is a special mobile indoor space,in which various facilities are abundant and provide diversified services for passengers.However,due to its complex structure and indoor space layout and some uncertainties of human activities,it is inevitable that there are potential safety hazards.The purpose of introducing indoor positioning technology into the cruise space is to provide intelligent and humanized services for the passengers on board,and to ensure the public safety on board.The passengers have multiple activities and long residence time in the ship,resulting in a variety of long-time moving tracks.So that the personnel positioning in the cruise space is different from that in the normal indoor space.The work of this paper starts with the UWB positioning system in the cruise,obtains the potential information through the passenger trajectory data collected by the UWB positioning equipment,and puts forward some trajectory data mining strategies suitable for the internal environment and the positioning characteristics of cruise personnel.This paper mainly studies the identification of passengers accompany relationship based on moving trajectory data.(1)A time-sequence similarity measurement model of cruise passenger trajectory is constructedTrajectory similarity measurement is the foundation of trajectory data mining.In order to accurately measure the distance between the moving trajectories of the passengers,the calculation method of the proximity of the position points of the passengers is well defined,which is based on the characteristics of UWB indoor positioning and the data structure of the moving trajectories of the passengers.In the meanwhile,indoor location semantics is introduced into the trajectory distance calculation based on Hausdorff method.Finally,a passenger trajectory timing similarity measurement model based on improved Hausdorff distance is constructed,which is combined with the timing similarity calculation of trajectory.(2)A time-varying clustering algorithm based on trajectory time-sequence similarity is proposedIn this paper,the issue of passenger accompany relationship recognition is solved by time-varying trajectory clustering and intersection algorithm.The traditional accompany relationship recognition algorithm realize the process by the point clustering and intersection results in time sequence,which has low efficiency.In this paper,a clustering intersection algorithm based on time-varying trajectory segments is proposed.The experimental time is divided into multiple intervals.The trajectory segments of each interval are clustered based on the trajectory timing similarity measurement model,and the intersection operation of clustering results is performed along the time sequence.The final set that conforms to the definition is the result of accompany relationship recognition.Compared with other algorithms,the clustering algorithm with improved trajectory distance measurement is verified to be effective in identifying passenger accompany relationship.(3)An algorithm of accompany relationship recognition based on LSTM neural network is presented.The quality of trajectory clustering algorithm is closely related to the accuracy of trajectory similarity measurement method and the characteristics of clustering algorithm itself.In order to solve the problem of high computational complexity in trajectory similarity calculation and clustering,this paper presents a accompany relationship recognition model based on LSTM neural network.This model is used to train the passenger trajectory and the determined accompany relationship contained in it,and learn the end-to-end correspondence between trajectory data and accompany pattern,which can realize efficient recognition of passenger accompany relationship.