Research And Implementation Of Key Techniques For Indoor Movement Object Trajectory Prediction

In the development of indoor mobile object location technology,wireless interconnect technology(Wi-Fi)has been used in combination with mobile devices to obtain object location information.However,in a large indoor environment with a large population density,such as airport terminal buildings,train stations,shopping malls,etc.,there may be insufficient Wi-Fi bandwidth,signal dead angle,and the like,which makes it difficult to obtain the position of a moving object in real time.So need to predict where the moving object may move to in the future.Based on the project " Intelligent Airport"(project number: 2017GZ0034)supported by the provincial science and technology department of Sichuan province,This thesis takes indoor moving object trajectory data as the research object,and solves the problems such as poor network transmission environment,large number of track data requests and low positioning delay and low positioning accuracy.First of all,by analyzing the characteristics of interior space,the indoor space is divided by the grid method,and build the interior space model.DBSCAN(Density-Based Spatial Clustering of Applications with Noise)clustering algorithm is used to extract the eigenvalues of the trajectory data and cluster important site points to generate cluster libraries of important location points.In order to verify the effectiveness of the preprocessing method and clustering algorithm,this thesis pre-processes and clusters the public data of UJIIndoor Loc website,and proves the effectiveness of the algorithm.Then,combining with the indoor space model,the trajectory data of the original moving object generated by the current indoor positioning technology is studied,and a method of preprocessing the original trajectory data is proposed.By this method,the trajectory data can be encapsulated into a specific trajectory data model;By using the data obtained in the second step to simulate the prediction model,verify the validity of this forecasting model.On this basis,the historical trajectory data is compressed by combining the cluster library of important locations to simplify the data.Thirdly,for the Hidden Markov Model(HMM)model,the problem of state retention caused prediction failure,combined with the generated state-of-the-point cluster library and the state retention time of the trajectory points,improve the classic HMM model and Establish an improved trajectory prediction model to achieve trajectory prediction of moving objects;Finally,an analog system for indoor location services is designed and developed to implement the application of clustering and prediction models for mobile trajectory data.