Similarity Calculation And Trajectory Prediction Of Indoor Moving Object Trajectories

The continuous acceleration of the urbanization process makes the indoor space become the main place for people’s activities.Research based on indoor trajectory data has a wide range of applications in commercial promotion,indoor navigation,location prediction,etc.,and with the expansion of indoor space and the diversified development of indoor positioning technology,the similarity research and Trajectory prediction also poses more challenges.The current trajectory similarity measurement based on continuous positioning data format is mainly based on the method of character equivalence comparison.These methods pay more attention to the alignment of trajectory points in the space dimension and do not fully consider the position noise and asynchronous scattered sampling encountered in the trajectory acquisition process.,leading to poor robustness of the algorithm;in addition,the existing indoor trajectory prediction is mainly based on the data mining method of the Markov model,which does not well integrate the spatial distribution characteristics of the trajectory and the semantic law of the trajectory into the model for prediction.In response to the above problems,this paper conducts research on the similarity calculation of indoor trajectories and indoor trajectory prediction.The main work is as follows:(1)Combining trajectory sampling and indoor space characteristics,mainly from the perspective of space and time,this paper proposes a new spatiotemporal metric method,called spatiotemporal similarity STS,which can be used to evaluate the spatiotemporal overlap between any two trajectories degree.This method takes into account the influence of position noise and sporadic sampling,based on kernel density estimation,to evaluate their colocation probability at different time stamps and then compare the similarity of trajectories.Object locations are estimated by establishing a personalized spatiotemporal probability distribution,which effectively reduces the need for training data.Then,the similarity of any two trajectories was deduced by calculating the colocalization probability.The experimental results show that the accuracy and robustness of the STS algorithm are higher than the existing algorithms.(2)For the indoor trajectory prediction problem of discrete positioning,in addition to the spatial and temporal characteristics of the trajectory,the hidden features brought by the trajectory semantic information are further considered,and a prediction method based on the hidden Markov method is proposed.The proposed method first clusters Wi-Fi access points based on their similarity;then,the trajectory data is processed based on frequent subtrajectories,which is used to capture and simplify users’ mobility patterns.In addition,the method assumes that the customer’s visit history has some habitual semantic patterns,and designs a continuous semantic algorithm to mine the semantic hidden information of the trajectory;finally,the method integrates the method based on frequent sub-trajectories with continuous semantic prediction into a unified framework,which further improves the predictive ability.In this paper,extensive experiments using indoor datasets show that the prediction accuracy of our algorithm is higher than that of existing algorithms.