| With the popularity of mobile communication,intelligent terminal and other tools that can locate and record GPS trajectory data,some location-based service applications have generated a large number of spatio-temporal trajectory data of moving objects.Identifying stops from GPS trajectories is one of the main concerns in the study of moving objects.The deeper mining of the staying behavior information is conducive to more accurate modeling of the trajectory of moving objects from the semantic perspective,and is benefited from the development of valuable applications,such as taxi hotspots and bus stop settings,the best location for tourism viewing,personalized friendly recommendation,commercial advertising push location and other features of services.It provides decision and support for achieving smart city planning and intelligent traffic management.Although the spatial and temporal characteristics of trajectories have been widely investigated for the identification of stops,few studies have concentrated on the impacts of the contextual features,which are also connected to the road network and nearby POIs.In order to obtain more precise stops information from moving objects,this paper proposes and implements a novel approach that represents spatio-temproal dynamics relationship between stopping behaviors and geospatial elements to detect stops.The main research contents are as follows:(1)Recognize the stops of trajectory based on existing methods.In this paper,based on the existing traditional time-distance threshold method is used to extract the stop candidates,taking account the characteristics of time,distance and speed of the trajectory points,and different constraints are proposed to achieve the purpose of dividing the trajectory,making the recognition of track points more refined,improving the recognition performance and being able to dig more and deeper levels trajectory information.(2)Consider the extraction of stops features under surrounding environmental factors.According to the Mobility Context Cube(MCC),combined wth different types of POIs and road network structures at different times around the stops,the stop duration,the average speed,the average distance between candidate stops and 11 types of POIs,the number of 11 types of POIs,the total number of POIs,the number of road intersections,the number of road,and the grade of road to the nearest section are selected as input features of machine learning models for stops identification.(3)Classfity stops.The 31 spatio-temporal features were used as input elements,different machine learning methods are evaluated to realize the classification of stops.The process includes attribute selection,parameter adjustment,performance evalution and other steps.This paper first extracts stop candidates based on traditional methods,discusses the relationship between it and surrounding environment elements and extracts staying features by constructing a mobility contexts cubes,and then uses three machine learning models for classification to accurately identify stays in the trajectory.The experiments of this research on real trajectory datasets also prove the effectiveness in improving the accuracy of recognizing stops.The experimental results show that:(1)the stops of operating vehicles is mainly concentrated in 5 time periods: 7:00-8:00,11:00-12:00,13:00-14:00,17:00-18:00,23:00-24:00.The number of stops between 14:00 and 15:00 is the least.(2)From 0:00 to 1:00,staying behavior is most likely to occur in the type of accommodation POIs;Between 7:00 and 8:00,the stops are mainly concentrated in the places of accommodation,transportation facility,medical services and education,such as High School Attached to Hunan Normal University,bus station and so on;From 17:00 to 19:00,the operating vehicles during this period tend to stay in POIs such as accommodation,catering,and corporate institutions,and there are mainly concentrated in places such as Lugu Enterprise Square and Wuyi Square Business Circle;Between 23:00 and 24:00,it is more likely that operating vehicles will stay near POIs in accommodation services,medical services,and financial services,such as the Xiangya III Hospital of Central South University and Lufeng Hotel.The staying behavior in these places is in line with people’s living habits.(3)Compared with other algorithms for extracting stops,the accuracy of our method is improved;The accuracy comparison of the three stops classification methods shows that SVM is better than Random Forest and better than KNN. |
