| As a regional complex dynamic comprehensive system that integrates many aspects of politics,economy,technology and culture,the city has a great impact on people's daily life such as travel,work,social activities,entertainment and so on.The rapidly development of urbanization and the continuous expansion of urban size have put forward unprecedented challenges to urban transportation system and urban management.On the one hand,people's daily activities are recorded and stored by ubiquitous sensors and widely applied functional systems at all times,providing us with multi-sources,heterogeneous and massive spatio-temporal data which have the advantages of update rapidly,economy and convenience.On the other hand,people's needs are changing from basic query and search to the more personalized and personalized location services.For example,when searching for a large shopping mall,people are more likely to be able to get personalized recommendation services according to their preferences and habits on the basis of the smooth arrival.Therefore,how to carry out the correlation analysis of the crowd activity pattern by using these data,reveal the temporal and spatial regularity and preference of urban crowd activities,and provide the basis for people's travel,traffic management and urban planning,has gradually become a hot topic in the academic field.However,the analysis of the existing research related to the theme of urban population activity pattern is mainly from the single angle of road network,population density,city layout and so on,and in the larger granularity(usually administrative division)to carry out the urban crowd activity pattern,but lack of research on the interaction between different factors at a more detailed level.Therefore,from the view of road network,this paper determines the content of urban diversity road network extraction based on mass trajectory,integrated geometric and topological structure characteristics of road network matching,solving the problem of data source of road network.From the perspective of community characteristics,the paper carried out two aspects of research:extracting community aggregation features from multi-source POI,and analyzing spatio-temporal characteristics of crowd mobility on the community level.The more specific research contents are as follows:(1)The urban road network with wide coverage and high accuracy is the basis for analyzing the interaction between road network and the spatial-temporal pattern of crowd movement.Aiming at the spatial-temporal characteristics of trace points such as multi noise,sparse,massive,and uneven distribution,the problems in existing road extraction methods based on trajectory points are analyzed,such as low geometric accuracy,erroneous topology connection,and road network simplification.This paper taking advantage of the median's anti-noise and anti-outlier characteristics,combining principal component analysis(PCA),inverse distance weighted(IDW)smoothing,and kernel density estimation(KDE),a road network extraction algorithm based on the median clustering of mass trajectory points is proposed.The algorithm directly calculates the mass scattered GPS trajectories,which can reduce the effect of noises and uneven distribution on the efficiency of road extraction.And then KDE for the road feature points with weak linear is used to extract cross or turnning regional roads to keep the diversity of characteristics and the road network topology connectivity.It can improve the geometric and topological precision of the road network,and provide the data base for the analysis of crowd movement mode.(2)The matching of different road networks is the key link to provide high quality,different levels of detail and different coverage of road data for crowd movement mode analysis.The concept of road segment matching and road network matching are introduced,and the difference between road network data in geometry(distance,shape,length,angle)and topology are analyzed.Aiming at these differences,based on the introduction of graph theory and graph matching,a new road network matching algorithm based on neighborhood map and C4.5 algorithm is proposed.The algorithm uses a neighborhood graph structure to measure the topology of the road,considering the geometric discrepancy of the road itself and considering the difference of topological adjacency roads;furthermore,a decision tree method based on C4.5 is proposed to construct training models for roads matching prediction.Experiments show that our algorithm can express the topology of roads well and without matching direction problem,and achieves good matching accuracy and high matching efficiency,which providing a up-to-date,accurate and complete road network data for the whole city in the experimentation area.(3)The traditional way to obtain community collection characteristic data through community survey not only takes a lot of manpower and material resources,but also with difficult to obtain a large number of samples data that can reflect the characteristics of the community.In this paper,a multi source POI method is proposed to obtain comprehensive and massive community characteristic data.In order to eliminate the repeated same entities and reduce repetition in the multisource POI,a removal algorithm based on entropy weight and semantic understanding is proposed.The experiment shows that the recalculation method is superior to the existing method in accuracy and recall.Experimental results of community aggregation extraction show that the distribution of various types of communities in different administrative regions shows great differences,which,to a great extent,reflects the main functions,business and population aggregation of the administrative region in the city.It shows that,use the appropriate and efficient repetition removal algorithm,it is feasible and effective to analysis community aggregation characteristics,urban functional area division and business zone discovery by using multi-source POIs.(4)Through the analysis of the time and space characteristics of the road network running state,three different types of road running state can find,and the same running state has obvious difference in the characteristic curve of the different city regions.The analysis of spatial-temporal characteristics of the community population movement shows that the distance of the crowd,the main driving force and the changing trend,etc.are varies with time.For example,during the period of 6:00-10:00 and 18:00-24:00,the movement of people in the central and non-central urban areas shows a strong direction.At this period,the population movement with a long distance,but the crowd between 10:00-12:00 and 16:00-18:00 mianly has a smaller and shorter distance within the administrative region.The analysis results of the running state of the road network and the spatial and temporal characteristics of the community show that the different road operating states in different regions are influenced by the community aggregation characteristics around the road to a certain extent and the congested road,transition road in different regions usually have different measures(traffic control,diversion,setting up tidal lanes,issuing traffic forecasts and optimizing commercial layout)to prevent and alleviate congestion.For example,for the transition roads in the transition zone,the congestion of the morning and evening peak can be alleviated by setting tidal lanes.However,the setting of tidal lanes for the transition roads in the central urban area may aggravate the congestion. |
PDF Full Text Request