Research On The Method Of Urban Intention Region Recognition Based On Volunteered Geographic Information

At present,urbanization has shown a global trend,and the government work report of China points out that during the 14 th Five-Year Plan,China plans to increase the urbanization rate to 65%,so the urbanization process in China is also in a rapid development stage.The process of urbanization should develop in harmony with the regional characteristics of urban areas.The city image,which is the impression of the city,can represent the development of the characteristics of urban areas to a certain extent.Therefore,the division and identification of urban imagery areas can help analyze the distribution of various types of urban imagery areas,and also play a role in guiding the impressions of various areas of the city,providing a theoretical basis for the coordinated development of urban areas with regional characteristics.At present,the research on urban imagery areas in China is still in the initial stage,and the classification and identification of urban imagery areas have not yet formed a standard,and most of the methods are based on the data collected by human resources,while the research methods based on big data are less used.The innovation of electronic information technology brings massive big data and various data mining techniques,which also provides open source and voluntary geographic big data and refined geographic data analysis techniques for the research of urban imagery area delineation and identification.The purpose of this thesis is to explore the method of dividing the image area of the city based on volunteered geographic information(VGI),combining small scale with large scale,and to propose a method of boundary extraction of urban image region.Firstly,the map of central city of Nanchang is divided into 1km grid,and the image area of city is identified by feature vector.This method is a traditional small-scale recognition method based on large data.Secondly,the study uses the big data analysis technology,uses the clustering algorithm to identify the city image region from the large scale,avoids the subjective interference as far as possible,combined with POI weight assignment to identify urban image regions,and compared the advantages and disadvantages of several common clustering algorithms and identification characteristics,summed up its applicable scenarios.At present,no scholars have carried out research on the boundary division of urban image region,mostly by road or administrative territorial entity division,but the construction land on both sides of the road is of the same type,and there is a certain error in division by road or administrative territorial entity,finally,a KDE based boundary extraction method for urban image regions is proposed.The POI is weighted,and the following conclusions are obtained:(1)The overall recognition accuracy of small-scale gridding is 83%,and the Kappa coefficient is 0.79,which indicates that the classification consistency of imagery areas is high.(2)The K-means algorithm has the feature of simple and fast algorithm,which is suitable for imagery regions with boundaries close to squares or circles,and can be used as a reference for other clustering methods;the GMM model is suitable for imagery regions with polygonal geometric features on the boundaries,and can be used instead of the k-means algorithm without considering the computation time and computer performance;DBSCAN can fit imagery regions with more kinds of shapes The HDBSCAN algorithm is an extension of DBSCAN,which further reduces the subjectivity factor;therefore,the GMM algorithm and HDBSCAN algorithm should be considered first when doing the imagery region classification.(3)Based on the KDE method to extract single-class imagery region boundary from global large scale,the precise extraction of single-class imagery region has high applicability.In summary,this paper analyzes the urban imagery areas from the perspective of machine learning based on the volunteered geographic information,and provides new methods and ideas for the division of urban imagery areas and the extraction of imagery area boundaries.