Research On Urban Functional Zone Identification Based On Multi-feature Latent Semantic Information Fused High-resolution Image And POI Data

With the rapid development of urbanization,it is difficult to maintain the urban development mode of "incremental expansion",and the new trend of urban development in the future will be to focus on the refined space "inventory to explore potential".Urban functional zones are an important part of the stock space.Making clear the layout of urban functional zones can strengthen the stock utilization and optimize the urban industrial structure.Therefore,accurate identification and division of urban functional zones plays an important role in rationally planning urban development and solving urban problems.Remote sensing technology has the advantages of short revisit period,low cost and wide coverage,and with the development of remote sensing technology,high-resolution remote sensing images contain more and more rich spatial details.Therefore,remote sensing technology has unique advantages in the extraction of urban spatial information.At the same time,the emergence of urban data such as social media data provides rich data resources for urban research and application,and complements the inherent social and economic attributes of buildings that are missing from remote sensing images.However,how to bridge the semantic gap between the low-level features in high-resolution remote sensing images and the high-level semantic features in social media data to achieve remote sensing image interpretation is still a key issue.The probabilistic topic model can identify all kinds of ground objects in remote sensing images by mining the potential information in the images,which is one of the effective methods to fuse multi-source and multi-feature information and bridge the semantic gap.Therefore,aiming at the existing problems in the identification of urban functional zones,the two typical area as the study area in Ningbo city,fused the low-level features of high-resolution remote sensing images and the high-level semantic features of social media data,and uses the Embedded Topic Model(ETM)to mine their potential semantic information to train support vector machines,and finally realizes the accurate identification of urban functional zones.The main research contents and innovation work of this paper are as follows:(1)This paper systematically summarizes the research status and existing problems in the identification of urban functional zones at home and abroad,introduces the characteristics of high-resolution remote sensing images and POI data,and analyzes in detail the feasibility and application prospect of the integration of high-resolution remote sensing images and POI data to realize the identification of urban functional zones.(2)This paper introduces the main principles of probabilistic topic model in detail,and innovatively introduces ETM probabilistic topic model to mine the potential semantic features of the fused feature information to realize the recognition of urban functional zones,aiming at the problem that the traditional topic model recognition accuracy is not high,and obtains good recognition results.(3)The influence of different feature combinations on the identification accuracy of urban functional zones in the case of ETM probabilistic topic model is deeply explored.The ETM probabilistic topic model is compared with the traditional probabilistic topic model in the ability to identify urban functional zones,and the optimal parameters are explored.A method framework for urban functional zone identification based on high resolution remote sensing image data and POI data is constructed.