Study On Urban House Information Extraction Automatically From Quick Bird Images Based On Space Semantic Model

With the appearance of the high resolution remote sensing image like IKONOS, QuickBird, Spot 5, Cosmos, Orbview and so on, automatic quick accurate extraction of urban object(such as house, road etc.) information has become the focused research topic in the currentinformation extraction and application fields for high resolution remote sensing image. Inrecent years, the progress of artificial intelligence domain made information extracting modelfound on the whole knowledge and semantic network model of the object's relationshipinstead of the individual research model. Space semantic model takes the images as the basicanalysis object and semantic surrounding, emphasizes on adequately taking advantage of thecontext information of the image, analyzes its semantic feature of the space and attribute, andbuilds perfectly semantic network model to express spatial knowledge.The automatic recognition and accurate extraction of the house are meaningful for manyapplications such as getting GIS data, understanding image, big scale mapping and so on.Based on the introduction to the characters and constructing flow of space semantic model,the feature space and context of house information in high resolution remote sensing imageare analyzed. Candidate nodes for building house semantic network model are acquired byedge detection, edge tracing and region segmentation methods respectively. The complexlinkages among semantic networks make themselves be conjoined together and turn intomulti-scale semantic network model. In the end according to the semantic network modeltransforming idea put forwards by Milko Marinov, semantic network model was transformed logisticmodel composed of many functions, which is programmed using Prolog programming language, andfinally different semantic object automatically extracted.This paper was divided into five chapters as follows: The first chapter summarized thepresent study status, and then explained the main content, significance, aim and study flow ofthis paper. The data and its processing methods of the study were also introduced here. Thesecond chapter introduced semantic network model theory in detail based on ontology,including model structure, knowledge expression and inferential mechanism. The thirdchapter was the important part of this paper, the feature space and the whole contextinformation and local context information in high resolution remote sensing image wereanalyzed, and the house semantic network model of Quick Bird image was also constructedbased on region segmentation and edge extraction of image. In the fourth chapter automaticextraction experiment, model exercising and evaluating were accomplished. The fifth chaptersummarized study content of this paper, brought forwards its innovation and stated thelimitation of this paper.