| With the popularization of high-resolution remote sensing images,object-oriented image analysis technology has been applied to many aspects,and landslide recognition is one of them.However,in the data-driven processing mode,landslide remote sensing recognition is mainly for specific landslide cases,based on the interpreter’s understanding of landslides,based on spectral features of image objects and a small number of auxiliary features.Due to the one-sidedness and difference of different interpretations of landslides and the limitations of data-driven processing modes,remote sensing landslide identification is not only subjective and low precision,but also has poor universality and low degree of automation.Aiming at the above problems,the paper introduces the concept of ontology in knowledge engineering,and explores the method of remote sensing identification of landslide through the approach of landslide geographic ontology modeling.The specific research contents and results are as follows:(1)Landscaping concept ontology description.According to the characteristics of landslide remote sensing recognition,the landslide remote sensing recognition knowledge is objectively summarized from four aspects: geographic knowledge,remote sensing image features,image object features and expert knowledge,and the landslide concept ontology is further constructed based on the intrinsic relationship between landslide identification knowledge.(2)Landscaping geographic ontology modeling.According to the method of ontology modeling,the model construction of landslide geographic ontology is carried out,including landslide knowledge modeling(including sensory image ontology modeling,landslide recognition category ontology modeling,image object ontology modeling,etc.)and classifier modeling.Finally,a landslide recognition semantic network model is formed.The model built is the basis for subsequent landslide identification.(3)Geographical ontology-driven landslide identification.Combined with the characteristics of object-oriented image analysis methods,the geographic ontology-driven landslide identification workflow is proposed,and the processing methods of each link are discussed in detail.(4)Landslide recognition experiment.Geographical ontology-driven landslide object recognition was carried out using WorldView-2 remote sensing imagery in Heifangtai area of Gansu Province,which verified the effectiveness of geographic ontology-driven landslide identification method.Through research,this paper proposes a feasible geographic ontology-driven object-oriented landslide recognition method,which provides a reference framework for objective and comprehensive knowledge representation of landslide identification,and builds a slidable and scalable landslide geographic ontology model for landslide identification. |
