Study On Theory And Method Of Landslide Intelligent Interpretation Based On Remote Sensing

As one kind of slope geological hazards, landslide can have severe impact on the safety of people’s lives and properties because of its outburst, wide influence, graveness, unpredictability, etc. At present, a main research on landslide is region-wide landslide risk assessment and mitigation, in which Remote Sensing (RS) plays a very important role and is used for landslide catalogue, landslide mapping and so on.Nowadays, in the landslide RS application, visual interpretation is the main method to obtain the regional landslide information. But this method is inefficient, subjective, and requires higher demand of knowledge and experience of readers. Besides, it easily leads to the phenomenon of landslides missing. So it influences the output of RS application badly. Therefore, it’s important to do more research on how to entirely use RS data and RS intelligence interpretation technology on regional study of landslide recognition and getting landslide information. It has both theoretical and practical significance on the development of RS applied theory and the prevention of landslide disaster.Intelligent landslide RS interpretation refers that landslides embodied in RS images are recognized, described, understood and classified through the process of computer simulation the sophisticated process usually doing by earth science experts. Currently the method has not reached the practical level because of some theoretical and practical problems. This paper mainly focuses on these problems and has made the following results and innovation:1. This paper studies on the basic content frame of the research about landslide remote sensing intelligent interpretation. And the paper indicates that the further studies should be done on the following aspects such as:interpreting knowledge base structure, interpretation index system, interpreting index quantification method, intelligent interpretation methods, accuracy of interpretation results and so on. So that the basic research direction on landslide intelligent interpretation has been established, in which way the research becomes more systematic. 2. Based on the knowledge applied by earth science experts when doing their visual interpretation, the paper puts forward that the interpretation knowledge base should be composed of recognition knowledge about landslide elements such as back wall, body, etc; and the knowledge of each element includes its spectral, spatial, thematic, DEM and comprehensive experience and so on. The paper also has methodized the visual interpreting knowledge representation of landslide and provided theoretical guidance for the landslide interpreting knowledge which is represented in the computer environment.3. Considering the development of digital image analysis technology, this paper has studied on the landslide remote sensing intelligent interpretation index system and the quantitative method. The paper also establishes index systems which are based on pixel level and object-oriented respectively. Besides, the paper has proposed the method that can be used for quantification representation of landslide spatial form and spatial relationships on images, which uses the image geo-statistical texture index and a series of object-oriented index. The index system enhanced the systematicity, objectivity and operability of knowledge representation for landslide interpretation. The quantitative method mentioned above can help solve the problem that the spatial characteristic features are very limited in landslide of intelligent interpretation.4. Concerning the difficulty of quantifying the comprehensive experience index of landslide recognition, the paper has proposed that the index can be quantified based on the theory and method of slope stability analysis in engineering geology field. How to evaluate the stability of slope with Limit equilibrium theory is one of the research focuses of the paper, in which one difficulty lies in model parameters acquirement. The paper puts forward that the model parameters can be obtained indirectly using RS data. Through the way of landslide geo-hazard susceptibility evaluation the paper has realized comprehensive experience index quantification and shows a new way for solving the problem that geological experts’ subjective experience plays extremely important role in landslide interpretation process.5. The paper has done research on methods of landslide remote sensing intelligent interpretation. After comparative analyzing the three methods which are variety of features based on pixel landslide interpretation method, object-oriented landslide interpretation method and the visual interpreting method of landslide remote sensing, the paper indicates that object-oriented landslide interpretation method is superior to the interpretation methods based on pixel in many aspects such as:landslide interpreting knowledge representation, realizing the analysis of the drive to knowledge and the representation of interpretation results. However, the object-oriented method still needs further study.The ultimate goal of landslide remote sensing intelligent interpretation is to realize full-automatic on the interpretation. What the paper has researched only shows a way to go further in this field. To achieve the goal thoroughly, many work still remains to be done.