Research On The Remote Sensing Dynamic Character Monitoring And Prediction Analysis Of Landslide Disaster

This paper takes advantage of dynamic monitoring of high resolution image of landslide (ex. Hiroshima, Japan), on the basis of systematic and overall summarization of the mechanics of the damage formation, and gets the variation rule of the movement. In addition, it establishes the dynamic space-time database frame for the nature and environment factors of the landslide disasters, on the basis of which, the paper predicts the generation and development of the landslide disaster by making use of concerning regression analysis and nonlinear methods of pattern recognition.For the mechanics and the image characteristics of the disaster formation of the geological landslide in Hiroshima, the research adopts remote sensing image of high resolution to interpret the landslide information, and extracts influence factors according to the concerning data analysis. In this way, the research reaches the aim of dynamic monitoring and prediction of the landslide disaster taking the advantage of remote sensing technique. The data analysis and process mainly include: geometry correction, space registering, Multi-wave bands image digit composition, inlaying, data transformation and other interpretation preprocessing of the remote image of the landslide research district, and then using the 1:50000 landform map and other concerning geology materials, extract superficial geological specification map of the landslide areas, land use specification map, tilt area specification map, soil specification map, terrain specification map and so on. Get the normalization difference vegetation index chart (NDVI index chart) by ratio operation processing of the remote sensing image in the landslide areas. On the basis of that, obtain the degree of weathering specification map, superficial geological specification map, land use specification map, degree of saturation specification map, tilt area specification map, soil specification map, and landform specification map and so on. Finally, inverse the form of the subject category on the concerning landslide factors, and predict the cause, time, site and scale of the landslide accurately, to the aim of identification, analysis, characterization of the landslide.Additionally, besides improving the remote dynamic monitoring, data processing and extent of the management automation, the research builds the dynamic space-time data base frame. Moreover, on the basis of mechanics of landslide formation, with the application of 3S space information technique and support vector machine method on the basis of statistics theory, the research discusses the prediction analysis of the landslide disasters.The contributions of the thesis are summarized as follows:(1) The object classification approach to the landslide disaster was brought forward. This algorithm based on the landslide image object was increased the more classified parameters, besides spectrum brightness value, but also it fully uses the object neighbor relations and the levelrelations, and has also introduced geometry and the structure information and so on, includes object-shape, size, texture in the landslide classification process, so this methed increases the classified precision. By landslide object classified precision analysis, we may see in this classification to participate in classified 6 kind of places categories of things mean value nearly all above0.90, the maximum value are 1, but the standard difference all are smaller than 0.10, the overall classification precision achieves 93.83%.(2) It is the first time to apply the multiple regression analysis of multivariable analysis in the remote sensing dynamic monitoring and data processing and prediction. By taking advantage of the rainfall amount, analysis of the amount of precipitation, degree of weathering and cross section of the research area and so on, the time, site and scale of the disasters were predicted in the Hiroshima area, in the meantime, the automation processing of the amount of precipitation was realized in the remote sensing dynamic monitoring and data processing and prediction.(3) On the internal and external factors of the landslide formation in an overall and systematically, this paper analyzes the basic thought and conceptual frame of the establishment of dynamic time-space database; the data construction adopts mixed mode, and the database mode adopts the mode of the combination of graphs and digits. This thought and method not only reserves the requirements of the GIS space analysis on the data, but put forward the new dynamic time-space requirements.(4) The data of the landslide disaster monitoring dynamically and large scale ground environment information were integrated into the the landslide prediction mode with the help of the "3S" space information technology, and the methed of the prediction of the landslide was realized by GIS space analytical function, and this methed makes the all-information prediction in the landslide disaster.(5) It is the first time that Hiroshima area in japan was taken as a case study area, by use of its basedata, geological landform data, remote sensing image data, the prediction mode of remote sensing image of the landslide disaster based on the support vector machine technology was carried out. The prediction accuracy is improved evidently and superior to those of other methods with similar scale by use of the support vector machine theory, and the region appraisal and the development tendency of the landslide disaster are rapidly realized.