An Automatic Identification And Extraction Method Of Co-seismic Landslide Hidden Danger Based On Fine DEM

Earthquakes frequently occurring in the mountainous areas of western China and co-seismic landslides distributing widely which always cause huge destroy,the investigation of hidden landslide hazards has become the focus of geological disaster prevention.In face of the high-elevation and hidden co-seismic landslides,the traditional methods for group detection and prevention and optical remote sensing images have shown obvious limitations.How to take good use of the high-precision terrain data to extract hidden landslide hazards has become the key to urgent breakthroughs.In this paper,we take use of the high-precision DEM data generated by airborne LiDAR point cloud,taking the landslide group near the Panda Sea of Jiuzhaigou as the research area,and,from the perspective of fractal theory and geo-statistics,respectively,propose a co-seismic hidden landslide hazards extraction method based on fractal theory and a co-seismic hidden landslide hazards recognition method based on the combine of the Variance function model and support vector machine.The results of extraction and identification of co-seismic landslide hazards in the study area corresponding to the field investigations and verifications indicate the feasibility of fractal theory and semi-variance function model for co-seismic hidden landslide hazards extraction and identification.The main research results achieved in this paper are as follows:(1)Making a comprehensive analysis of the examples of co-seismic landslides in Maoxian and Jiuzhaigou,and analyzing the characteristics and topographic performance of co-seismic landslides based on the study of landslide mechanics.From the perspective of terrain attributes and terrain spatial structure morphology,we select the terrain index,such as slope,terrain curvature,surface roughness,terrain wetness index and semi-variance function,and construct a quantification system of hidden earthquake landslide hazards.(2)Based on the characteristics of spatial self-similarity and spatial heterogeneity between features,the spatial self-similarity and spatial heterogeneity of the co-seismic hidden landslide terrain factor field are analyzed.Based on the variation function in geo-statistics,we propose a method for automatic identification of hidden co-seismic landslides based on semi-variance function and support vector machine.(3)In view of the fact that the co-seismic landslides with different degrees of development have different topography and landform representations and cannot be extracted reliably using the unified factor threshold,from the perspective of fractal,the fractal nature of co-seismic landslides is pointed out.An F-A fractal model is proposed to use the fractal theory model based on a specific terrain index to initially extract and study the potential hazards of co-seismic landslides.This paper analyzes the differences between the co-seismic hidden landslide hazards and the high background values,such as steep slopes and steep cliffs,and proposes a method of extracting co-seismic landslides with terrain constraints.(4)Taking the landslide group near the Panda Sea in Jiuzhaigou as the study area,the above two research methods were used to identify and extract hidden dangers of the landslide,and the extraction effects of the two methods were compared and analyzed in combination with the field survey.Based on the fractal theory,the co-seismic landslide hazard extraction can extract the trailing edge and side edge of the landslide hazard more completely,and it can also extract the landslide that has hidden hazards well;The method of co-seismic landslide hazard extraction and recognition based on the semi-variance function model and support vector machine can well identify the obvious development of high steep slope crack hidden areas,and has a good effect on the identification of suspended hidden hazards.