Research On Landslide Risk Assessment Based On UAV Remote Sensing

Taking Honghe Town,Pengyang County,Ningxia as an example,this paper obtains the orthophoto image of the study area through UAV remote sensing technology,generates the corresponding data layer,identifies the landslide and extracts the information of the landslide elements at risk,and realizes the landslide susceptibility,hazard and risk assessment in the study area.The main achievements of this paper are as follows:(1)Using DJI Phantom 4 RTK and JOUAV CW-30,the UAV aerial photography system is constructed,and the UAV remote sensing spatial information acquisition scheme for the loess hilly area of Southern Ningxia is formulated.The 0.13m UAV remote sensing Orthophoto Image(DOM)of Honghe town has been generated,covering an area of 164km2,which provides data support for subsequent loess landslide identification and development feature analysis,landslide elements at risk information extraction and regional landslide risk assessment.(2)Based on the identification marks of loess landslides and field verification,180 landslides are determined.The landslides in Honghe town are UAV remote sensing identification marks with unique shape,color,water system,texture and surface characteristics.On the whole,the Loess Landslide in Honghe town is more in the South and less in the north;There are more semi-circular and irregular landslides;The length of landslide is mostly 30-300m,and the length and width have a certain correlation.There are many landslides with the length width ratio between 0-1;The effect of intermittent water flow against the ditch creates conditions for the development of landslide on the loess hilly slope in Honghe town;Most of them slide to the north and northwest;The trailing edge of landslide is generally clear and recognizable,and it mostly occurs in opposite sliding landslide;The sliding surface is generally straight,and the sliding surface of the landslide on the shady slope is easier to grow trees;The landforms of the sliding body are diverse,with the phenomenon of double ditch homology.Landslide terraces,drum mounds and other landforms can also be seen.(3)After analyzing the image characteristics of landslide elements at risk,the experimental area is set,and two object-oriented landslide elements at risk information extraction methods of multi-scale segmentation and single-scale segmentation are adopted.The accuracy and time efficiency of landslide elements at risk information extraction based on the two scale segmentation are compared,and the landslide elements at risk information extraction in the study area is realized through scheme optimization,Finally,the optimization scheme of single segmentation scale in the study area is determined to quickly extract the information of landslide landslide elements at risk by using two single segmentation scales:100(shape factor 0.2,compactness 0.5)and 259(shape factor 0.1,compactness 0.5).(4)The landuse data,road data,river data,VDVI data,landslide elements at risk data,landslide data obtained by UAV remote sensing and the data obtained by non UAV remote sensing data sources are used to evaluate the landslide risk in Honghe town.Among them,the landuse data,road data,river data,VDVI data and landslide data participate in the landslide susceptibility assessment,and the ROC value of the final prediction result is 0.814;Landuse data,road data,landslide elements at risk data and landslide data participated in vulnerability and risk assessment.The medium and high risk zone in Honghe town are mainly concentrated in the terraces of Honghe River valley and the southern tableland in the form of dots or slices.The practice results show that the data obtained by UAV remote sensing can participate in landslide risk assessment.