| A landslide is a natural disaster that is second only to earthquakes in the world today.It has the characteristics of strong suddenness,wide coverage,great hazards,and difficulty in correct warning.It often causes huge economic losses and casualties to humans.China has many mountain ranges,complex geological environment structures,and widespread landslide hazards.The prevention and control of landslide disasters is of great significance and arduous tasks.The traditional landslide research work is often very inefficient,heavy workload,high cost,and difficult to carry out in complex mountainous areas and dangerous areas;remote sensing technology as a powerful means to obtain information without entering the site,is already in the landslide Disaster investigation,dynamic monitoring and early warning,real-time disaster investigation and loss assessment play an irreplaceable role.In recent years,with the tremendous changes in information technology,from "Internet +" to cloud computing,to the development and application of big data and artificial intelligence technology,has brought opportunities and challenges to landslide research.On the basis of reviewing previous studies,the thesis mainly discusses the work of extracting landslide information and predicting the spatial position of landslides in the research of optical remote sensing images in landslide geological disasters.The main contents and results of the research are as follows:(1)Object-oriented(OBIA)landslide recognition based on high-resolution remote sensing images.Based on an in-depth analysis of the object-oriented landslide identification method,a set of processes suitable for obtaining landslide information from multi-temporal remote sensing images is summarized;and the image object characteristics of landslide are quantified from the perspective of static and dynamic characteristics.At the same time,the key contents of image preprocessing,image segmentation,classification method and accuracy evaluation in OBIA technology are discussed.(2)Landslide recognition based on the dynamic characteristics of image objects.The concept of image correlation analysis is introduced into object-oriented remote sensing change detection to obtain the dynamic characteristics of image objects;the correlation coefficient model for single spectral features is improved,and a correlation coefficient model based on combined features is proposed.The thesis further uses the change detection method of multi-temporal image object correlation analysis to carry out the information extraction experiment of landslide disaster.The experimental results show that this method can quickly determine the surface change area,and then achieve the effective extraction of landslide spatial position information;as the correlation coefficient that quantifies the correlation of image objects,it can also be used in other fields with multi-temporal remote sensing images OBIA is used as a dynamic feature.(3)Landslide spatial location prediction based on landslide susceptibility assessment.Based on the analysis of landslide risk assessment,the evaluation factors and methods of landslide susceptibility are summarized.Based on the principle of random forest algorithm,a landslide susceptibility calculation model was constructed,a corresponding program was developed based on Python language,landslide susceptibility calculation was carried out in the study area,and the landslide spatial location was realized according to the idea of "high susceptibility location is most likely to occur landslide" prediction.The experimental results show that the prediction accuracy of the landslide susceptibility evaluation model in the study area based on the random forest algorithm is very high,and it is known from the landslide susceptibility distribution map in the study area that the areas with high risk and above account for 47.88%;Monitoring of high-prone landslide areas. |
PDF Full Text Request