Advanced data collection and analysis methods using UAV-LiDAR/optical/multi-spectral data for landslide analysis

Abstract

A landslide is a phenomenon in which soil collapses and flows downstream, triggered by external causes such as rainfall and earthquakes. With characteristics such as high speeds, large impact forces, and long runouts, landslides are challenging to predict and hence can cause enormous damages. In recent years, landslides have occurred with higher frequency in urban areas, resulting in significant damage due to locally heavy rainfall. Such catastrophic disasters can be prevented by accurately predicting landslide characteristics (i.e., location and volume) and the extent of landslide-related damage through a flow analysis. Furthermore, countermeasures against landslides should be installed in optimum locations through prior survey and analysis. The resolution of topographic information significantly impacts the accuracy of landslide prediction. Ground-based platforms have been employed to acquire topographic information. For example, a point-based method has been used to acquire data at individual points using hand-held equipment, but this method is not applicable in large areas or sites with steep slopes or cliffs that are difficult to access due to time and manpower constraints. Terrestrial light detection and ranging (terrestrial LiDAR), also referred to as terrestrial laser scanning (TLS), is increasingly applied as it is a relatively capable method of obtaining high-resolution points. However, this method is considerably affected by occlusion, a phenomenon in which data are not obtained due to objects being blocked off or closed up. Occlusions can occur when the target is placed in the direction of the source propagation of the LiDAR sensor. Although there is a method to utilize satellite data, certain data may not be available depending on the orbit cycle, revisit period, and weather conditions, and there is the additional issue of low resolutions that could result in inadequate data for landslide analysis purposes. Few studies have been conducted to obtain high-resolution topographic information. In this Ph. D. dissertation, a study is conducted on the development of analysis methods for landslide analysis and the maintenance of barriers by acquiring high-resolution three-dimensional (3D) LiDAR point clouds and multi-spectral data using an unmanned aerial vehicle (UAV). This dissertation describes the development of a system comprising an UAV with a LiDAR sensor capable of efficiently obtaining high-resolution topographic data (0.2 m horizontal resolution) according to flight altitude and velocity. To verify the multiscale model-to-model cloud comparison (M3C2) method for topographic change analysis, an experiment was performed and the result indicated very high accuracy (0.15% error rate). The UAV system was applied to a recent landslide site in South Korea (occurred on August 7, 2020) to obtain high-resolution topographic data within a short period of time after the landslide occurrence. The initial and final volumes were computed using the topographic information obtained from the 3D LiDAR sensor. Using multi-spectral data obtained from a landslide site, normalized difference vegetation indexes (NDVI) were derived to analyze the landslide area. In addition, multi-spectral data were also gathered to estimate the landslide damage area and the distribution of water content in the soil. Moreover, this study investigates the potential application of the UAV system for barrier sediment monitoring by analyzing a recently dredged closed-type barrier installed in South Korea using LiDAR and red-green-blue (RGB) point clouds. The UAV system results were compared with the results of a field survey that was conducted prior to dredging by the local government. Finally, this study derives a monitoring method to determine maintenance frequency according to the amount of sediment collected behind a barrier. The UAV-LiDAR/optical/multi-spectral camera system and analysis technique developed and verified in this study are an economical and efficient method to obtain high-resolution topographic/soil information related to landslides. Therefore, this study provides information necessary for landslide research and development of various construction automation technologies.