Date of Award

2022

Document Type

Open Access Dissertation

Degree Name

Doctor of Philosophy in Civil Engineering (PhD)

Administrative Home Department

Department of Civil, Environmental, and Geospatial Engineering

Advisor 1

Zhen (Leo) Liu

Committee Member 1

Stanley J. Vitton

Committee Member 2

Qingli Dai

Committee Member 3

Chunpei Cai

Abstract

Landslides cause major infrastructural issues, damage the environment, and cause socio-economic disruptions. Therefore, various slope stability analysis methods have been developed to evaluate the stability of slopes and the probability of their failure. This dissertation attempts to take advantage of the recent advancements in remote sensing and computer technology to implement a deep-learning-based landslide prediction method.

Considering the novelty of this approach, this dissertation leads with proof-of-concept studies to evaluate and establish the suitability of deep learning models for slope stability analysis. To achieve this, a simulated 2D dataset of slope images was created with different geometries and soil properties. Next, multiclass classification and regression models in deep learning were used to test the performance of the models. The model performance was evaluated in terms of accuracy and computation time, and satisfactory results were obtained. The results indicated the high potential of deep learning methods in slope stability analysis.

After confirming the feasibility and suitability of deep learning methods for slope stability analysis, a dataset of real-world image data was needed to test whether the deep learning models can predict landslides for more down-to-earth applications. For this purpose, multi-temporal high-resolution DEM data was used to compile a dataset of landslides that have occurred in a study area. The proposed landslide detection method could detect small to medium-size landslides, which was validated with landslide inventories and Google Earth imagery.

The detected landslides were used to train an instance segmentation model, i.e., Mask R-CNN, and a regression model in deep learning. The instance segmentation model was unsuccessful in localizing the landslides in the pre-event images but provided insight into the shortcomings of the adopted procedure and model. The regression model, however, showed encouraging performance in terms of accuracy and computation time.

Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.

Share

COinS