Landslide investigation using remote sensing and GIS techniques : a case study of Bamyan Province, Afghanistan | |
| Author | Nikzad, Ali |
| Call Number | AIT Thesis no.RS-18-06 |
| Subject(s) | Landslides--Afghanistan--Bamyan--Case studies Remote sensing--Afghanistan--Bamyan |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Remote Sensing and Geographic Information System, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Abstract | The landslide is a serious hazard worldwide, which causes significant loss of lives and damage to properties and lifelines every year. Many remote sensing techniques exist to identify the landslides that have their own disadvantages such as time and cost inefficiency. Therefore, an optimal model should be developed to better detect the landslides in the Bamyan province, Afghanistan. The main objective of this research is assessing the automated method of landslide detection using object-based satellite image analysis, which can be further divided into the following three Specific objectives: 1. Conducting object- based landslide detection with several scale parameters to measure accuracy and computing time by the parameter. 2. Finding best features objects and rule sets automatically way to reduce the time computing. 3. Experimenting and comparing the efficiency of different classification algorithms, such as SVM, decision tree, k-NN, random tree, Bayes, and Std. NN. To reach to above objectives, the following process is done as part of my methodology which includes the following: Data entry, Image segmentation, Accuracy assessment for segmentation, Sample selection, Object feature extraction, Image classification, post- processing, Accuracy assessment for classification. Worldview-2 high-resolution satellite images of August 2017 are used in eCognition software. Second, different scales are selected to segment objects based on trial and error method, and accuracy assessment is done for them. Then, Computing time is recorded for different scales to compare the time computation change in terms of scales. After that ruleset is extracted automatically to reduce the time consuming by best feature selection and given membership function value. Finally, Rule-based classification versus other algorithms such as KNN, SVM, Bayes, Decision Tree, Random Tree, and Std. NN. is applied to compare the results based on accuracy assessment. As a result, Rule-based classification has higher accuracy in comparison to others algorithms. For future study, I suggest to develop a better method for segmentation scale selection and using detail geology and soil map as a thematic layer to differentiate better between bare land and landslide objects. |
| Year | 2018 |
| Type | Thesis |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Remote Sensing (RS) |
| Chairperson(s) | Miyazaki, Hiroyuki |
| Examination Committee(s) | Virdis, Salvatore G.P. ;Nakamura, Tai |
| Scholarship Donor(s) | Ministry of Higher Education (MoHE), Afghanistan |
| Degree | Thesis (M. Sc.) - Asian Institute of Technology, 2018 |