Please use this identifier to cite or link to this item:
Title: Change Detection for Surface Mining Boundary Based on Multi-source Remote Sensing Images
การตรวจสอบการเปลี่ยนแปลงของขอบเขตการทำเหมืองแร่แบบเปิด โดยการประยุกต์ใช้แหล่งข้อมูลจากการสำรวจระยะไกล 
Authors: Kawipa Sukkee
กวิภา สุขขี
Burapha University. Faculty of Geoinformatics
Keywords: Remote Sensing
Random Forest Algorithm
Digital Elevation Model
Surface Mining
Issue Date:  11
Publisher: Burapha University
Abstract: Mining is an important industry in Thailand. It can levy gross mineral royalties an average of 3 billion Thai baht per year, and The Minerals Act, B.E. 2560, regulates the industry. The Department of Primary Industries and Mines (DPIM), part of the Ministry of Industry, is in charge of monitoring and promoting the mining industry, including mineral trading, as well as establishing safety and pollution-control regulations. In the past, mining outside the permissible limits frequently occurred. It negatively affects royalty storage and the environment. Because of mining supervision, there are also limitations to tools, methods, personnel, expenses, as well as regulatory frequency. This study applied data from freely available satellite data and open-source software. To assess the suitability of satellite technology applications to detect changes in horizontal and vertical mining in small mining areas to suit mining areas in Thailand. There are two study areas, selected from the average size of all mines currently mining. The satellite data used in this study include Sentinel-1, Sentinel-2, and Landsat 8. The validating data is from the allowed agent to ensure the reliability and accuracy comprising topographical mining data measured by unmanned aircraft (UAV) from DPIM which have high resolution, and the Digital Elevation Model (DEM) from the Royal Thai Survey Department (RTSD). The research methodology used in this study is to extract the boundary of horizontal mining by applying Sentinel-2 data and Landsat 8 by Mean-Shift algorithm and classifying mining areas with Random Forest (RF) algorithms obtains classified into two classes: Mining and Non-mining. The performance of the classification result was assessed based on the confusion matrix formed using the 32 observations for study area 1 and 34 observations for study area 2 from the test samples. The overall accuracy was calculated using the confusion matrix. The vertical boundary mining analysis has applied Sentinel-1 data to extract DEM using InSAR techniques. Then used the DEM compared with RTSD DEM, statistically analyzed by using a coefficient of determination (R²) and root mean squared error (RMSE). Analyzing changes in vertical mining using DEM data obtained from the InSAR technique and analyzing the volume changes of two periods. The result of horizontal mining boundary extraction from Sentinel-2 and Landsat 8, In the first study area round 1 has an overall accuracy of 95.66% and 86.57%. Round 2 of the first study area is 97.47% and 96.50%. In the second study area, round 1, the overall accuracy is 100% and 99.35%. Round 2 of the second study area is 99.26% and 95.90% respectively. Based on validation results, the satellite data from Sentinel-2 is more accurate than the horizontal boundary of mining compared to Landsat 8 data. When using the horizontal boundary of mining from sentinel-2 data to analyze changes in horizontal mining areas, the data of the mining area was used to analyze the changes in the horizontal mining area. In study area 1, mining expansion was 11.64% of the original mining area, according to the reference. And in study area 2, mining expansion was 11.79% of the original mining area, according to the reference. The result of DEM extraction has obtained the result as 14 m resolution of DEM and correlates when compares to DEM from UAV. The result found that the R² and RMSE values are 0.6038 and 34.279 for the study area 1 of the first round, 0.5621 and 35.731 for study area 1 of the second round, 0.2947 and 55.704 for study area 2 of the first round, and 0.2666 and 57.603 for the study area 1 of the second round.  However, the DEM extracted from the Sentinel-1 is highly accurate, but it is not enough to need of vertical mining change analysis of a small mining area. Finally, the application and method of this research to use in change detection of horizontal and vertical surface mining boundaries. Sentinel-2 has a medium level of suitability for change detection of horizontal mining boundary since the change characteristic is similar to the reference data. Landsat-8 is not a suitable choice for horizontal change detection in small area mining and Sentinel-1 is not suitable for detecting the change in vertical mining in the small mining areas.
Description: Master Degree of Science (M.Sc.)
วิทยาศาสตรมหาบัณฑิต (วท.ม.)
Appears in Collections:Faculty of Geoinformatics

Files in This Item:
File Description SizeFormat 
63910060.pdf7.34 MBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.