Establishment of a 3D on-site model using UAV image – a case study of a landslide dam in Huisun Forest

• Main Authors: Wei-Ting Chen, 陳韋廷
• Other Authors: Ming-Der Yang
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/nt2c4c

碩士 === 國立中興大學 === 土木工程學系所 === 102 === The geology of Taiwan mostly is weak shale and slate that causes a great potential of mud flow occurrence and soil scour after heavy rainfalls, especially after Chi-Chi earthquake in 1999. Taiwan often suffers nature hazards as earthquakes and typhoons. After the Chi-Chi earthquake, the geographical feature in Taiwan has been significantly changed and shaken. During the typhoon season in summer and autumn, mountainous streams are tend to be blocked by falling rocks and debris from watersheds and cause landslide dams due to heavy rainfalls, such as Typhoon Morakot. The other triggering force is a strong shock by severe earthquakes, such as Chi-Chi earthquake. Monitoring and investigation of landslide dam has become the most essential task to deal with in Taiwan as well as other countries with a great percentage of mountainous area. Remote sensing has been broadly applied to earth surface monitoring, but very limited in landslide lake investigation due to the of imaging conditions. This proposal expresses a three-year successive research project to evaluate and apply novel remote sensing techniques to landslide lake monitoring and investigation. At the first year, worldwide literatures about applying various remote sensing techniques to landslide lake monitoring and investigation will be collected and reviewed. A variety of remote sensing techniques will be compared at cost, efficiency, operation convenience, and environment limitation. Two major triggering factors of landslide dams, such as severe earthquakes and heavy rainfalls, will be analyzed. Also, the monitoring and investigation of those landslide lakes caused by Typhoon Morakot will be followed up. 3D model of landslide dams will be constructed by Image-Based Model (IBM), such as Build the algorithmic poplin such as Structure From Motion (SfM),Visual-Hull, and Minimum cuts for optimal surface reconstruction, base on close-range images shot by amateur digital cameras built on various play forms. In one-scale transformation, the results for the error amount to 1.4963 m, two-scale transformation the results for the XY error amount to 0.8760 m, the results for the Z error amount to 0.2563 m. The Scale transformation is more suitable for the current situation in less amount of information, but also a relatively large error amount. In the coordinate conversion, the results for the X-axis error amount to 0.3238 m, Y-axis error amount to 0.8842 m, Z-axis error amount to 0.7820 m. Finally, using quaternions to get the UAV fight attitude. The results show that the impact of our UAV in flight by the wind, leading 3D reconstruction to produce error.