| DC 欄位 |
值 |
語言 |
| DC.contributor | 地球科學學系 | zh_TW |
| DC.creator | 陳薇安 | zh_TW |
| DC.creator | Wei-An Chen | en_US |
| dc.date.accessioned | 2023-6-19T07:39:07Z | |
| dc.date.available | 2023-6-19T07:39:07Z | |
| dc.date.issued | 2023 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:444/thesis/view_etd.asp?URN=110622008 | |
| dc.contributor.department | 地球科學學系 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 本論文開發了一個可自動檢測台灣大都會區的建築物面積的Mask R-CNN 深度學習，並依此建立台灣地震風險模型數據庫的重要元素。為了有效利用 Mask R-CNN 檢測建築物面積，提出了三個步驟：圖像預處理（通過影像銳化處理與從紅外和近紅外波段計算的標準化植被指數來預先處理遙測影像）、模型訓練（以產生通用模型）、後處理（使用遮罩來移除非建築物辨識結果以改善檢測成果）。此模型應用於全台灣，並獲得了約180萬個建物足跡，分別從三個台灣直轄市（新北市、台北市、台中市）計算準確率並得出平均F指數為0.67。另外本研究將建物面積分別與場址效性參數Vs30及斷層最近距離作比較，得出人類活動地區與Vs30與斷層距離均具反比關係，案是人類活動好聚集於場址效應較為顯著之平原區域。本研究的應用成果顯示，深度學習模型提供了更有效的方法來建立建築物資料庫，並且提供了更全面評估自然災害風險的寶貴見解。在實際應用中，本研究所開發的建築物足跡數據庫將為相關部門提供更好的資源分配策略，以應對自然災害風險。 | zh_TW |
| dc.description.abstract | A deep learning model training is implemented to detect building footprints automatically for the metropolitan area in Taiwan. Building a database is one of the main components to develop a natural hazard risk model for Taiwan. This thesis produced building footprint data through Mask R-CNN, a convolutional neural network (CNN) widely used in image segmentation to predict individual objects. To detect building footprints based on the Mask R-CNN, the thesis is proposed of three procedures: image preprocessing (for image preprocessing, a pan-sharpening multispectral image was obtained from the remote sensing data, and a normalized vegetation index was calculated using the red and near-infrared wavebands to increase data information); model training (for training the deep learning model using the preprocessed data to produce a general model); and post processing (considering masks used to remove non-building-like objects to improve detection results). I applied our model to Taiwan and obtained approximately 1.8 million building footprints with an average F-score of (New Taipei City, Taipei City, Taichung City) of 0.67. In addition, this study compared building area with Vs30 and the nearest distance to the fault, and found that the human activity areas have an inverse relationship with both Vs30 and the fault distance. The deep learning model proposed in this thesis provides a more efficient way to build our building information database, which in turn enhances natural hazard risk assessment. Notably, the application of our model can also be extended to seismic risk assessment, providing a valuable tool for disaster management and prevention efforts. | en_US |
| DC.subject | Mask R-CNN | zh_TW |
| DC.subject | 建築物資料庫 | zh_TW |
| DC.subject | 地震風險分析 | zh_TW |
| DC.subject | Mask R-CNN | en_US |
| DC.subject | Building database | en_US |
| DC.subject | natural hazard risk assessment | en_US |
| DC.title | 利用Mask R-CNN 辨識建物輪廓與地震風險分析:應用於台灣都會區 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.title | Detect building footprints by utilizing Mask R-CNN : Application to Metropolitan Taiwan | en_US |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |