生態農業：農業地景生態監測及復育【土木工程學系蔡慧萍助理教授】

論文篇名 英文：Timely and quantitative damage assessment of oyster racks using UAV images
中文：應用UAV影像進行即時量化之蚵棚損害評估
期刊名稱 IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
發表年份,卷數,起迄頁數 2018, 11(8), 2862-2868
作者 Yang, M. D.(楊明德), Huang, K. S., Wan, J., Tsai, H. P.*(蔡慧萍), Lin, L. M.
DOI 10.1109/JSTARS.2018.2839767
中文摘要 目前台灣以人工方式進行現地蚵棚判識，此方式經常於災損評估時產生補償爭議。因此，本研究提出了一套有效的蚵棚辨識方法，利用無人機（UAV）在低潮位(tide elevation)與太陽仰角(solar angle)較低時進行影像拍攝，以降低鏡面反射影響與獲取清晰蚵棚影像。獲取之影像以形態學為基礎進行開運算與拼接處理，再輔以Canny邊緣檢測法、以邊緣線補齊和多尺度分割技術整合判識出蚵棚邊界，並進一步精確計算出蚵棚的數量與損害比例。本研究於2015年9月4日和10月5日在東石漁港23區進行案例研究，以評估颱風杜鵑的影響。根據蚵棚辨識分析結果，颱風的影響造成了8個固定式(Horizontal rack culture)蚵棚全部損毀及其餘不同比例的損壞，但浮筏式(Raft-string culture)蚵棚增加了60個。此增加的浮筏式蚵棚可合理推測是因為颱風影響，從鄰近區域飄移而來。本研究透過影像精確分析出蚵棚的損壞比例，其中有38個固定式蚵棚由於其損壞面積超過20%，因此符合農作物天然災害補償資格。目前的人工方式現場損壞評估過程需時約一個月，而本研究所提出之方法約於一周內便可有效提供蚵棚判識定量的評估，對於未來蚵棚災損評估之速率與有效性深具實用價值。
英文摘要 Oyster racks identification relies onmanual in situ assessment and often leads to a compensation dispute in aquacultural damage assessment. This study proposes an efficient classification method to identify oyster racks using unmanned aerial vehiclesw (UAVs) images. After image preprocessed by the morphology-based opening operation and mosaicing, the Canny edge detection algorithm, an edge line reparation algorithm, and the multiresolution segmentation techniques are applied to recognize the boundary of oyster racks and the number of oyster racks are further obtained. In this study, a case study was carried out in the District 23 of Dongshi fishing port on September 4, 2015 and October 5, 2015 to evaluate the influence of the Typhoon Dujuan. Based on the image identification, eight horizontal racks were destroyed by the storm surges and strong winds, but the raft-string racks were increased by 60, which were steered from neighbor districts. The damaged ratio of oyster racks was successfully identified and 38 racks were eligible for disaster relief. Comparing to the current manual in situ damage assessment process taking over one month, the proposed process provides a timely and quantitative assessment by efficiently identifying oyster racks on the UAV images within one week, which demonstrated that the proposed process is a promising way for the efficient damage assessment of oyster racks.