應用圖像辨識與公民科學於海洋資源永續管理

Applying Image Recognition and Citizen Science to the Sustainable Management of Marine Resources

杜杰(Jay Tu)

指導教授 ： 溫丹瑋

淡江大學/商管學院/管理科學學系企業經營碩士班/碩士(2023年)

https://doi.org/10.6846/tku202300090

摘要

為了解決日益嚴重的海洋資源枯竭問題，完善魚類資料庫並用於生態管理是不可或缺的辦法，本研究採用公民參與科學、YOLO圖像辨識技術與整合性的地理資訊系統為海洋資源管理建立一套魚類辨識決策支援系統，讓釣魚民眾從中取得釣魚資訊與氣象變化並上傳魚類圖片至資料庫中，所蒐集之魚種、漁獲量、地點等資訊集結後可提供即時的海洋資源監控資料給管理者作為監控海洋資料使用。經實驗測試判定為YOLOv7-E6E效果最好，因此選為本研究之魚類辨識模型，本系統之圖像辨識模型是由訓練6種具有特殊性魚種作為示範，同時訓練兩種不同大小YOLO預訓練模型來進行比較。本研究建構整合型網站作為連接使用者與系統之間的橋樑，網站本身運用響應式網頁設計使得網站能在各種裝置上操作順暢。網站建構完成後經內部測試，確認系統運作順暢乃發佈至網路上，提供註冊會員以及訪客帳號供釣魚民眾來使用。此系統發布後，接著於4月1日開始在各大釣魚社群宣傳網站並同時發佈問卷，收集使用者對於網站的反映與建議。截至5月15日，本研究之網站總計有382位會員，另外共收集568份有效問卷。有關使用者滿意度的統計分析結果顯示釣魚民眾對於本系統整體之資訊、設計、辨識、反應及運作上都持正面態度，並且釣魚民眾有意願將來持續使用本系統。統整系統使用紀錄以及問卷之反饋，得知本系統增加整合YOLO圖像辨識、地理資訊系統、氣象與海象及視覺化界面確實能提升公民參與海洋資源永續管理，進而完善魚類即時資料。本研究可以減輕管理者與決策者收集資料的難度，增加資料的即時性及完整性，提供海洋資源管理決策的參考。

關鍵字

公民科學 ； 圖像辨識 ； 海洋資源永續管理 ； 決策支援系統

並列摘要

In order to solve the increasingly severe depletion of marine resources, it is
indispensable to improve the fish database and use it for ecological management. In
this study, a fish recognition decision support system was developed for marine
resource management, citizen science, YOLO image recognition technology, and an
integrated geographic information system. This allows fishing public to access fishing and weather information and upload fish images to the database. The collected data on fish species, catch quantities, and locations can provide real-time monitoring data on marine resources for administrators. YOLOv7-E6E was determined to be the most effective fish recognition model through experimental testing and was therefore selected for this study. This study use web-based as a way to connect users to our system. After the completion of website and internal testing. The system was confirmed to operate smoothly and was subsequently released online. Registered members and guest accounts were provided for recreational fishing public to use. Starting from April 1, the website was promoted in few fishing communities and a questionnaire was simultaneously distributed to collect user feedback and suggestions. Up to May 15, the website had a total of 382 members and 568 valid questionnaires were collected. Statistical analysis of user satisfaction revealed that fishing communities held a positive attitude towards the system's overall information, design, recognition, responsiveness, and operation. They also expressed willingness to continue using the system in the future. These findings demonstrate that the integration of YOLO image recognition, geographic information systems, weather and oceanographic data, and visual interfaces in this system effectively enhances public participation in sustainable marine resource management and improves real-time fish species data. This research alleviates the difficulties faced by managers and decision-makers in data collection, enhances data timeliness and completeness, and provides a reference for marine resource management decisions.

並列關鍵字

Citizen Science ； Image Recognition ； Sustainable Marine Resources Management ； Decision Support Systems

參考文獻

英文參考文獻

Al Muksit, A., Hasan, F., Emon, M. F. H. B., Haque, M. R., Anwary, A. R., Shatabda, S. (2022). YOLO-Fish: A robust fish detection model to detect fish in realistic underwater environment. Ecological Informatics, 72, 101847.

Alves, C. (2021). Marine resource management and fisheries governance in Belize exhibit a polycentric, decentralized, and nested institutional structure. Ocean Coastal Management, 211, 105742.

Arranz, I., Brucet, S., Bartrons, M., García-Comas, C., Alcaraz, C., Bardina, M., Barquero, P. N., Casals, F., Caiola, N., Duran, M. C. (2022). Individual body mass and length dataset for over 12,000 fish from Iberian streams. Data in Brief, 42, 108248.

Barange, M. (2003). Ecosystem science and the sustainable management of marine resources: from Rio to Johannesburg. Frontiers in Ecology and the Environment, 1(4), 190-196.