白蝦養殖投放飼料與生長曲線轉換之水下識別技術研究

隨著全球人口增長和食品需求的增加，白蝦養殖業已經成為全球重要的糧食供應源之一。然而，傳統的白蝦養殖方式存在著許多問題，如高成本、低效率、環境污染和動物福利問題。為了解決這些問題，本研究對智慧養殖白蝦進行了研究。其中以蝦的觀察為重，蝦是底棲動物，在觀察與決策方式完全以經驗法則做判斷，有經驗的漁民能透過水色、殘餌、泳姿、活動情況來做決策，這方面的經驗需要長時間的養成，並且在過人中容易驚擾到白蝦，嚴重甚至造成死亡。本研究主要以南美白對蝦為實驗對象，針對白蝦養殖過程分別在實驗場域與實際場域收取水質資料及水下照片，將水下照片經過MSRSR增強後得出殘餌量及白蝦體長，並且得知白蝦的生長變化曲線、體長體重的關聯以及飼料轉換率，並將其帶入機器學習，進而將養殖的經驗法達到一個可解釋的數據化。實驗結果得知，實際場域的水質變化主要與溫度有關；實驗場域投餵飼料時，pH與OPR會有所改變；白蝦體長推算體重的預測，其決定係數為0.889；白蝦體重增加的預測，其決定係數為0.988；YOLOv7有最高的精確度0.973，而YOLOv5有著最快的訓練速度；MSRCR影像增強後，精確度從0.931增加到0.959。

關鍵字

南美白對蝦；水下影像增強；影像處理；水質監測系統； YOLO ；機器學習；飼料轉換率(FCR)

並列摘要

With the growth of global population and the increase of food demand, white shrimp farming has become one of the important sources of food supply in the world. However, there are many problems in the traditional white shrimp farming methods, such as high cost, low efficiency, environmental pollution and animal welfare issues. In order to solve these problems, this study conducted a study on the intelligent farming of white shrimp. Among them, the observation of shrimp is the most important. Shrimp is a benthic animal. The observation and decision-making methods are completely judged by empirical rules. Experienced fishermen can make decisions based on water color, residual bait, swimming posture, and activity conditions. Experience in this area requires It takes a long time to grow, and it is easy to disturb the white shrimp when passing people, and even cause death in severe cases. This study mainly takes Penaeus vannamei as the experimental object, collects water quality data and underwater photos in the experimental field and actual field respectively for the white shrimp breeding process, and obtains the amount of residual bait and white shrimp body after the underwater photos are enhanced by MSRSR. Long, and know the growth curve of white shrimp, the relationship between body length and weight, and feed conversion rate, and bring it into machine learning, and then make the empirical method of farming into an interpretable data. The experimental results show that the water quality change in the actual field is mainly related to the temperature; the pH and OPR will change when the feed is fed in the experimental field; the prediction of the weight of the white shrimp body length is 0.889; The increased prediction has a coefficient of determination of 0.988; YOLOv7 has the highest accuracy of 0.973, while YOLOv5 has the fastest training speed; after MSRCR image enhancement, the accuracy increases from 0.931 to 0.959.

並列關鍵字

Penaeus vannamei ； underwater image enhancement ； image processing ； water quality monitoring system ； YOLO ； machine learning ； eed conversion rate (FCR)

參考文獻

[1] 鄭金華. 2017. 赴南非參加「2017世界水產學會」發表論文及參訪。

Google Scholar

[2] 漁業局. 2021. 臺灣行政院農業委員會水產局，水產品貿易

Google Scholar

[3] FAO. 2009. Penaeus vannamei. In Cultured aquatic species fact sheets. Text by Briggs, M. Edited and compiled by Valerio Crespi and Michael New.

Google Scholar

[4] Rome. 2020. The State of World Fisheries and Aquaculture 2020. Sustainability in action, THE STATE OF THE WORLD, p30

Google Scholar

[5] Barange, M. 2019. Fishery and aquaculture statistics. FAO yearbook. Fishery and Aquaculture Statistics= FAO Annuaire. Statistiques des Peches et de l'Aquaculture= FAO Anuario. Estadisticas de Pesca y Acuicultura, I-82.

Google Scholar

主題瀏覽