English 網站地圖 臺大首頁 中文 Site Map NTU SDG
國立臺灣大學生物資源暨農學院網站LOGO
最新消息
news

臺大生物機電工程學系郭彥甫教授榮獲2021中技社AI創意競賽第一名   

   台大生物機電工程學系郭彥甫教授長年致力於機器視覺在農林漁牧方面的應用,並於今年領導團隊以「仔豬觀察員:守護台灣豬」之主題獲得2021中技社AI創意競賽第一名之殊榮。

       中技社以「引進科技新知,培育科技人才,協助國內外經濟建設及增進我國生產事業之生產能力」為宗旨,自1963年起開辦多項獎學金、學術講座等活動。並於2019年起開辦AI創意競賽,獎勵更多跨領域人才結合AI與創意,期望加強台灣AI發展。在2021的AI創意競賽,即是「AI與農林漁牧」。

       郭彥甫教授帶領之機器學習與機器視覺實驗室(MLMV)研究主軸為機器視覺在農林漁牧上的應用。於農業方面,利用深度學習與番茄葉片病斑影像偵測十一種常見病蟲害,並配合通訊軟體供農民即時於田間辨識;於林業方面,開發了國內首款木材物種辨識智慧型手機應用程式,透過木材橫切面之影像辨識高達41種木材,希望建立驗證台灣貴重木材之辨識系統;於漁業方面著重在魚種辨識,辨識十種延繩釣漁船常見的魚種及量測其體長,並自動產生觀察員報告。於牧業方面,研究著重於禽畜類的活動力偵測,如偵測商用雞舍中雞隻活動力異常行為,和豬隻哺乳舍中母豬之哺乳頻率與仔豬活動力異常行為。

       台灣養豬業之產值為全國畜牧業之冠,豬肉在國內的自給率更是達到90%。但在此產業中,從業人口老齡化、人力資源不足、及仔豬育成率偏低等問題仍然等待解決。其中,仔豬育成率為豬農普遍關心的議題,因此郭彥甫教授與其學生以「仔豬觀察員:守護台灣豬」之主題參與此次2021中技社AI創意競賽。該研究透過架設攝影系統拍攝哺乳舍欄位之俯視影像,再訓練深度學習模型追蹤仔豬活動力與辨識母豬之姿態,並結合兩模型辨識後的資訊找出於母豬哺乳時哺乳異常之仔豬,為利用AI為畜養豬產業達到照護智慧化與省工化的目的再更邁進一步。

       

 

Dr. Yan-Fu Kuo received 1st place honor in the 2021 CTCI Foundation AI Innovation Competition 

  Dr. Yan-Fu Kuo, Professor of the Department of Biomechatronics Engineering at National Taiwan University, is specialized in machine vision applications in agriculture. This year, he and his students received 1st place honor in the 2021 CTCI Foundation AI Innovation Competition.The title of their work is “Piglet Observer: Saving Taiwan’s Pigs”.

  CTCI Foundation’s main mission is: “Introducing cutting-edge knowledge of science and technology, cultivating talents, assisting domestic and overseas economic development, enhancing productivities of manufacturing industries”. Since 1963, CTCI Foundation has offered multiple scholarships and academic lectures every year to commend outstanding contribution in various academic fields, encouraging students to innovate and engage in interdisciplinary study. Starting from 2019, CTCI Foundation has hosted the AI innovation competition, rewarding interdisciplinary students that apply AI in innovative ways, with the hope of accelerating the development of AI in Taiwan. The theme of the 2021 AI Innovation competition is “AI and agriculture, forestry, fishery and animal husbandry”.

  The Machine Learning and Machine Vision (MLMV) Lab led by Dr. Kuo focus on applying machine vision and machine learning techniques to agricultural applications. In crop farming, 11 types of tomato pests and diseases are identified using deep learning models and tomato leaf images. A chatbot is also developed to deliver the identification results to farmers. In forestry, 41 wood species are identified using deep learning models and cross-section photos of wood specimens. A mobile application is developed to facilitate the identification. In fishing, 15 longline fish species/types are identified using deep learning models and the images of harvested fish. The body lengths of the fish are also evaluated automatically. In animal husbandry, research is focused on the activity monitoring of chickens and pigs, such as chicken activity monitoring in commercial chicken farms and piglet movement tracking combined with sow lactation detection in farrowing houses.

  The pig farming industry has the highest production value in the animal husbandry industry in Taiwan. However, issues such as an aging workforce, a lack of laborers, and high piglet pre-weaning mortality rate still remain unsolved. The relatively high piglet pre-weaning mortality rate in Taiwan is of concern especially to pig farmers. This is why Dr. Kuo and his students conduct the research “Piglet Observer: Saving Taiwan’s Pigs” and join the 2021 CTCI Foundation AI innovation competition. The research uses embedded surveillance systems to capture piglet and sow videos in farrowing houses and trains deep learning models to track piglets and to recognize sow postures. After training, the output from the two models were combined to further analyze complex behavior such as finding piglets that do not feed properly. The research is a foundation to realize the automatic care of piglet and sow. 

回最新消息