2020 Fiscal Year Final Research Report

Developing algorithm for estrus detection based on social relationships between cows

Research Project

PDF

Project/Area Number	18H03294
Research Category	Grant-in-Aid for Scientific Research (B)
Allocation Type	Single-year Grants
Section	一般
Review Section	Basic Section 61030:Intelligent informatics-related
Research Institution	Tokyo University of Science
Principal Investigator	OHWADA Hayato 東京理科大学, 理工学部経営工学科, 教授 (30203954)
Co-Investigator(Kenkyū-buntansha)	堂腰顕地方独立行政法人北海道立総合研究機構, 農業研究本部酪農試験場, 研究主幹 (40506606) 窪友瑛地方独立行政法人北海道立総合研究機構, 農業研究本部酪農試験場, 研究職員 (50825338) 大島一郎鹿児島大学, 農水産獣医学域農学系, 准教授 (60465466) 鍋西久北里大学, 獣医学部, 准教授 (90575151) 古山敬祐地方独立行政法人北海道立総合研究機構, 農業研究本部酪農試験場, 研究職員 (50611026)
Project Period (FY)	2018-04-01 – 2021-03-31
Keywords	機械学習 / 帰納学習 / 乳牛 / 繁殖 / 視覚化
Outline of Final Research Achievements	In this study, to develop an algorithm for estrus detection based on social rank among individual dairy cows, we created a model for automatically estimating social rank in a herd from surveillance camera images. The model was trained by yolov5, which is a fast object detection model of deep learning method, from the dataset for individual identification, and the fit rate was about 95%, which enabled practical individual identification. In addition, we examined the characteristics of individuals based on the time measurement of behavioral classification and compared the rankings by the tank prioritization method with the rankings in this study and found that the rankings of seven out of eight animals were the same. In addition, it was suggested that the rankings were close to each other when the distance between individuals in the video was small. This suggests that it is possible to detect estrus by estimating the social rank from the video and tracking everyone.
Free Research Field	知能情報学
Academic Significance and Societal Importance of the Research Achievements	本研究成果では，センサデバイスを使用せず監視カメラ映像のみで牛群内の社会的順位を推定することが可能となったが，その手法として深層学習の一般モデルであるyolov5を利用したことが特徴的である．これは，汎用的なモデルでも高性能に牛群の個体識別・追跡が可能であることを示した．更に，高速推論モデルの利用によりリアルタイム監視システムが可能となり，発情発見後の即時対応が可能となる．これらにより，ICT技術の生物学への適応と酪農経営における省力化に貢献していると考えられる．