乳牛熱緊迫影像監控系統之建置與資料分析

Development and Data Analyses of an Image Monitoring and Control System for Heat Stress on Dairy Cow

蔡雨錡(Yu-Chi Tsai)

指導教授 ： 林達德

國立臺灣大學/生物資源暨農學院/生物產業機電工程學研究所/碩士(2018年)

https://doi.org/10.6342/NTU201803830

摘要

泌乳牛熱緊迫現象會導致產乳量下降三成左右，甚至導致疾病如乳房炎或是跛足。因此，本研究建立了一套影像監控和環境控制系統用來幫助牧場的管理者或是擁有者能夠監控他們的牛隻健康狀況。該系統通過影像和環境數據檢測泌乳牛熱緊迫程度，並根據情況調控環境降溫系統，以預防泌乳牛產生熱緊迫現象。該系統能夠通過物體偵測和機器學習方法自動分析泌乳牛的飲水頻率和行為模式。利用熱緊迫與THI關係的預測模型，系統可以根據不同熱緊迫程度來改變風扇、噴霧和灑水系統等降溫設備的控制策略。影像模組由嵌入式系統Raspberry Pi 3 B+以及pi camera v2建構，用於收集影像和溫度/濕度數據，並對這些數據進行前處理。在前處理之後，系統將這些處理過的數據發送回伺服器並在現場的控制面板以及網頁上顯示，並將結果發送到控制端的變電箱，以調整泌乳牛舍所需的降溫策略。
本研究在五個場域進行多次實驗以驗證該套系統的有效性，第一個實驗場域為台大泌乳牛舍飼料區，在該場域驗證影像系統並改進了物體偵測演算法以及機器學習演算法；第二個實驗場域在苗栗的四方牧場，在該場域測試了系統能夠有環性的自適性並新增了客製化的降溫及警報設定；第三個實驗場域在台南的煌經牧場，在這裡將整套系統模組化並做了多場域的資料結果比較；第四個實驗場域在台大懷孕牛舍，該牛舍分成兩區域做比較，在這邊驗證了使用噴霧系統自動控制的情況下，熱緊迫現象會減少。第五個實驗場域在台大運動場，在此做的實驗為牛隻站臥行為辨識，做為熱緊迫程度的其中一種指標。本研究開發的影像監控和環境控制系統可以幫助減少熱緊迫問題，節省人力和節能，並在網站以及line機器人中提供客製化的的降溫控制策略和警報服務，讓牧場管理者以及擁有者可以監控牛群狀況。

關鍵字

影像處理 ； 物件偵測 ； 監控系統 ； 嵌入式系統 ； 熱緊迫

並列摘要

The heat stress of lactating cows can cause milk production to decrease about 30%, and even lead to diseases such as mastitis or lameness. Therefore, this study established an image monitoring and environmental control system to help farm managers or owners to monitor their dairy cow health conditions better. The system detects the symptom of heat stress of lactating cows through image and environmental data, and regulates the environmental cooling system according to the different situations to prevent heat stress in lactating cows. The system automatically analyzes the drinking frequency and behavior patterns of lactating cows through object detection and machine learning methods. Using a predictive model of the relationship between heat stress and THI, the system can change the control strategy of cooling devices such as fans, mist, and sprinkle systems based on different level of heat stress. The imaging module was built by embedded system Raspberry Pi 3, pi camera v2 and temperature/humidity sensor to collect image and temperature/humidity data and to pre-process these data. After the pre-processing, the system sent the processed data back to the server and displayed it on the web page and the control panel of the site, and sent these results to the transformer substation to adjust the cooling strategy required by the dairy farm.
This study conducted multiple experiments in five fields to verify the effectiveness of the system. The first experimental field was the feeding area of the dairy farm in National Taiwan University, where the image and environmental control system was validated and the object detection algorithm and machine learning algorithm was improved. The second experimental field was Fourways dairy farm in Miaoli, where the system was tested for the ability of self-adaptability, and the customized cooling settings and alarm settings were added. The third experimental field was Huang Jing dairy farm in Tainan, the whole system was modularized and the results of multiple fields were compared. The fourth experimental field was the pregnant area of NTU dairy farm. The area was divided into two areas for comparison. It was verified that the heat stress level was reduced under the automatic control of mist system. The fifth experimental field was the playground in NTU dairy farm. The experiment done here was for the identification of the standing and lying behavior of the dairy cow, which was one of the indicators of the heat stress degree. The image monitoring and environmental control system developed in this study can help on heat stress reduction, labor saving and energy saving, and provide customized cooling control strategies and alarm services on websites and line bots, allowing farm managers and owners to monitor the status of the herd in their dairy farms.

並列關鍵字

Image Processing ； Object Detection ； Monitoring System ； Embedded System ； Heat Stress

參考文獻

黃揚文。2015。合作式影像系統在農業環境作業模式分析之應用。碩士論文。台北：臺灣大學生物產業機電工程學研究所。

徐濟泰。2008。乳牛之飼養管理。畜牧要覽草食家畜篇。297-320。

Aernouts, B., E. Polshin, J. Lammertyn, and W. Saeys. 2011. Visible and near-infrared spectroscopic analysis of raw milk for cow health monitoring: Reflectance or transmittance?. Journal of dairy science, 94(11): 5315-5329.

Allen, J. D., L. W. Hall, R. J. Collier, and J. F. Smith. 2015. Effect of core body temperature, time of day, and climate conditions on behavioral patterns of lactating dairy cows experiencing mild to moderate heat stress. Journal of Dairy Science, 98(1):118-127.

Armstrong D.V. 1994. Heat stress interactions with shade and cooling. J. Dairy Sci. 77:2044–2050.