- 學位論文
非接觸式乳牛呼吸頻率監測系統
Developing a Noncontact Respiration Rate Monitoring System for Dairy Cows
摘要
本研究之目的為利用毫米波頻率調變連續波雷達與影像建立了一個乳牛的呼吸頻率監測系統，以進行乳牛群體與個體之熱緊迫現象的偵測及分析。當乳牛遭遇熱緊迫時，會對牛隻的健康、繁殖及泌乳產生負面的影響，造成產業上重大的損失。過去的研究顯示乳牛的呼吸頻率與乳牛遭遇熱緊迫的程度具有高度相關性，而目前產業上廣泛使用溫度濕度指數作為熱緊迫程度的指標。透過乳牛的呼吸頻率，可以進一步考慮到牛隻對熱緊迫的個體差異，然而目前測量呼吸頻率的方法主要是透過人工測量，一個自動化的乳牛呼吸頻率監測系統可以節省牧場大量的勞力及時間。本研究的監測系統包含監測裝置及伺服器兩個部分，監測裝置以Raspberry Pi嵌入式開發板作為系統核心，搭配溫濕度感測器、相機及Texas Instruments IWR1443BOOST毫米波頻率調變連續波雷達感測器，裝設於榨乳室中的欄位上並對準牛隻的側腹部。雷達感測器所收集到的資料、溫濕度資訊及相機所拍攝的側腹部影像會定期透過無線網路回傳至伺服器，在伺服器上利用開發的演算法得到牛隻在榨乳欄位上的時間段及對應的呼吸頻率，並利用側腹部影像透過深度學習演算法辨識牛隻個體，進而達成個別牛隻的呼吸頻率監測。監測系統於實驗場域進行了長時間的實地測試，呼吸頻率的監測結果並與溫度濕度指數及產乳量進行群體與個體之分析。
並列摘要
This study aims to develop a noncontact respiration rate monitoring system of dairy cows using millimeter-wave frequency modulated continuous wave (FMCW) radar and images in order to monitor and analyze the heat stress of dairy cows both by group and individuals. Heat stress was found to affect health and fertility of dairy cows and cause significant reduction in milk production, resulting in great losses for the dairy industry. Based from previous studies, the respiration rate is highly correlated to the level of heat stress, of which Temperature Humidity Index (THI) is the most commonly used environmental indicator. By using the respiration rate of cows, it is possible to accommodate the individual differences under heat stress. However, respiration rate measurement of cows is usually done by observing the flank movements using human eyes, which can be labor-intensive and relative to different observers. The proposed system consists of two main part: the monitoring device and the server. The monitoring device includes a Raspberry Pi, as an embedded system for sensor interface and data transmission, a temperature and humidity sensor, a camera, and a Texas Instruments IWR1443BOOST FMCW radar sensor installed on the side of the dairy cow inside the milking parlor area. The collected data from the radar sensor, temperature and humidity sensor, and side abdomen images are transmitted to the server via Wi-Fi. The actual timestamps of the cows standing in front of the sensor and the corresponding respiration rates were determined in the server. The identities of the individual cows were also recognized using side abdomen images via deep learning method to achieve individual monitoring. The system was tested in an experimental dairy cow farm for continuous monitoring of dairy cow respiration rate, also the data were analyzed with respect to the environmental and milk yield data by both group and individuals.