2023-01-012024-05-13https://scholars.lib.ntu.edu.tw/handle/123456789/653335家禽是國內重要經濟來源，也是國內主要的肉類來源之一。在傳統管理方式中，業者多透過人力親自巡視雞舍以觀察雞隻與墊料狀態，例如雞隻散佈程度不均勻、活動力下降、體溫異常或墊料過於潮濕等。然而此方式不但費時、管理效率低，且頻繁進出雞舍容易傳播病原菌。故本計畫將開發滑軌式巡場載具，並搭載定位系統、可見光與熱影像模組，應用5G網路高速且低延遲的優點將影像傳輸至雲端伺服器，再透過訓練好的深度學習模型，進行雞隻與墊料狀態之自動監測。本年度將設計滑軌式巡場載具、載具定位系統、載具用的影像模組，與5G網速之雲端傳輸與儲存模組。滑軌式巡場載具將行駛在巡場軌道系統上，定速於雞舍內巡航。定位系統則用於自動偵測載具於雞舍中之位置。影像模組可蒐集可見光與熱影像，用於拍攝雞隻與墊料。蒐集的影像再透過5G網路，回傳至雲端伺服器。此套系統之開發可輔助業者自動監測雞舍內部環境，降低傳統飼養人員巡視之勞動成本，輔助雞舍之管理。 Poultry is an important economic industry in Taiwan, constituting a major part of the daily protein diet in the country. Conventionally, poultry farmers patrol in chicken farms to monitor the status of chickens and litters. Common issues to be monitored include inhomogeneous distributions of chickens in the farm, low vitalities of chickens, abnormal body temperatures of chickens, and wet litters. However, manual patrol is time-consuming and inefficient. Moreover, frequent patrols may increase the risk of avian influenza while entering and exiting chicken farms. This project proposes to develop a rail-based overhead surveillance system to monitor chickens in a chicken farm automatically. The system will comprise of a vehicle, a positioning module, an imaging module, and a 5G internet module. The vehicle will be designed to automatically navigate in a chicken farm. The positioning module will provide the position information of the vehicle. The imaging module will comprise of an RGB camera and an infrared camera. The images acquired by the RGB and infrared cameras will be rapidly sent to a cloud server through 5G internet for further usage (e.g., for training deep learning models). The developed rail-based overhead surveillance system is expected to reduce labor costs from manual patrolling and to aid in chicken farm management.深度學習;自動監測;滑軌式巡場載 具;;Deep learning;Automatic detection;Rail-based overhead surveillance vehicle