- 期刊
透過可見光魚眼影像評估火雞活動力方法之建置與比較研究
Development and Comparative Study on the Methods of Evaluating Turkey Activity Through Visible Light Fisheye Imaging
摘要
火雞視為西方國家重要節日中不可或缺的食材，近年來由於健康飲食和高蛋白質食品的需求增加，火雞肉因高蛋白且低脂肪之特性而被認為是一種健康的肉類選擇，促使消費者對其需求增加。在台灣因市場需求，火雞的飼養周期長達5個月，因此須導入相關智能監控技術，常態進行火雞群中的健康檢測，以便即時採取相應的防治措施，降低飼養風險。據業者所述，雞隻活動力的狀態是判斷其是否健康的關鍵因素之一。本研究於商業火雞禽舍內建置可見光魚眼影像蒐集系統截取影像資料建構活動力演算法，並與紅外線熱影像活動力監測系統之活動力密度法演算法進行比較。其中可見光魚眼影像資料分別採用光流法與像素點變化法量化標定活動力指標，期望能選配適用於可見光魚眼影像中的活動力演算法。結果顯示可見光魚眼影像在使用像素點變化法時前處理須採用分段二值化閾值的方式進行，統計時則以黑變白與白變黑像素點數之和為指標值，而光流法若能克服光照變化對光流之影響，則兩者計算之活動力R^2（決定係數）可達到0.93，效率部分光流法運算速度較快為3筆／秒。本研究使用之可見光魚眼影像監測系統，除監測範圍可提升4.5倍以上，還擁有管理者更熟悉的直覺式管理視覺，系統未來若能再整合餵飼環境、種類與雞齡等條件參數，將可成為商業禽舍內活動力偵測與管理系統之實用選項。
並列摘要
Turkey is considered an indispensable holiday essential for important festivals in western countries. In recent years, since the increasing demand for a healthy diet and high-protein food, turkey meat has been considered an excellent choice due to its high protein and low-fat characteristics. This demand also led to an expansion of the feeding scale in Taiwan. As the breeding cycle of turkeys is as long as five months in Taiwan, it is necessary to employ intelligent monitoring technology to conduct health checks in the turkey flock, execute accurate time response, and reduce the risk of rearing. According to the feeders' experience, poultry's mobility is one of the critical health symptoms. In this thesis, a visible fiseye image collection system was installed in the commercial turkey sheds to intercept images and construct an activity algorithm for the activity index. This algorithm is then compared with the activity index using the density method algorithm of the infrared thermal image. The activity index of the visible fishing images was quantified by the optical flow method, the pixel dot transformation method, and the activity density algorithm. The results show that the pre-processing of the visible fisheye images using the pixel dot transformation method needs phased implementation with different binary thresholds, and the sum of black-to-white and white-to-black pixels is used as the index value for the statistics. However, if the optical flow method can overcome the influence of light changes, the R^2 (coefficient of determination) of activity indexes calculated by optical flow and results of the pixel dot transformation method with pixel’s sum of block-to-white and white-to-black can reach 0.93. The efficiency of the optical flow method has a faster calculation speed of 3 strokes/second. The visible fisheye image monitoring system used in this study has a lower setup cost, a wider monitoring range that can be extended to 4.5 times more, and a familiar, intuitive image. Suppose the system could be integrated with the parameters of feeding environment, species, and age of poultry in the future. In that case, it will be an excellent practical option for commercial poultry houses' activity detection and management systems.