畜禽產業為台灣農業的重要產業之一，台灣單季的肉雞飼養量超過5000萬隻。面對禽舍中大量的肉雞，飼養業者在禽舍管理經常面臨人力不足，使得雞隻健康狀態難以隨時注意，在養期間人員經常進出禽舍也會增加傳染病風險。近年來，家禽產業智慧化逐漸受到重視，已逐漸普遍開始採用智慧監控系統管理禽舍。隨著高速網路如5G技術的進展，使得高解析度影像如4K級的視訊監控用於禽舍更具可行性。本研究透過建置高速網路環境與架設4K及4台Full HD影像系統，進行智慧化禽舍監控。為了有效即時且長期記錄觀測肉雞外觀顏色變化，期望提供有效且準確的色彩資訊供專家遠端監控與診斷雞隻健康狀態。
本研究於商業禽舍架設4K級與HD級視訊監控系統，透過5G網路架構達到遠端監控情境，並於禽舍內架設檔案伺服器NAS蒐集肉雞一週期飼養，共計90天用於實驗分析。為了達到自動化分析與監測，我們對現地場域考量不同時間與天氣之環境光條件，從環境取得顏色產生自製色彩導表，針對攝影機進行色彩校正與分析，最終選擇色差較小的校正矩陣，其平均色差約為6。接著我們透過深度學習技術(YOLOv4)，從860張不同條件的影像提取雞冠特徵進行訓練，訓練後偵測準確率達90%，並且將該功能用於自動提取攝影機中的雞冠顏色。
透過上述環境的建立，我們針對一期肉雞飼養進行顏色觀測與分析。實驗結果顯示，肉雞雞冠顏色隨著日齡增長而改變，其中雛雞雞冠顏色a*值自40~60區間逐漸向+a*方向，b*值約為5~50。至肉雞成熟後，b*值降至10~15。而氣溫變化也會使雞冠顏色受到影響，在氣溫驟降時L*a*b*三個值皆下降。透過本研究可提供一個自動提取雞冠顏色的系統，未來可協助專家應用於與色彩相關之肉雞健康診斷。

The livestock and poultry industry is one of the essential industries in Taiwan's agriculture. There are poultry houses all over Taiwan, and more than 50 million broilers are being raised in a single season. In the face of a large number of broilers in the poultry house, the breeder often faces a shortage of workforce in the management of the poultry house, making it difficult to pay attention to the chickens' health at any time. The frequent entry and exit of the poultry house during the breeding period will also increase the risk of infectious diseases. In recent years, the intelligence of the poultry industry has gradually received attention. To improve the feeding situation of the poultry house, an intelligent monitoring system is used to manage the poultry house. The experts use the monitor screen to diagnose the health status of the chickens remotely.
In this study, by establishing the intelligent video surveillance system in apoultry house suitable for comb color extraction, we observed the appearance characteristics of broilers and performed a statistical analysis of color. It was applied to the breeding environment and poultry growth management, monitoring the health status of broilers, reducing the spread of poultry diseases, and improving poultry production.
The video surveillance system includes a 4K-level camera and 4 Full HD cameras connected to the 5G network and set up in the poultry house to observe the breeding cycle of broilers for a total of 90 days. After the monitoring image is tested under the pressure of storage space, the appropriate camera parameters are selected, and a color guide table is designed for the breeding environment of the poultry house to perform the best color correction on the monitoring image.
In this study, YOLOv4 was used to detect chicken combs to capture color and observe broilers' distribution of external characteristic colors in the CIEL*a*b* color space. The analysis results of broiler comb color show that the color of broiler comb changes gradually with age. The a* value of the comb color of the 4-week-old chicks gradually increases from 40 to 60 to the +a* direction, and the b* value is about 5~ 50. At 11 weeks of age, the b* value drops to 10~15, and the temperature change will also affect the color of the comb. When the temperature drops sharply, the brightness and chroma of the comb will decrease, and the color will be dull; when the temperature increases, the color of the comb will be brighter red, which will improve the chroma. Healthy broiler combs have an apparent L*a*b* color gamut. In the future, an automated color detection and warning system can be used to remotely monitor the status of broilers and develop intelligent poultry house management.
Through establishing the environment mentioned above, we conducted color observation and analysis for the first phase of broiler breeding. The experimental results showed that the comb color of broiler chickens changed with the increase of age. The a* value of the chicken comb color gradually increased from 40 to 60 to the +a* direction, and the b* value was about 5 to 50. When the broiler matures, the b* value drops to 10-15. The temperature change will also affect the color of the cockscomb. When the temperature drops sharply, the three values of L*a*b* all drop. This study provides a system for automatically extracting comb color, which can assist experts in color-related broiler health diagnoses in the future.

[1] 行政院農業委員會，農委會109年年報，單元壹、國內外農業發展情勢，2021。
[2] 行政院農業委員會，糧食供需年報(109年)，Ⅳ、主要項目產品別詳表，2021。
[3] 財團法人中央畜產會，108年家禽重要疾病診斷研判圓桌會議結論，2019。
[4] S. Sunoj, C. Igathinathane, N. Saliendra, J. Hendrickson and D. Archer, "Color calibration of digital images for agriculture and other applications", ISPRS Journal of Photogrammetry and Remote Sensing, vol. 146, pp. 221-234, 2018.
[5] R. Zhou et al., "Relationship between gross primary production and canopy colour indices from digital camera images in a rubber (Hevea brasiliensis) plantation, Southwest China", Forest Ecology and Management, vol. 437, pp. 222-231, 2019.
[6] J. Fernandez-Gallego, S. Kefauver, T. Vatter, N. Aparicio Gutiérrez, M. Nieto-Taladriz and J. Araus, "Low-cost assessment of grain yield in durum wheat using RGB images", European Journal of Agronomy, vol. 105, pp. 146-156, 2019.
[7] F. Sala, C. Popescu, M. Herbei and C. Rujescu, "Model of Color Parameters Variation and Correction in Relation to “Time-View” Image Acquisition Effects in Wheat Crop", Sustainability, vol. 12, no. 6, p. 2470, 2020.
[8] S. Sarkar, A. Ramsey, A. Cazenave and M. Balota, "Peanut Leaf Wilting Estimation From RGB Color Indices and Logistic Models", Frontiers in Plant Science, vol. 12, 2021.
[9] D. Swayne, M. Boulianne, C. Logue, L. McDougald, V. Nair and D. Suarez, Diseases of Poultry. Newark: John Wiley & Sons, Incorporated, 2020.
[10] 王松、胡益源、魏剛才，雞類症鑒別診斷及防治，化學工業出版社，2018。
[11] 陳冠鈞、王思文、廖秀津，家禽場疾病監測的實務與應用：白色/有色肉雞篇，行政院農業委員會動植物防疫檢疫局初版，2021。
[12] G. Cherian, "Symposium: Metabolic and Cardiovascular Diseases in Poultry: Nutritional and Physiological Aspects", Poultry Science, vol. 86, no. 5, p. 983, 2007. Available: 10.1093/ps/86.5.983.
[13] Y. LeCun, Y. Bengio and G. Hinton, "Deep learning", Nature, vol. 521, no. 7553, pp. 436-444, 2015.
[14] G. Marı́a, J. Escós and C. Alados, "Complexity of behavioural sequences and their relation to stress conditions in chickens (Gallus gallus domesticus): a non-invasive technique to evaluate animal welfare", Applied Animal Behaviour Science, vol. 86, no. 1-2, pp. 93-104, 2004.
[15] H. Kristensen and C. Cornou, "Automatic detection of deviations in activity levels in groups of broiler chickens – A pilot study", Biosystems Engineering, vol. 109, no. 4, pp. 369-376, 2011.
[16] D. Pereira, B. Miyamoto, G. Maia, G. Tatiana Sales, M. Magalhães and R. Gates, "Machine vision to identify broiler breeder behavior", Computers and Electronics in Agriculture, vol. 99, pp. 194-199, 2013.
[17] S. Abdanan Mehdizadeh, D. Neves, M. Tscharke, I. Nääs and T. Banhazi, "Image analysis method to evaluate beak and head motion of broiler chickens during feeding", Computers and Electronics in Agriculture, vol. 114, pp. 88-95, 2015.
[18] X. Zhuang, M. Bi, et al, Development of an early warning algorithm to detect sick broilers. COMPUTERS AND ELECTRONICS IN AGRICULTURE, vol. 144, pp. 102-113, 2018
[19] H. Zhang and C. Chen, “Design of Sick Chicken Automatic Detection System Based on Improved Residual Network,” Proc. 2020 IEEE 4th Inf. Technol. Networking, Electron. Autom. Control Conf. ITNEC 2020, no. Itnec, pp. 2480–2485, 2020.
[20] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once: Unified, real-time object detection,” Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit., vol. 2016-December, pp. 779–788, 2016.
[21] J. Wang, N. Wang, L. Li, andZ. Ren, “Real-time behavior detection and judgment of egg breeders based on YOLO v3,” Neural Comput. Appl., vol. 32, no. 10, pp. 5471–5481, 2020.
[22] A. Bochkovskiy, C.-Y. Wang, and H.-Y. M.Liao, “YOLOv4: Optimal Speed and Accuracy of Object Detection,” 2020.
[23] N. Tavakoli, "Entropy and Image Compression", Journal of Visual Communication and Image Representation, vol. 4, no. 3, pp. 271–278, 1993.
[24] A. Sparavigna, "Entropy in Image Analysis", Entropy, vol. 21, no. 5, pp. 502, 2019.
[25] Z. Wang, A. Bovik, H. Sheikh and E. Simoncelli, "Image Quality Assessment: From Error Visibility to Structural Similarity", IEEE Transactions on Image Processing, vol. 13, no. 4, pp. 600-612, 2004.
[26] Q. Huynh-Thu and M. Ghanbari, "Scope of validity of PSNR in image/video quality assessment", Electronics Letters, vol. 44, no. 13, pp. 800, 2008.
[27] D. Montgomery, E. Peck and G. Vining, Introduction to linear regression analysis, 5th ed. Hoboken, NJ: Wiley, pp. 223–253, 2012.
[28] G. Hong, M. R.Luo, and P. A. Rhodes, “Colorimetric Characterization Based on Polynomial Modeling,” Color Res. Appl., vol. 2, no. Iso 17321, pp. 76–84, 2000.
[29] 陳重吉，數位相機色彩校正模式之研究，世新大學圖文傳播暨數位出版系，2005。
[30] C. S. McCamy, H. Marcus, and J. G.Davidson, “Color-Rendition Chart.,” J Appl Photogr Eng, vol. 2, no. 3, pp. 95–99, 1976.
[31] D. Pascale, “RGB Coordinates of the Macbeth Color Checker,” The BabelColor Company, 2006.
[32] K. Jacobson, "Universal Processing and Printing of Colour Film", The Journal of Photographic Science, vol. 9, no. 2, pp. 92–92, 1961.
[33] C. S.McCamy, H. Marcus, andJ. G.Davidson, “Color-Rendition Chart.,” J Appl Photogr Eng, vol. 2, no. 3, pp. 95–99, 1976.
[34] C. A.Poynton, “Rehabilitation of gamma,” in Human Vision and Electronic Imaging III, vol. 3299, 1998, pp. 232–249.