Artificial intelligence and Infrared thermography – the future of Dairy Production Management in India

Artificial intelligence and Infrared thermography – the future of

Dairy Production Management in India

Gayathri S. Lal

Ph. D. Research Scholar, LPM Division, ICAR-NDRI, Karnal-132001

Email: gayathrisherlylal@gmail.com

Abstract

India continues to be the largest producer of milk in the world. The Government has initiated several productivity enhancements measures to increase milk production, and significant improvement is evident in the last five decades from 20.8 million tonnes (1970) to 209.96 million tonnes (2021). But the technical innovation based on thermal imaging and artificial intelligence is lacking among the Indian dairy production management scenario. The advancement can provide precision in tackling the burning issues of sub-clinical and clinical mastitis, one of the greatest production diseases causing huge economic loss to the dairy farmers, lameness and other temperature related issues in any herd.

Key words: artificial intelligence, dairy management, milk, precision, sub-clinical mastitis, clinical mastitis, thermal imaging,

Introduction

India is an agrarian country whose wealth lies in dairy production and its dairy animal population. Henceforth, the four pillars of livestock production – housing, feeding, breeding, and management plays a significant role in productivity enhancement, followed by the sustainable economic growth of dairy sector in India. The improvement in milk production in India is technology driven. Smallholder dairy farmers contribute maximum milk in the countries’ milk production. The estimated number of rural families in India is approximately eight crores in dairy production. The rural market consumers (Kaur et al., 2014) consume more than half of the total milk produced. About 75% of milk is consumed at the household level, which is not a part of commercial dairying (NDDB, 2019). Most consumers prefer loose milk and is having the largest market in India as it is perceived to be fresh. In any livestock farm, several problems occur on a day-to-day basis and need addressing in one way or the other for the smooth functioning of the farm. Some common problems encountered in high-producing dairy animal farms pertain to production activities and milking environment. The related issue includes mastitis, metritis, lameness, heat stress during lactation, teat and udder damage, etc. Among them, mastitis is one of the most challenging issues in front of the dairy farmers. Most of the above problems have a common factor, i.e., change in temperature. Scientific studies reported that the udder surface temperature increases from the normal temperature during mastitis (Metzner et al., 2014) and sub-clinical mastitis (Digiovani et al., 2016). Besides that, temperate change (0.62-0.66°C) was observed in the teat during stress, in case of machine milking and milking tubes of milking machine of the corresponding quarter affected with mastitis (Marrero et al., 2020). All these temperature change-related aspects under various physio-pathological conditions can be assessed using effective infrared thermography (IRT), which forms the ideal solution for detecting several problems in dairy farms. IRT is an easy, efficient, onsite, and non-invasive diagnostic tool. It has highly sensitive thermal cameras, which can detect minute variations in surface temperature. It can be used as the most convenient, supportive, and portable diagnostic tool in various conditions of livestock farm management (Sinha et al., 2018). The recent development of sensor-based technology may be suitable for large commercial dairy farms, which is lacking in the Indian scenario. Therefore, IRT caught the attention of researchers for its multipronged usability as an assessing tool for detecting any surface temperature change pertaining to various physio-pathological conditions.

India is one of the largest milk producers in total milk production, with a share of 20.17 percent worldwide (NDDB, 2019). However, India stands first in milk production. Still, mastitis beholds a lion’s share position in terms of economic losses among the farmers of dairy industry as well as among the diseases of dairy cows with high prevalence and incidence rates. Mastitis is one of the most common multi-etiological diseases affecting milk quality. The quality of milk is affected by mastitis, and the quantity of milk and udder health status is also affected. Two decades back, there were low clinical (1 to 10%) and subclinical mastitis (10 to 50%) incidence rates among cows in India. However, recent studies reported an increase in the incidence of subclinical mastitis (20 to 83%) in cows. During the last five decades, the economic losses in India due to mastitis increased 115 folds (6,000 crore). A delay in detecting sub-clinical mastitis contributes to the increased incidence of clinical mastitis. The losses incurred are due to loss of milk production (temporary or permanent), poor milk quality, disposal of milk from affected animals, and reduced productive life or pre-mature culling of the cow. The loss due to subclinical mastitis is higher than the clinical mastitis, and milk yield loss from mastitis ranges from 100 to 500 kg/cow/lactation (Srivastava, 2015). Production loss associated with clinical mastitis in cows is 700 kg (first lactation) and 1,200 kg (second or higher lactation) (Wilson et al., 2004). Mastitis causes the dairy industry an annual loss of approximately ₹7165.51 crores in India, while from sub-clinical mastitis, ₹4151.16 crores, accounting for 57.93% of the total loss. Economic loss due to subclinical mastitis in cross-bred cows was ₹592.87 per lactation, and loss due to decrease in milk production was ₹700.18 (Kumari et al., 2018). For early detection of mastitis, several methods available till date. These includes evaluation of EC, color, CMT, pH, SCC, detection of enzymes such as lactate dehydrogenase and N-Acetyl glucosaminidase, and culturing the bacteria. Advanced techniques are also used, such as PCR, proteomics, ELISA, and mass spectrometry for biosensor systems, biomarker detection, and chip-based diagnostic techniques, which are used under field conditions and in precision dairying (Deak et al., 2019). However, these diagnostic methods are laboratory oriented, lack complete accuracy, and need considerable time of farm staff or milker (Polat et al., 2010; Hovinen et al., 2008; Metzner et al., 2014). Therefore, a cost-effective, rapid, non-invasive cow side diagnostic technique with a potential application in the field is essential for monitoring udder health. In such circumstances, early assessment of mastitis can bring a lot of difference, and Infrared thermography (IRT) is a promising tool. Researchers have used IRT as a tool to assess the udder skin surface temperature (USST) changes in mastitis affected as well as healthy quarters of dairy animals (Sathiyabarathi et al., 2016; Polat et al., 2010; Hovinen et al., 2008; Metzner et al., 2014) and its analysis of thermograms (Deak et al., 2019). These studies were conducted under an induced model of mastitis infection or simple assessment of healthy and mastitis-affected quarters. However, early assessment of mastitis and sub-clinical mastitis during the natural course of infection in Indigenous cows and buffaloes is yet to be explored. Faulty attachment of the milking machine and other factors pertaining to milking machine components such as teat liners and vacuum pressure may hamper the harmony of cell lining of teat leading to trauma (Sharma et al., 2012). Marrero et al. (2020) reported that the milking tube temperature assessed using IRT showed an elevated temperature in mastitic quarters. Recently, image-based analysis by developing an algorithm and training software for automatic detection of the surface temperature (Lowe et al., 2020) in dairy animals is another way forward for precision farming. Thus, algorithm-based model development is the need of the hour

Infrared Thermography

Covid-19 pandemic has taught mankind a new era of digitalization and usage of technology in a mode virtual reality. The situations prevailed during pandemic showed us how thermal imaging can be utilized effectively in temperature screening of the individuals in any common platform, be it in a railway station or airport. The people were screened automatically via passing through the detector, which was monitored using an operator. Thus, thermal imaging became a common platform for the common.   Infrared thermography (IRT) commonly referred to as thermal imaging, is a technological advancement developed initially for military purposes, during 1940s in industry and during 1959 for medicine.  The basic principle underlying IRT is that every object whose temperature is above absolute zero emits infrared radiation that is not visible to our naked eyes. The ability of an object to emit radiation is emissivity. The basic law that governs the proportion of radiation emitted and temperature is Stefan-Boltzmann law. The detector of the highly sensitive IR camera captures such radiation and converts it into electrical signals. Then it is further processed in to thermal image form that is visible to our naked eye. Sir William Herschel first identified Infrared radiations (Wave lenghth-700 nm to 1 mm) during the year 1800. IRT is a simple, effective, on-site, and non-invasive tool that uses surface body temperature to generate images without causing radiation exposure. The thermal images/ thermograms, depicts the warmest region as white/red and the coolest as blue/black (Colak et al., 2008). The emissivity range of cattle external body surfaces is 0.93 to 0.98, and further depends on skin and hair color, and density of hair. Maximum IR temperature and its change are preferred in scientific study purposes rather than average and minimum IRT of external body surfaces as maximum IRT is better associated with changes during lactation variables, behavioral responses of emotions, and metabolism. Minimum and average IRT calculation are susceptible to variation in positioning of the frame of thermogram, and external aspects such as adhesion of water droplets, mud or faeces on external body surfaces. In addition, usage of single pixel to estimate a temperature, i.e., when maximum IRT values are used, pixel error is an outward outcome, but this usually has minor variance than that of minimum or average temperature (Uddin et al., 2020). According to Gloster et al. (2011) in cattle the core body temperature is reflected by the maximum IRT of the eyes, specifically the skin surface around the inner corner of eye socket, because eyes are located close to hypothalamic thermosensitive site, which reduces the lag time in response. In addition, authors claim that, IRT of the eyes was not predisposed by ambient temperature, and eyes have superficial capillaries beds in caruncula lacrimalis and posterior border of eyelids, that are highly innervated by sympathetic nervous system. Uddin et al. (2019) reported a positive relationship between eye IRT and right laterality, and a negative relationship between eye IRT and milk fat content, which suggests that anxious cows had higher eye IRT and lower milk fat content. Another aspect is that flight or fight response is mainly controlled by the right side of the brain.  Also, left eye is connected to right side of the brain and passing on the right allows an animal to view the probable threat more easily with their left eye. So, IR images must be taken only after 10 minutes of rest in the milking shed if the eye temperature needs to be captured during milking hours. Idris et al. (2021) suggest that IRT of cows’ external body surfaces indicates surface temperature that is linked to peripheral temperature than the core body temperature, particularly in the presence of environmental or emotional stress.

Assessment of infectious diseases and estimation of welfare status among farm animals using IRT is effective though accurate diagnosis is not possible by IRT, evidence of pathology for diseases, such as mastitis is evident. Minute change in the udder surface temperature can be assessed using sensitive IR cameras. In addition, the temperature of the udder reflects the body temperature of normal cows.   Thus, IRT as a convenient portable tool in livestock management is effectual.

Artificial neural network in mastitis management

McCulloch and W. Pitts developed the idea of artificial neural network based on the concept that sophisticated algorithms can imitate the learning process of human. In ANN, an input layer, one or more hidden layers consisting of small computing units called nodes or neurons, and a final output layer are formed by the arrangement of a large number of input parameters. The information propagates in neural network models initiates when the data is supplied in the input layer. Each node in the following layer receives the inputs after they have been weighted. To create the nodal output, the weighted inputs are then added together and transmitted through a transfer function. The estimated value of pre-determined error function is then transmitted back through the network via multiple approaches including back propagation, and using the output results as a comparison to the original results. This method modifies the weights of each link based on the information in order to slightly lower the value of error function. The network will change to a scenario where the size of errors will be minimal after repeated correction for numerous training cycles.

An ANN model is useful in routine health monitoring to detect respiratory diseases among pigs, lameness among cattle, and sheep, and heat stress among cattle, which will be useful for adopting the precise management decisions as soon as possible to lower a herd’s mortality rate. One of the most significant diseases linked to financial loss for dairy farmers is mastitis among dairy cows and is linked to the animals’ welfare too as mastitis causes usage of antibiotic therapy and physical damage to udder.  ANN model can also be constructed using a number of milk parameters, including somatic cell count, electrical conductivity, and milk composition index. The use of this model in milk recording devices will enable earlier and more accurate pathogen detection before clinical manifestation. As a result, the incidence of mastitis in a dairy herd can be decreased. Self-organizing networks (SOM) and multilayer perceptron ANN models can identify the presence or absence of mastitis in animals at early stages with improved accuracy. In addition, ANN models are used in detection of estrous in cattle, and sheep, calving date prediction, probability of dystocia prediction and epidemiological outbreak of diseases (Yu et al., 2022).

Thermal image analysis for mastitis and Algorithm development

Zaninelli et al. (2018) conducted a study to evaluate the udder health status using IRT. Here, thermo-graphic images were analyzed using algorithm software and showed a significant logarithmic correlation between (Udder Skin Surface Temperature) and SCC (Somatic Cell Count). Sensitivity and specificity in classification of udder health were- 78.6% and 77.9%, respectively, considering the level of SCC of 200,000cells/mL as a threshold to classify a subclinical mastitis or 71.4% and 71.6%, respectively when a threshold of 400,000 cells/mL was adopted. Thus, the said study opened up the window for automatic screening of mastitis through IRT, which needs further validation via extensive research activities. Similarly, Khakimov et al. (2022) obtained a linear model indicating the relationship between SCC and USST. The authors claim that the basis of their algorithm is the relationship between USST and milk yield. A strong negative correlation for USST range of (36–39 °C) to milk yield was proved among mastitis affected cows. Thus, using SCC, USST-milk yield-based algorithms will aid in faster detection of mastitis among cows in any dairy farms. Lowe et al. (2020) conducted a study to frame an automated system for collecting thermal infrared images of calves and on the development and validation of an algorithm capable of automatically detecting and analyzing the eye and cheek regions from those images. Thus, the said study lit light for a path of new thought in precision farming pertaining to calf health. Yan et al. (2021) constructed thermal models to envisage skin temperature of Holstein dairy cows by mixing the existing models and the formulas related to heat production and loss from the respiratory systems and body surface. The skin surface temperature obtained from the models were compared to thermograms. Results revealed that the two-layer model showed optimum prediction performance when maximum skin surface temperature was considered with THI ≥ 72 rather than mean skin surface temperatures. Authors claim that maximum skin surface temperature of the thermograms is the most appropriate representation of skin surface temperature as it is less sensitive to environmental parameters and is in correlation with the core body temperature. Joy et al. (2021) conducted a study among the sheep to predict the rectal temperature when subjected to heat stress using a combination of IRT and machine learning techniques. They have developed artificial neural network (ANN) models using three different backpropagation algorithms with IRT temperatures, and THI as inputs and rectal temperature measured manually as targets. Results revealed that the thermograms of eye and forehead had the highest correlation with THI and rectal temperature. The model developed was successful in obtaining real-time body temperatures among sheep and detection of heat stress with minimal restraint. Wang et al. (2022) developed a deep learning model based on YOLOv5 networks using left and right USST, OST (ocular surface temperature). You Only Look Once (YOLO) is a unified and real-time target detection algorithm (Redmon et al., 2016) that recombines the target detection task into a single regression problem and uses a single neural network structure to directly predict the position and category of the target box from the image. The authors claim for an accuracy rate 87.62% for mastitis detection with specificity of 84.62%, and sensitivity of 96.30%. Comprehensive detection method using left and right USST manifested a classification accuracy and sensitivity of 6.67 and 62.97% higher, respectively, and correspondingly OST and USST difference detection method showed 19.05 and 25.65% higher, respectively. The Average precision for udders and eyes was >96 per cent and eyes alone was > 98 per cent. The mean Average Precision value of the YOLOv5 algorithm was 96.1%, the highest and detection speed of 116.3f/s, the fastest in comparison to other models indicating the accuracy of the model in detecting the key parts of the cow. Thus, the authors claim that the said deep learning model is the keystone for automatic mastitis detection in dairy farms.  Similarly, Zhang et al. (2020) used deep learning technology using OST and USST difference of 0.8 °C as threshold, and detection accuracy of mastitis for their model was 83.33%. In the future, a depth camera can be added to the collection of thermal images of cows to obtain more complete and accurate temperature information of cows’ eyes and udders and further improve the accuracy of the mastitis detection system.

Conclusion

Livestock management demands advancement in terms of technology and precision farming is its output. The burning issues of mastitis among the dairy farmers needs to be addressed in an effective manner. Rapid, non-invasive cow side diagnostic tool with prospective use in field is necessary for assessing udder health. In such circumstances, early assessment of mastitis can bring a lot of difference and Infra- red thermography (IRT) can be a promising tool. Researchers have used IRT as a tool to access the udder skin surface temperature (USST) changes in healthy and mastitis-affected quarters of dairy animals and its analysis of thermograms based on algorithms- artificial intelligence that are proved to be effective.

References

• Colak, A., Polat, B., Okumus, Z., Kaya, M., Yanmaz, L.E. and Hayirli, A. 2008. Short communication: early detection of mastitis using infrared thermography in dairy cows. Journal of Dairy Science, 91(11):4244-4248.
• Deak, F. L. G. B., Chacur, M. G. M., de Souza, C. D., Andrade, I. B., Cornacini, G. F., Garcia, A. R., Filho, L. R. A. G. 2019. Effects of physiological stage and season on infrared thermograms of different body areas of dairy cows raised under tropical conditions. Animal Reproduction, 16(2):311-316.
• Digiovani, D. B. Marcelo, H. F. B., Victor, H. G. G., Bruna, F. M., Guilherme, M. B., Thiago da R. S., Patrícia, C. F., Flavio, A. B. J., Flavio, G. L., Celso, K. J. and Edson, L. A. R. 2016. Infrared thermography as diagnostic tool for bovine subclinical mastitis detection. Revista Brasileira de Higiene e Sanidade Animal, 10(4): 685 – 692.
• Gloster, J., Ebert, K., Gubbins, S., Bashiruddin, J. and Paton, D. J. 2011. Normal variation in thermal radiated temperature in cattle: Implications for foot-and-mouth disease detection. BMC Veterinary Research,7:73.
• Hovinen, M., Siivonen, J., Taponen, S., Hanninen, L., Pastell, M., Aisla, A. M. and Pyorala, S. 2008. Detection of clinical mastitis with the help of a thermal camera. Journal of Dairy Science, 91: 4592-4598.
• Idris, M., Uddin, J., Sullivan, M., McNeill, D. M., Clive, J. and Phillips. C. 2021. Non-Invasive Physiological Indicators of Heat Stress in Cattle. Animals, 11:71-86.
• Joy, A., Taheri, S., Dunshea, F., Leury, B., DiGiacomo, K., Osei-Amponsah, R., Brodie, G., and Chauhan, S. 2021. Non-invasive measure of heat stress in sheep using machine learning techniques and infrared thermography. Small Ruminant Research, 207:106592.
• Kaur, A., Kaur, P. K., Singh, S., and Singh, I. P. 2014. Antileishmanial compounds from Moringa oleifera Zeitschrift für Naturforschung C, 69: 110-116.
• Khakimov, A. R., Pavkin, D. Y., Yurochka, S. S., Astashev, M. E. and Dovlatov, I. M. 2022. Development of an Algorithm for Rapid Herd Evaluation and Predicting Milk Yield of Mastitis Cows Based on Infrared Thermography. Applied Sciences, 12:6621-6633.
• Kumari, T., Bhakat, C. and Choudhary, R.K. 2018. A Review on Sub Clinical Mastitis in Dairy Cattle, International Journal of Pure and Applied Bioscience, 6(2): 1291-1299.
• Lowe, G., McCane, B., Sutherland, M., Waas, J., Schaefer, A., Cox, N. and Stewart, M.    Automated collection and analysis of infrared thermograms for measuring eye and cheek temperatures in calves. Animals, 10:292-305.
• Marrero, M. G., Chris Rijos-Fernández, Yomar Vélez-Robles, Guillermo Ortiz-Colón, Héctor Sánchez-Rodríguez, Esbal Jiménez-Cabán, Jaime Curbelo-Rodríguez. 2020. Short-milking-tube infrared temperature as a subclinical mastitis detection tool in tropical dairy farms, Applied Animal Science, 36 (3): 329-334
• Metzner, M., Sauter-Louis, C., Seemueller, A., Petzl, W., and Klee. W. 2014. Infrared thermography of the udder surface of dairy cattle: Characteristics, methods, and correlation with rectal temperature. Veterinary Journal, 199:57–62.
• Metzner, M., Sauter-Louis, C., Seemueller, A., Petzl, W., and Klee. W. 2014. Infrared thermography of the udder surface of dairy cattle: Characteristics, methods, and correlation with rectal temperature. Veterinary Journal, 199:57–62.
• 2019. Annual Report. https://www.nddb.coop/sites/default/files/NDDB-AR-2019-ENGLISH-24022020.pdf
• Polat, B., Colak, A., Cengiz, M., Yanmaz, L. E., Oral H., Bastan, A., Kaya, S. and Hayirli, A. 2010.Sensitivity and specificity of infrared thermography in detection of subclinical mastitis in dairy cows. Journal of Dairy Science, 93:3525–3532.
• Redmon, J., Divvala, S., Girshick, R. and Farhadi. A. 2016. You only look once: unified, real-time object detection. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 27–30 June 2016, Las Vegas, NV, USA, pp. 779-788
• Sathiyabarathi, M., Jeyakumar, S., Manimaran, A., Jayaprakash, G., Pushpadass, H. A., Sivaram, M., Ramesha, K. P., Das, D. N., Kataktalware, M. A., Prakash, M. A. and Kumar, R. D. 2016. Infrared thermography: A potential noninvasive tool to monitor udder health status in dairy cows. Veterinary World, 9(10): 1075-1081.
• Sharma, N., Srivastava, A. K., Bacic, G., Jeong, D. K. and Sharma, R. K. 2012. Bovine Mastitis. 1st Edn, Satish Serial Publishing House Delhi, India.
• Sinha, R., Bhakat, M., Mohanty, T. K., Ranjan, A., Kumar, R. and Lone, S. A. 2018. Infrared thermography as non-invasive technique for early detection of mastitis in dairy animals. A review. Asian Journal of Dairy and Food Research, 37(1):1-6.
• Srivastava, A. K. 2015. Mastitis in Dairy Animal: Current Concepts and Future Concerns. Satish Serial Publishing House, Delhi, India. 1-5.
• Uddin, J., McNeill, D. M., Lisle, A.T. and Phillips, C. J. C. 2020. A sampling strategy for the determination of infrared temperature of relevant external body surfaces of dairy cows. International Journal of Biometeorology, 64:1583–1592.
• Uddin, J., Phillips, C. J. C., Goma, A. A., McNeill, D. M. Relationships between infrared temperature and laterality. Applied Animal Behaviour Science, 220:104855.
• Wang, Y., Kang, X., He, Z., Feng, Y. and Liu, G. 2022. Accurate detection of dairy cow mastitis with deep learning technology: a new and comprehensive detection method based on infrared thermal images. Animal, 16 (10): 100646.
• Wilson, D.J., González, R.N., Hertl, J., Schulte, H.F, Bennett, G.J., Schukken, Y.H and Gröhn, Y.T. 2004. Effect of Clinical Mastitis on the Lactation Curve: A Mixed Model Estimation Using Daily Milk Weights. Journal of Dairy Science, 87:2073- 2084.
• Yan, G., Liu, K., Hao, Z., Shi, Z. and Li, H. 2021.The effects of cow-related factors on rectal temperature, respiration rate, and temperature-humidity index thresholds for lactating cows exposed to heat stress. Journal of Thermal Biology, 100:103041.
• Yu, L., Pu, Y., Cen, H., Li, J., Liu, S., Nie, J., Ge, J., Lv, L., Li, Y. And Xu, Y. 2022. A light weight neural network- based method for detecting estrus behaviour in ewes. Agriculture, 12(8):1207.
• Zaninelli, M., Redaelli, V., Luzi, F., Bronzo, V., Mitchell, M., Dell’Orto, V., Bontempo, V., Cattaneo, D. and Savoini, G. 2018. First evaluation of infrared thermography as a tool for the monitoring of udder health status in farms of dairy cows. Sensors, 18: 862.
• Zhang, X., Kang, X., Feng, N. and Liu, 2020. Automatic recognition of dairy cow mastitis from thermal images by a deep learning detector. Computers and Electronics in Agriculture, 178: 105754.

Please follow and like us: