Dairy cattle farming plays a crucial role in livelihood of farmers and are considered as an important source of nutrition and employment in India. At present, India has total cattle population of 192.49 million, out of which female cattle population is 145.12 million (Livestock Census, 2019). The Exotic/Crossbred and Indigenous/Non-descript Cattle population in the country is 50.42 million and 142.11 million respectively. India has total 50 registered breeds of cattle. Due to changing scenario of Indian dairy sector, intensive management is becoming important factor for maximum profitability at large commercial dairy farms. With advancement in the field of information and communication technology (ICT) field, useful algorithms are being developed for fast, precise and accurate decision making of farmers. Linear type traits are the basis and important criterion for classification of best dairy cows. This includes measurement the degree of individual type traits of a cow, which directly or indirectly reflects her milk producing ability (International Committee for Animal Recording, 2018). These traits are heritable and thereby can help improve herd life and production level of a cow (Atkins and Shannon, 2002). The linear type traits are the morphological characteristics of body parts of a dairy cow which are directly or indirectly linked with dairyness of a cow (Dubey et al., 2014). Linear classification is done on the basis of measurement of individual type traits, which narrate/describe the degree of trait. (ICAR, 2018). Use of linear scoring is advantageous because these traits are measured/scored individually, any variation in the trait is easily recognisable and instead of desirability, degree of a trait is measured. The standard linear traits should fulfil the following conditions (ICAR, 2018): linear in biological meaning, single trait and heritable, must have some economic value, possible to measure, there should be variation in the target population and single linear trait able to describe a unique body part of cow which is not usually described by a combination of type traits. There are 18 approved standard traits and 5 common standard traits (ICAR, 2018). Approved standard traits includes: 1. Stature 2. Chest Width 3. Body Depth 4. Angularity 5. Rump Angle 6. Rump Width 7. Rear Legs Set 8. Rear Legs Rear View 9. Foot Angle 10. Fore Udder Attachment 11. Rear Udder Height 12. Central Ligament 13. Udder Depth 14. Front Teat Placement 15. Teat Length 16. Rear Teat Placement 17. Locomotion 18. Body condition score, plus five common standard Traits; 19. Hock development 20. Bone structure 21. Rear udder width 22. Teat thickness 23. Muscularity. Most countries use a scale from 1 to 9 or from 0 to 50 as linear scoring scale. Linear traits give a description of the cow, measured by the eye of a classifier.
Early indication of productivity in dairy animals can be accessed by evaluating their body confirmation and linear type traits as they are closely related to production potential, reproductive performance and fitness of lactating animals (Batanov et al., 2019). Selection of cows based on type traits can improve the milking, health and overall dairy characteristics. Animal’s response to feeding and health management can be evaluated by measuring body conformations which act as quality evaluation criteria (Communodet al., 2013). Therefore, evaluation of linear type traits can act as a tool for indirect measurement of productivity (Zhang et al., 2018).
Appraisal of linear traits by computer vision technique (image analysis technique)
In India, evaluation of body conformation and linear type traits is mostly done by conventional manual methods based on subjective visual appraisal which require expertise, measuring equipment which may physically interfere with animals and thus causes stress to them besides being time consuming and laborious methods. Due to advancement in the field of computer vision technique (image analysis technique) empowered with machine learning algorithms, problems encountered in visual assessment can be reduced. This technique can be used to automatically assess the images of morphological characters (or linear type traits) to classify an animal. Considering the limitations of manual methods and requirement of best dairy animals in country, innovative technique of measuring these traits based on computer-assisted visual image and digital image can be introduced as it is automated, easy, accurate and faster method. It decreases the chances of manual errors, personal biasness and also it is animal friendly (Qian et al., 2008). Recording of body traits with the help of computer vision technique or digital image analysis technique can increase the efficiency of selecting better producing cows, saves times and help in automation in dairy husbandry (Teira et al., 2003). This technique would be useful for small as well as large commercial farms for selection of animals with high production potential, at early age based on linear body conformation traits in a more accurate way. We can also identify different breeds based on their unique linear body traits.
Image processing technique has been applied to various fields like control and guide, pharmacy industry, virtual reality etc. and now it is being introduced to research and production in animals. This technique can be used for dairy animals by developing algorithms for processing of images and extraction of required linear type traits. Other applications of this technique are body condition scoring (Halachmi et al., 2013), diagnosis of lameness through back posture recording in dairy cattle (Viazziet al., 2013); mobility, automated body prediction using 3D vision technique (Hansen et al., 2018).
Some studies have used automated computer assisted assessment and checked the accuracy of these methods by comparing with subjective methods. In foreign countries like China, Russia, USA, UK etc. this technique has been used for monitoring the weight and beef quality in cattle (Teira et al., 2003), determination of pig body dimensions relationship with skeleton composition, size and shape with help of 3D images (Wu et al., 2004). The real cow body measurements and body weight of a cow were taken cow from its photographs with image analysis by software for 18 coordinate points (Tasdemir et al., 2008). A linear appraisal system for dairy cows based on image processing and used vision softwares for image filtration and processing was conducted (Qian et al. (2008) in China and they reported <3 min time per cow for appraisal with 0.9% error rate and 4-point difference between manual appraisal method and computer appraisal method. Salau et al. (2014) studied the possibility of Time-of-Flight-camera-based system for measuring cow body traits based on the software designed for image recording, evaluation and extraction of 13 needed body traits. They found an error rate of 0.2% for ischial tuberosity, 1.5% for base of the tail, 0.1% for rump and backbone dishes, and 2.6% for hips and 0.89 coefficients of determination for each individual animal.
Methodology involved in image processing
Image processing technique is a concept developed on the basis of different positions and anatomical dimensions of body parts of a cow, which are clearly visible on the digital images (Tasdemir et al., 2008). In this technique, the shape, length, width, angle of attachment, different body parts etc., are measured which forms the basis for classification of animals into different categories. The numerical value would be generated by specialised spatial calibrated software after analysis of reference points on the images clicked. Important points to be kept in mind are image quality, image background and contrast (Jimenez et al., 1998–2002) to reduce error chances. The measurement of type traits would be done by measuring the distance between two points or angle of three points for a particular trait by the help of a trained software. The desired system will take the images of different traits as input and expected outcome will be description of animal. The recommended linear type standard traits by International Committee for Animal Recording (2018) for a dairy cow are normally measured by this technique.
Need of image processing technique in India
Indigenous zebu cows like Sahiwal, Gir, Tharparkar etc have smaller and different body size; body conformation compared to Bos taurus cows. Therefore, there is need to work in respect of digital image processing technique for automation of process of evaluating the animal’s linear type traits of dairy cows. As, it is a non-contact method, animal friendly and promotes animal welfare by reducing stress reactions of an animal and increase precision dairy farming practices. This technique might be useful for farmers for selection of elite animals, at early age based on linear type traits. Therefore, a feature learning system from the images of morphological characteristics to support decision of dairy farmers for selecting the best dairy animal need to be developed.
Conclusion
In recent times, there has been an increasing interest in exploring the benefit of so-called precision dairy farming/smart farming. This can be realised by application of sensing technologies, to offer positive interventions in dairy farming. Image processing technique is a concept developed on the basis of different anatomical dimensions of body of a cow, which are clearly visible on the digital images. To increase the efficiency and accuracy of selection, evaluation of animal’s external conformation and linear traits is necessary to improve the lifetime productivity of animals. Therefore, for comprehensive evaluation of animal linear conformation traits, development of a technique like digital image technique based on sensors, algorithms with scientific and practical significance might be needed for indigenous cattle of our country.
Indu Devi, Sumit Mahajan
Scientist, ICAR-CIRC, Meerut, UP