Abstract
This paper will discuss how facial recognition can help to revolutionize farming. Also, how the way the multi-UAV framework works. An onlooker ramble floats 90-270 feet over the group. This automaton utilizes descending confronting sound system cameras to follow the movement. It decides the area and direction of the cows. Three laborer rambles take the area data given by the spectator automaton and use it to follow a particular bovine. The laborer rambles at that point to play out the wellbeing checking errands. To test self-governing automatons in synchronized flight, the group set up a dry run community in a cellar lab of the mechanical designing structure. Here, tall roofs permit cameras roosted on the dividers to go about as the spectator ramble, utilizing retroreflective markers to triangulate the situation of the automatons and dairy animals in the space. Programming run by a close-by PC takes that data and gives the specialist rambles arrange and flight directions comparative with where the cow is. There are no genuine bovines in the lab, however, there is a dairy animals model.
Introduction
This restrictive programming will depend on consideration regarding pictures to screen the two appearances and creature covers up. Their innovation, Cargill and Cainthus’ fight, will make it conceivable to utilize PC imaging programming to screen nourishment and water admission, internal heat level, resting and sitting time, and ecological conditions so as to monitor creature wellbeing and government assistance.
Another issue that faces the cultivating business all the more extensively is work deficiencies. The California Farming Bureau detailed that 56% of ranchers can’t recruit enough laborers to stay aware of their requests. Numerous employments are brief, regardless of whether they are agreement or regular, to stay aware of change underway. While trying to convince more specialists to join the cultivating business, bosses are offering all-day work regardless of whether it isn’t required. What’s more, almost 50% of all homestead laborers are not legitimately approved to work in the nation, making representatives transient. This has brought about numerous ranchers going to innovation and motorization to fill holes in labor. Facial acknowledgment assists ranchers with handling both of these issues.
Background
It should be recognized that the modern agricultural sector has undergone significant changes compared to the industry in which people worked a few decades ago. At the end of the twentieth century, the farmer is usually an employee of a small ranch caring for cattle, where the stock of animals is hardly extremely large if it is not big meat processing plants. However, coupled with the general increase in the world population, the availability of meat products became higher, which led to an intensification of needs: demand for meat products increased. This fact can be seen in the analysis of comparative statistics on the dynamics of the total number of cattle, both individually on the continents and in total. In particular, Figure 1 shows that over half a century, the number of cattle increased by 43% for Asia, by 190% for Africa, by 84% for North and South America, by 52% for Australia, and decreased by 38% for Europe. Without stopping to discuss the decline in the number of cattle in Europe, it should be noted that the general trend of population growth is difficult to notice: the total change in cattle population over this period was +331%.
The increase in the number of cattle certainly has an impact on the productivity and quality of the agricultural sector. First of all, it concerns the employees’ professional experience: while personal and individual care used to be the priority early, nowadays, the specialists should mainly adjust computer equipment and monitor the proper functioning of intelligent systems (Plug and Play Tech Center, 2019). In other words, a cow as an individual animal has lost the level of care that it used to receive and instead began to play a role in a large chain of farm production.
There is no doubt that the welfare of livestock should be of paramount importance for agronomists. Even though eventually the animal will be killed for the sake of food resources, it must be in a safe and healthy condition throughout its life. This concept has served as a moral basis for the creation of an ethical principle of the new welfarism that protects the rights of animals (Taylor, 2018). Indeed, the consumer wants a delicious and quality meat product, and it is known that the absence of stress in the life of an animal directly affects the structure of the meat. The general determinants of livestock welfare are the lack of transportation, the friendly attitude of workers, the quality of food and drinking water, fresh air, and long walks (Tarantola et al., 2020; Cozzi et al., 2016). Deprivation of cattle any of a fundamental need can be fatal, so it is essential to provide the animal with all the necessary resources and to examine its quality of life regularly.
Unfortunately, animal welfare guidelines have been ignored by some suppliers. As a result of inadequate management policies, livestock is grown in unsanitary conditions, fed on inappropriate feed without vital nutrients, and suffers from regular safety and health problems. While there are enough practical examples of such scenarios, the case of a Canadian farmer from Ontario deserves special attention because his indifference and irresponsibility killed more than 1,500 pigs (“Ontario man pleads guilty,” 2018). The farmer’s animals lived in an unhygienic environment without proper ventilation and care for their health, which eventually led to premature mass death. While this story is an example of the outrageous and unacceptable attitude of the farm owner, it also shows the reality of the most unimaginable scenarios in the agricultural sector. In addition, these facts only confirm the need to develop a monitoring system to ensure regular care for the welfare of livestock.
The search for tools to monitor livestock 24 hours a day, seven days a week, has attracted a great deal of research, each of which offers its developments as a solution to a practical problem. For instance, the widely known difficulty associated with grazing cows on pastures is the loss of the animal as a result of uncontrolled actions by livestock. Watching cows seems particularly difficult if the pasture is not on a plain with clearly visible species but in close proximity to forests, ravines, and marshlands (Cheng, 2019). Increasing the number of simultaneously walked cows directly affects the possibility of their regular identification: the more objects to be observed, the harder it is for a shepherd to cope with this task. The solution to this problem lies in the use of GPS technology: with the help of small elements modified by the tracking function — such as collars, chips, or bracelets — the owner of the farm can find geodata of the lost animal and take timely rescue measures (Schieltz et al., 2017). It should be noted that while this technology is relatively cheap and easy to use, it requires regular battery replacement and can cause discomfort to the animal.
An alternative technology solution for the agricultural sector could be the introduction of feeding management systems. The issues of nutritional adequacy and monitoring of the fullness of the diet are central predictors of an intelligent approach to livestock welfare. Inadequate use of resources can lead not only to problems with farm animal health but also to economic damage to the farm. In order to neutralize the undesirable effects, feeding management systems can ensure that every meal is accounted for, and workers are notified if an animal is undernourished. Specific algorithms for creating such programs can be the use of sound signals received from each animal as a result of jaw muscles (Rau et al., 2020). However, such models’ practical use may seem unlikely if we consider a large-scale production replete with different sounds and audio signals.
In general, it should be clarified that technological models for managing livestock life are not limited to the observation of food and control of the animal’s presence on the farm. On the contrary, the variety of practical tasks has given rise to numerous models that are used for monitoring. Some examples of already mentioned management programs may include breeding control systems, sleep analysis, care of animals with diseases, including control of timely medication intake (Van Den Pol-Van Dasselaar et al., 2017). At the same time, through a database that is updated periodically, farm owners have constant access to complete information about an individual animal, including date of birth, breed, susceptibility to disease, behavioral and nutritional characteristics, and maturation times.
Although the advantage of such programs for modern agricultural production is evident, as interaction with cattle is simplified, it is necessary to note the difficulties that may hinder the successful use of monitoring systems. First of all, it concerns practical skills: not all farmworkers have professional computer experience, and the need to regularly access the database, tracking the regularity of livestock development can be a problem for employees. In addition, it should be clearly understood that agricultural activities in themselves are a multi-component and complex industry where employees are expected to be multitasking. Animals must receive daily medical care, eat and sleep on time, and get fresh and clean air: to perform all these tasks, farmworkers must have the multi-potential to keep everything under control. Adding a dozen different functions to the range of responsibilities expressed in the variety of monitoring systems will only make the farm more complicated and will not bring any positive results for either the animal or the human.
The expected solution, in this case, is to use an integrated program to monitor all the above described livestock conditions. The system should identify any changes and deviations from the animals’ usual behavior and decipher and record them in a visually friendly form. For example, a caregiver can use the application to evaluate all the parameters for an individual animal or the entire herd at once. Moreover, in case of any emergency — whether it is an exit from the farm territory, an unexpected mating, or loss of consciousness — the unified monitoring system urgently notifies the owner about it. Such a solution should not seem fantastic to the reader since it already exists. Digital identification of the person becomes available to farm owners who want to take care of the welfare of animals.
Details and Description
Modern systems for tracking and identification of biometric user data are increasingly being introduced into the everyday life of society. Voice recognition, fingerprint recognition, face, or retina scanners are already widely known technologies for using smartphones, bank services, or for accessing other confidential data. The transfer of some of these technologies to the agricultural industry has a high potential to control animal behavior and respond to changes in a timely manner.
Face recognition, discussed in detail in this project, is a simple technology available even for use with a cell phone. According to Bora (2020), Indian entrepreneurs have already trained the program to distinguish faces of animals with high accuracy, which only confirms the predictions for the rapid practical implementation of mass face recognition systems throughout the farm. Given the diversity of everyday farming tasks and the frequent combination of different animal species on the same ranch — it can be pigs, cows, sheep, or chickens — it is important to note that the technology can be used to solve multiple tasks in parallel.
For example, for some animals, the biometric data obtained is needed to support conservation and well-being efforts. This applies not only to monitoring their behavior but also tracking the vital functions of the body: temperature, quality of sleep, nutrition. If a cow, for some reason, is not recognized by other animals in the cattle, it can be evaluated with the help of video surveillance systems. Thus, the animal will receive the necessary care and attention, and thus the problem of well-being will be solved.
In addition, location data can be useful for finding animals and catching poachers. If, during grazing, a cow, horse, or sheep fought off the herd and did not return to the stall, this can be clarified with the help of computer vision skills. When the system detects the absence of the animal in its place, it will inform the owner of the farm about the loss and connect the rest of the face sensors available throughout the ranch. So, for example, having established the last location of the cow before the loss of the signal, the system can predict its probable route. If this area is known to be potentially threatening to livestock health and safety due to poachers, the system will let the user know.
Installation of recognition sensors on the territory of the whole ranch and in the most frequently visited places, in particular, has a high practical potential. Using this information, the researchers can monitor food and water consumption, determine body temperature and deviations in behavior. If an animal, for whatever reason, refuses to eat and consumes extreme amounts of water, this may indicate a health problem. For this reason, the following use of facial recognition systems can be expressed in medical care. Animals can also be monitored to obtain new data about them or to predict diseases they may suffer in the future.
It is worth noting another area of use of real-time monitoring systems that simplifies the work of the agricultural sector, namely, crop production. Although this work is focused on facial recognition of animals, vegetation analysis embodied through computer vision can perform similar functions. Plant breeders use computer vision to recognize diseases of crops: both at the micro level, by close-up images of leaves and plants, and at the macro level, revealing early signs of plant diseases or pest spread according to aerial photographs. Of course, photography is not the best source of data, and many essential aspects of plant life are better studied in other ways, such as using spectroscopic photometers, video surveillance, or in person. Often the condition of plants can be better understood, for example, by collecting hyperspectral images with special sensors or by performing laser scanning: such methods are increasingly used in agronomy (Wang et al., 2019). These kinds of data are usually of high resolution and are inherently closer to medical images than photographs; one of the planting surveillance systems is called AgMRI (Saik, 2018). Although special models are needed to process these data, their spatial structure allows for the application of modern computer vision technologies, in particular convolution neural networks.
Methodology
There is no doubt about the success of the programs described: it can bring the company more confident management and peace of mind for the safety of livestock. In this case, it is essential to discuss the methodology of implementing facial recognition systems on the farm. The primary task in designing the monitoring ecosystem is to address the issue of data relevance. It is evident that no one on the farm will look through hundreds of hours of video footage to track any remarkable trends in animal behavior. In that case, the solution to this problem should be to use machine learning.
It should be recalled that machine learning is traditionally understood as training a mathematical model on historical data in order to predict a specific event or phenomenon already on new data. In other words, this is an attempt to force the algorithms of programs to perform actions based on previous experience, not only on the basis of existing data (Beam & Kohane, 2018). Machine learning also includes a whole set of methods and algorithms, which can predict some results from the input data. For example, a farmer has information about the incidence of viral infection affecting reproductive systems in some animals, while machine learning algorithms can predict which of the data descendants of the entire herd will be most vulnerable. Although machine learning is already used in crop production, its widespread use as an application for animal welfare has not yet been discussed (Chen et al., 2019). In this regard, the use of computer algorithms seems to be the most appropriate task.
Video control systems, made as small video cameras, are installed throughout the farm to maximize coverage of land, premises, and surrounding borders. Thus, the main task of placing the cameras is to maximize the angle of view. The pre-installation step should be the training phase of the system: one way to implement this idea is to upload several thousand photos of animals into a database. In particular, the computer should be able to analyze the animal by face, but the torso shape, spot size, tail, or growth defects can become reference points for more accurate and consistent identification. Uploading photo data to the database with the animal identification code assignment is a long, unnecessary job for the program to learn to recognize the animal. In general, many companies report on the successful application of these models: the chance of accurately identifying an animal by image only is more than 95 percent (Kumar et al., 2016; Yao et al., 2019). On the one hand, this indicates the high reliability of such systems, but on the other hand, the need for further development of recognition algorithms.
The final implementation of the developed ideas includes video cameras with facial recognition sensors, an image decoding system, and a handy database application for interaction with animals. The daily scenario of such a system may be that the farmworkers in the morning explore these applications and assess the livestock dynamics. If a representative has a problem, insomnia, or a change in behavior patterns during the last night, the specialist is notified. Throughout the day, the application allows to monitor the quality of nutrition control, reminds one of the need to take drugs or other medical procedures. Facial recognition systems are crucial during the grazing of cattle. If the cameras are located high enough to monitor the herd from above, then it will allow the computer to assess trends in the movement of animals, to determine the “leader” of the herd and lagging animals. In the evening, when the workers go home, facial recognition systems monitor the animals to inform the employee about the detected problems in the morning. If in the night period, the animal needs emergency help, the application will give the signal.
Practical Analysis of Facial Recognition
Face recognition is a fully developed and logical system that combines the functions of individual monitoring programs. It should be admitted that this technology is not new, and by 2020 there are several large companies actively developing similar products. This section will provide some information about the result of the implementation of such systems and some nuances that should be taken into account when designing.
Perhaps the leader in the field of facial recognition systems research is the Irish company Cainthus, which studies herds of cows with the help of computer vision skills. Identification, which lasts only a few seconds, is used as part of an algorithm controlled by artificial intelligence and is ultimately passed on to the farmer. Cainthus strives to create a culture of well-being for animals and simplify economic farm management models, so it is natural to collect and analyze data such as water and feed supply and consumption, tracking behavior, mating times, and the amount of product output — whether milk, eggs or wool — is natural. It is fair to admit that the Irish company has reached individual peaks: Cainthus has tested its own results on fifteen dairy farms in Europe and Canada and recently focused on four factories in California (Cooke, 2020). The progress was significant as a large American company with a well-known brand history invested in the development of Cainthus, encouraging research (ModVegan, 2018). It follows that the practical application of facial recognition systems for animals is already a conceptual part of agriculture and can be expected to intensify in the future.
An essential part of the project implementation is the early discussion of some practical problems that may be faced by agricultural producers. In particular, CCTV cameras are sensitive electronic devices that are not resistant to natural forces. Devices installed outdoors, across pastures, may be exposed to rain, strong winds, or dust storms. In order to maintain the operability of the cameras and to extend their service life, it is recommended to place the device in an airtight container that does not prevent significant face recognition.
At the same time, an essential part of the analysis for the animals occurs at night, when most of the farmworkers are absent, and the livestock is alone. For successful surveillance during this period, the camera must have a night vision function that allows you to track all the same identification parameters as during the day. Indeed, such functionality technically complicates the cameras and increases their cost, but it must be recognized that the initial investment must be sufficient to ensure a smooth operation of the farm.
Advantages and Disadvantages
As with any other technology, the animal face recognition system being implemented for the agricultural industry has its pros and cons. The apparent advantage of the system is that it simplifies the interaction with animals. Employees more need to keep in mind all the details of animal behavior or remember which of the cows was worse fed or drank more water: the computerized database leaves employees with tasks to perform specific tasks in strict sequence. With face recognition, the farmers will be able to access information about each animal, including biographical notes, medical data, and economic justification. As previously described, thanks to these technologies, cattle will no longer be lost outside the grazing area, and any medical or mental deviations will be solved in an emergency. Moreover, an adequate personnel management policy will be a severe economic advantage, which includes the evaluation of the necessary and sufficient number of employees, quality control of the performance of their professional duties.
Nevertheless, there are a number of serious problems that need to be addressed in order to successfully implement the system. Ongoing developments do not address the issue of the legal protection of such information and do not take into account the ethics of the round-the-clock monitoring of animals. On the other hand, intelligent systems are vulnerable, which threatens the cyber-security of computer surveillance and facial recognition systems. Moreover, the system will remain imperfect, and even a high chance of correct recognition of an animal does not guarantee absolute correctness. As a result, the system may confuse the information of individual animals and cause a false alarm. In addition, the facial recognition system requires a significant financial investment in the initial stage and periodic monitoring of the performance of individual cameras and sensors.
Future Research
Future research in this technology industry should address the challenges that are currently disadvantages of animal facial recognition systems. Moreover, in the future, the system must be updated and modified as the demands of farmers will increase. In particular, it is proposed to evaluate the feasibility of using cameras for all animals at once and the resource intensity of the databases. It is crucial to understand if the combination of faces of different animals will not become an obstacle for the devices. The likely development of this technology may be aimed at crop production, which also needs regular monitoring and control.
Conclusion
To sum up, it should be noted that the agricultural sector is now experiencing incredible technological progress. The work of farmers is already noticeably different from that which was relevant for agricultural production ten years or more ago. One of the innovative solutions for the farm is the installation of animal face recognition systems, which is designed to solve several problems at once. Here are just a few of them: (i) animal welfare, (ii) research on trends in herd movement, (iii) identification of potential role models among the herd, (iv) medical control of the animal, (v) tracking of food and drink cycles. Although the proposed solution in this paper is conceptual, some companies, such as Cainthus, have already successfully integrated video surveillance systems for dairy cows. In the near future, it can be expected that these technologies will become a natural part of life in the agricultural sector and that scientific progress will have an even more significant impact on the work of farms.
References
Beam, A. L., & Kohane, I. S. (2018). Big data and machine learning in health care. Jama, 319(13), 1317-1318.
Bora, G. (2020). Facial recognition for cows: Why is Mooofarm giving cattle a digital identity? The Economic Times. Web.
Chen, C. H., Kung, H. Y., & Hwang, F. J. (2019). Deep learning techniques for agronomy applications. Agronomy, 9(3), 142-143. Web.
Cheng, A. (2019). Good and bad pasture management (examples). Shortform. Web.
Cooke, L. (2020). Cainthus uses artificial intelligence to watch cows 24/7. Darigold. Web.
Cozzi, G., Brscic, M., & Gottardo, F. (2016). Main critical factors affecting the welfare of beef cattle and veal calves raised under intensive rearing systems in Italy: A review [PDF document]. Web.
Kumar, S., Tiwari, S., & Singh, S. K. (2016). Face recognition of cattle: Can it be done? Proceedings of the National Academy of Sciences, India Section A: Physical Sciences, 86(2), 137-148.
ModVegan. (2018). Facial recognition for cows: Cargill invests in Cainthus, an Irish predictive imaging startup. Modern Vegan Living. Web.
Number of cattle, 1961 to 2018. (n.d.). Our World in Data. Web.
Plug and Play Tech Center. (2019). The cuture of AgTech: Cainthus [Video]. YouTube. Web.
O’Brien, C. (2019). Cainthus recognise need to strengthen with Kickert appointment. The Irish Time. Web.
Ontario man pleads guilty to animal cruelty in deaths of 1,500 pigs. (2018). Global News. Web.
Rau, L. M., Chelotti, J. O., Vanrell, S. R., & Giovanini, L. L. (2020). Developments on real-time monitoring of grazing cattle feeding behavior using sound [PDF document]. Web.
Saik, R. (2018). Monthly archives: August 2018. The Agriculture Manifesto. Web.
Schieltz, J. M., Okanga, S., Allan, B. F., & Rubenstein, D. I. (2017). GPS tracking cattle as a monitoring tool for conservation and management. African Journal of Range & Forage Science, 34(3), 173-177.
Tarantola, M., Biasato, I., Biasibetti, E., Biagini, D., Capra, P., Guarda, F.,… & Mioletti, S. (2020).
Beef cattle welfare assessment: Use of resource and animal-based indicators, blood parameters and hair 20β-dihydrocortisol. Italian Journal of Animal Science, 19(1), 341-350. Web.
Taylor, N. (2018). Whither rights? Animal rights and the rise of new welfarism [PDF document]. Web.
Van Den Pol-Van Dasselaar, A., De Vliegher, A., Hennessy, D., & Isselstein, J. (2017). Grazing in a high-tech world: Proceedings 5th meeting EGF Working Group” Grazing” in Trondheim [PDF document]. Web.
Wang, Y., Wen, W., Wu, S., Wang, C., Yu, Z., Guo, X., & Zhao, C. (2019). Maize plant phenotyping: Comparing 3D laser scanning, multi-view stereo reconstruction, and 3D digitizing estimates. Remote Sensing, 11(1), 63-69. Web.
Yao, L., Hu, Z., Liu, C., Liu, H., Kuang, Y., & Gao, Y. (2019). Cow face detection and recognition based on automatic feature extraction algorithm [PDF document]. Web.