S. O. Bazaev

The article examines the potential of video surveillance systems for contactless monitoring of cow behavior under dairy farm conditions. The aim of the study was to systematize data on the daily time budget of tethered cows and to determine approximate values of the main behavioral acts suitable for automated animal condition monitoring. The study included an analysis of scientific literature and the authors’ own observations of the daily behavior of 254 Black-and-White cows divided into three production groups: lactating cows, first-lactation cows, and young stock. For the comparative analysis, 41 publications containing quantitative data on rest, feeding, drinking, rumination, and locomotor activity were selected.Based on the comparison of published and original data, it was found that the most significant components in the structure of the daily time budget of cows are rest, feeding, rumination, drinking, and locomotor activity. Differences between published and original data were identified for certain activity types, which made it possible to determine approximate ranges of time-budget indicators for different animal groups. Based on the comparison of published and original data, a table of approximate changes in time-budget indicators was compiled; these changes may be considered behavioral signs of animal distress. The obtained results are of interest for the development of automated video monitoring systems and require further verification under clinically confirmed observation conditions.

The aim of the study is to describe the concept of a system for collecting and processing a large dataset from each livestock farm in a regional data center, as well as the possible risks associated with its operation, and to predict the economic effect of implementing the system in the industry. The concept involves collecting data from each sensor and machine regarding health, productivity, conformation, and the quality of all technological processes; further pre-processing of the data on the livestock complex’s server; and transmitting the information to the regional data center for processing and storage, with the aim of subse-quently providing the data to consumers. A table is presented with a description of the data and the methods used to collect it. The data include 88 automatic parameters (87.1 % of the total, such as rumen pH, temperature, and activity, with a reliability of >93 %) and 13 manual parameters (12.9 %, such as the results of genetic tests, with an input error of <8 %), collected using sensors, milking robots, monitoring systems, and manual input. Four stages of system implementation are described, with each stage in-volving an assessment of the factors affecting the final reliability of the information stored on the regional server, along with a description of the actions that increase the reliability of the information. Possible risks and the likelihood of their occurrence during the implementation and operation of the system are described, using an assessment of the measure of uncertainty of the possible states of the system, based on Claude Shannon’s formula for calculating the amount of information in a message about non-equiprobable events. Using the Monte Carlo method to model events associated with an uncertain probability, the entropy distribution of the risk system was estimated based on 10,000 possible scenarios. A model of the relationship between determinants (factors) and risks is presented, expressed as an effect with an assessment of its level of influence on the success of system imple-mentation. The obtained results are superimposed on a theoretically constructed productivity growth curve for a livestock farm, which could potentially implement systems with the four described development scenarios. Under the baseline scenario accounting for risks, the projected milk yield is expected to reach 10,000–11,000 kg per cow per year after three years and 13,000–14,000 kg after ten years, compared with the initial level of 8,100 kg. For a herd of one thousand cows, this corresponds to an additional annual milk output of 1,900–2,900 tonnes after three years and 4,900–5,900 tonnes after ten years.

The purpose of this study was to find the relationship between biochemical parameters of blood of pregnant cows and the development of rickets in newborn calves in order to assess the possibility of prognosis and prevention of this disease.The study was conducted on 32 pregnant zebu‐type Black‐and‐White cows and their offspring (n=30). Calcium‐phosphorus metabolism indicators and vitamin D levels were evaluated in the cows at 50 and 14 days pre‐calving, and in calves at 0, 1, 3, and 6 months of age.The concentration of vitamin D in cows turned out to be critically low (13.43–13.47 ng/ml at a rate of 28–35 ng/ml), and the ratio of Ca: The P did not correspond to the norm (1.36:1 and 1.2:1 with a norm of 1.5–2:1). The calves had vitamin D levels 80 % below normal at birth, and 80 % of the young showed signs of rickets in the first month of life. A strong positive correlation was found between the level of vitamin D in cows 50 and 14 days before calving and its level in calves at birth (r= +0.751 and r= +0.709, respectively). A stable average negative correlation was found between the level of calcium in cows and the activity of alkaline phosphatase in calves. However, the coefficient of determination (R2) for most bonds was low (11–39 %), which indicates a weak influence of the studied parameters.It was found that the indicators of calcium‐phosphorus metabolism in pregnant cows do not have a high prognostic value for reliable prediction of the risk of rickets in offspring. Based on the results, it was considered inappropriate to use these parameters for the prediction of rickets in a contactless animal assessment system.

The study compares two technologies of precision animal husbandry — the SCR Heatime collar (Allflex Livestock Intelligence, Israel) and the “Agrobiotest” bolus (“Agrobiotest” LLC, Russia) — for monitoring the physiological state of dairy cows. The experiment was conducted on 10 Holstein cows at the KubSAU farm. The cicatricial bolus demonstrated high accuracy (97.8%) in measuring the pH of the scar, which is critically important for assessing the acidity and health of the gastrointestinal tract. The SCR Heatime collar (Allflex Livestock Intelligence, Israel) proved to be more effective for tracking rumination (89.4%), which allows monitoring eating behavior. The combined use of the devices increased the overall accuracy of diagnosing the metabolic status of cows to 98.5%, providing a more complete picture of the animals’ condition. A moderate correlation (r = 0.73) was found between the rumination activity and the pH level, which confirms the relationship between eating behavior and metabolic processes. The study confirms that the integration of data from different sensors (collars and boluses) makes it possible to increase the effectiveness of early diagnosis of metabolic disorders and optimize diets. The results obtained open up prospects for the creation of digital animal twins, which contributes to the development of precision animal husbandry.

In modern precision animal husbandry systems, wearable and implantable devices with accelerometers are actively used to monitor the health and productivity of livestock. In this study, a comparative evaluation of two devices was performed — the SCR Heatime collar (Allflex Livestock Intelligence, Israel) and a rumen bolus “Cow Health” (Federal Scientific Agroengineering Center VIM, Russia) — to classify activity and chewing in dairy cows. The experiment was conducted on the basis of a farm of the Kuban State Agrarian University with the participation of Holstein cows for 90 days. The devices recorded data with a frequency of 10 Hz, which was further analyzed using machine learning algorithms to identify key behavioral patterns: feeding, chewing, walking, and resting. The results showed that collars demonstrated higher accuracy (94.2%) in detecting feeding and walking, while boluses surpassed them in monitoring chewing gum (97.8%). The combined use of both devices made it possible to achieve the maximum overall accuracy of behavior classification — 98.1%. Additionally, the effectiveness of the systems in detecting sexual hunting was investigated: the combined use of collars and boluses increased detection accuracy by 12% compared with separate use. The findings highlight the complementary role of wearable and implantable sensors in digital animal husbandry. The combination of technologies not only improves the monitoring of the physiological state of cows, but also opens up prospects for the creation of digital animal twins, which helps optimize herd management and increase the economic efficiency of dairy farms.

Currently, the process of collecting linear parameters of cow exterior is mainly carried out manually, which is a labor-intensive, complex process that depends on the skills of a veterinarian. For fast and accurate grading, it is proposed to develop an intelligent system for contactless digital assessment of the exterior of cattle based on the use of video cameras and modern image analysis technologies. Such monitoring will also allow for the diagnosis of diseases and their early detection.The purpose of the study is to compare two such cameras and determine their applicability in the system being developed.We used a laboratory stand with a large number of objects at a predetermined distance from the cameras, specially developed software for processing the results and constructing a linear regression linking the real distance with the measured one. A special adjustment has been developed to reduce the camera error. As a result, it was determined that both cameras of this type have an average error of ±5 mm when measuring objects at a distance of 382 to 671 mm, taking images with an aperture of up to 60 × 45° and a resolution of up to 480 × 480 pixels. It was concluded that the error of both cameras is acceptable for equipping the system under development, not exceeding 2% per 1 m.

Relevance. Increased values of temperature and relative humidity of the external environment lead to negative consequences for the animal body, forcing the thermoregulation processes to be turned on. These mechanisms allow the animal organism to adapt to new environmental conditions at the expense of productivity. In these cases heat stress is observed. It is established that its manifestation is observed after 17 hours and there is a possible decrease in productivity by 35–40%. To identify the effect of heat stress, accurate determination of temperature-humidity index (THI) is necessary.Methods. The materials and methods contain the most common formulas for determining the temperature and humidity index. The equipment and software package used for the research are presented.Results. A graph of the results of South Korean studies is presented to compare the effects of heat stress on productivity. The results and discussions display a modernized formula for determining the heat stress index and figures showing the level of heat stress at different temperature and relative humidity levels.

To increase the efficiency of livestock farming, scientists are developing information and communication technologies aimed at digitalizing the process of assessing the exterior of animals. This review should improve understanding of the development steps of systems applicable to the digitalization of animal conformation assessment using computer vision and deep learning neural networks. The search focused on several topics: computer vision systems; training datasets; image acquisition systems; deep learning models; neural networks for training; performance parameters and system evaluation. Machine vision is an innovative solution by combining sensors and neural networks, providing a non-contact way to assess livestock conditions as cameras can replace human observation. Two approaches are used to obtain three-dimensional images for digitalization tasks in animal husbandry: shooting animals using one 3D camera fixed in one place, and shooting from different points using several 3D cameras that record images of animals and individual parts of their bodies, such like an udder. The features extracted from the images, called dorsal features, are used as input to the models. The reviewed publications used a variety of deep learning models, including CNN, DNN, R-CNN, and SSD, depending on the task. Similarly, neural networks such as EfficientNet, ShapeNet, DeepLabCut and RefineDet have been mainly used for animal health monitoring, while GoogleNet, AlexNet, NasNet, CapsNet, LeNet and ERFNet are mainly used for identification purposes.

Monitoring the temperature and pH of the contents of the rumen can be useful for assessing the health status and detecting physiological heat in cows. The purpose of the research was to develop methods and means of livestock control in cattle breeding to manage the physiological state of the herd based on mathematical models for detecting the heat, upcoming calving, initial signs of diseases, monitoring the level of feeding and water intake. The work was carried out under the production conditions of farms using existing hardware and software. The measurements were carried out using a non-invasive control method using special sensors-boluses designed to monitor the health of cows. Boluses were placed orally in the rumen of the cows under research. As a result of the research, algorithms and mathematical models were compiled to identify the heat, upcoming calving, diseases, as well as monitoring the level of feeding and water intake. The source data was imported from a standard file format compatible with other applications (csv table). Correlations between the temperature and pH of the rumen, as well as the motor activity of cows were also analyzed. As illustrations, graphs of the main indicators of vital activity, as well as graphs of mutually correlative functions and an illustration of the working console of the program have been given. Table of the results of the program for each cow, average values and standard deviation have been given. A mathematical model is a set of algorithms and calculation results. In order to implement it, a program code was created in the Matlab R2019b software package. This mathematical model can be used to process and interpret data placed in the animal’s rumen of measuring elements (boluses).

Monitoring the temperature and pH of the contents of the rumen can be useful for assessing the state of health and detecting of cows in heat. The purpose of this research was to evaluate the influence of time of day and season of year on the pH of the contents of rumen and its temperature in dairy cows. The research was carried out on 10 clinically healthy fi rst-calf cows of Holstein breed. Cows were milked twice a day at 05:00 and 17:00 they were housed in in a free-stall barn and fed throughout the year at the same ration balanced in accordance with their physiological need. The temperature of the contents and pH of the rumen in cows were measured using special HealthBolus boluses for the care of animals, which were produced at the Federal Research Center Agro Engineering All Russian Institute of Mechanization. The temperature of the rumen began to increase 6 hours after evening feeding and milking, and 1 hour after morning milking it began to decrease. The lowest rumen temperature was observed in spring 38,81±0,001, and the highest in autumn 39,17±0,001. The concentration of hydrogen ions of the rumen (pH) began to decrease 3 hours after morning feeding, whereas 4 hours after evening feeding it began to increase. The lowest rumen pH was observed in the summer 5,99 ± 0,001, and the highest in the autumn-spring period 6,18 ± 0,001. It has been found that the temperature of the rumen in dairy cows has been affected by the time of day and season of year, the pH of the rumen changed similarly to temperature. It was also found that the pH of the rumen has been aff ected by the time of day and season of year.