Analysis of the physiological state dynamics in dairy cows based on video monitoring

In modern conditions in agriculturer, there is a growing need to develop conceptually new, effective technologies for collecting and analyzing information that provide operational monitoring of the health and physiological state of animals. Most cow diseases can be detected and prevented in the early stages by carefully recording and promptly responding to “cow’s signals”. It is necessary to have more advanced methods and intelligent video monitoring systems for the health and physiological state of highly productive cows at large dairy complexes. These systems must be economically feasible and provide the required increase in the efficiency of livestock farming with minimal costs for monitoring and processing video information. The lack of precise methods for substantiating the requirements for intelligent video monitoring systems for the health and physiological state of cows on dairy farms entails the risks of unjustified expenditure of funds and failure to achieve the goals pursued, which slows down the flow of investment in their development and implementation. To eliminate such risks, the article proposes a method for justifying the requirements for such systems, based on the developed Markov model of the life of a dairy herd and assessing the efficiency of production. Sixteen states of this process are identified, including states associated with diseases of cows: 1 - ranking; 2 - failure to inseminate healthy cows in the stabilization phase of lactation (extended service period); 3 - stabilization of lactation of healthy cows (hunting, insemination, first stage of pregnancy); 4 - decline in lactation of healthy cows (intensive fetal growth and decreased milk productivity; 5 - dry state of healthy cows (not milked, in the start-up); 6 - transit period before calving of healthy cows (in the maternity ward); 7 - calving and post-calving period of healthy cows; 8 - milking of healthy cows (increase in milk production, restoration of health after calving, preparation for insemination); 9 - non-insemination of sick cows in the phase of stabilization of lactation; 10 - stabilization of lactation of sick cows; 11 - decline in lactation of sick cows; 12 - dry condition of sick cows; 13 - transit period before calving of sick cows; 14 - calving and post-calving period of sick cows; 15 - milking of sick cows; 16 - forced culling of sick cows. The model proposed in the method can also be applied independently to analyze the dynamics of the physiological state of cows under various conditions and predict possible events. Using this model, analytical dependencies were obtained that link the income from productive cows with their physiological states and with the probabilities of recognizing the signs and diseases of animals by the intelligent video monitoring system. The dependencies are realized through the intensities of therapeutic and preventive transitions of cows from one state to another as functions of the parameters of development of various diseases and measures for their timely detection, prevention and treatment. It is shown that, given the desired income from productive cows, using such dependencies, it is possible to successfully justify the requirements for the accuracy and timeliness of solving video monitoring problems. Graphs are given that reflect the change in the integral efficiency of the dairy herd from various capabilities of the video monitoring system for the physiological state of cows in states 2, 3, 4, 8 and 9, 10, 11, 15. Proposals for the composition and structure of such a system are formulated, and possible options for its configuration are reflected. It has been numerically confirmed that the main direction for increasing the economic efficiency of dairy herds through the introduction of intelligent video monitoring systems is to increase the likelihood of timely recognition of signs of cow diseases and carrying out preventive measures to prevent them. This ensures not only a high level of health of the dairy herd, but also minimizes the costs of video monitoring itself. The c model does not contradict the objective laws of the productive life of cows in the herd. For the qualitative use of the model, it is planned to rely on the exact values of the intensities of cow transitions from one cow state to another. For this, it is planned to further analyze the accumulated data in the Leningrad region. In addition, along with solving the inverse analysis problem using the proposed model, we plan to predict the conditions of cows depending on the activities carried.