Abstract:

Animal husbandry is an important part of agriculture. At present, animal husbandry is developing towards large-scale and intensive development, which also increases the difficulty of diagnosis of livestock and poultry diseases. In recent years, in order to improve the level of animal welfare in livestock and poultry breeding, and to reduce the economic losses and public health safety risks caused by animal diseases and health abnormalities in livestock breeding, a number of automated methods for the diagnosis and treatment of livestock and poultry diseases through digital and intelligent means have emerged, such as machine vision analysis, animal audio analysis, infrared temperature perception, deep learning classification, etc. These methods could effectively improve the diagnosis efficiency of diseased or abnormal livestock and poultry, shorten the diagnosis cycle, and reduce the labor force of manual inspection in animal husbandry. The automatic diagnosis and treatment method of livestock and poultry diseases is different from the conventional diagnosis methods based on pathological knowledge, which mainly uses various sensors to automatically obtain various characteristics information of livestock and poultry during the breeding process, such as images, sounds, body temperature, heart rate, and excrement. The collected information is comprehensively analyzed and processed through mathematical models, such as Mel cepstrum coefficient, Logistics regression analysis and intelligent algorithms such as support vector machines and deep learning, and then the animal’s health status is evaluated and predicted. The current research progress of animal disease intelligent diagnosis technology and some basic method principles was summarized from several aspects, such as livestock and poultry morphological diagnosis technology, behavior diagnosis technology, sound diagnosis technology, body temperature diagnosis technology, and other physiological parameter diagnosis technology. Those methods were based on the digital characteristics of animal appearance and body size, behavior and movement, call and sound, body temperature, excrement, respiration and heart rate, the characteristics collected by the sensor, which were analyzed and classified in real time through mathematical models, and the analysis was basically achieved. The current research results on automatic diagnosis and treatment of livestock and poultry diseases were abundant, but most of the related diagnosis methods were carried out in an ideal environment. However, the interference factors in the actual production and breeding environment were very large, and the most of the current diagnostic methods could not eliminate the interference well and accurately extract the required characteristic information. Besides, the current digital livestock disease diagnosis methods were mostly based on the analysis and diagnosis of one kind of livestock feature information, which affected the diagnosis accuracy of the diagnosis system and the diagnosis results were not convincing. At the same time, the most of the current digital diagnosis methods for poultry and livestock diseases still had some problems such as poor diagnosis generalization ability and poor anti-interference ability, which restricted their promotion and application. The focus of future research on automatic diagnosis of livestock and poultry diseases is to improve the accuracy of its sensing algorithms and the applicability and robustness of mathematical models, and to develop an intelligent diagnosis and treatment expert system for livestock and poultry diseases based on multiple feature coupling and data fusion, realize real-time, efficient, intelligent and accurate livestock and poultry health diagnosis.

Key words: disease intelligent diagnosis for livestock and poultry, behavioral diagnosis, physiological diagnosis, sensor monitoring for livestock and poultry