Research On Methods And Devices For Monitoring And Predicting Poultry Body Temperature Under Infrared Thermal Imaging

Body temperature is the most important physiological characteristic of endotherms.Various complications in their pathological state are accompanied by temperature changes,such as the COVID-19 epidemic that has spread around the world since 2020.Therefore,the health status of poultry can be intuitively judged from the changes in body temperature.In recent years,with the rapid development of the livestock and poultry breeding,poultry breeding density and the number has increased dramatically,make the traditional contact-type temperature measurement way can not meet the needs of rapid monitoring of poultry body temperature,at the same time,caused by animal pathogenic microorganisms in the body of most of the epidemic diseases,such as bird flu,H1N1 outbreak also increasingly frequent,This not only brings huge economic impact to the breeding industry,but also has a great impact on human food health.Based on the above reasons,based on the flat green shell layers as the research object to a series of test temperature monitoring,machine learning and neural network technology was used to study the layers under the infrared image characteristic parts recognition model and the process of establishing temperature prediction model,and the infrared thermal imaging technology was adopted to design a kind of poultry non-contact infrared temperature sensors,The non-contact measurement of laying hens’ body temperature was realized,which provided technical support for preventing the outbreak of poultry disease.The research contents of this paper include:（1）The yolo V3-like neural network model was designed to identify characteristic regions of laying hens.The results showed that the mAP of the head region,leg region and body region of laying hens were 98.95%,97.80% and 95.15%,respectively.The mAP of the whole region of laying hens was 98.25%,and the mAP of the whole model was 97.54%.（2）The temperature data of infrared image CSV file were used to extract the temperature values of the corresponding regions through the coordinates of the prediction box of the feature regions.Finally,the change rules of body surface temperature of different feature regions in the health and illness states were calculated.The results showed that: When the infrared temperature of the head region of laying hens is above 41℃ for a long time and the infrared temperature of the body region and the leg region is below 35℃ for a long time,it can be used as a sign of the pathological reaction of laying hens.（3）The temperature distribution range of healthy and sick laying hens was determined by sample division.The results showed that: The incubation period of laying hens temperature distribution between 39-43 ℃,but when its temperature in 39-41 ℃ or 42-43 ℃,laying hens can appear ill trend,when 39 ℃ temperature is less than or greater than 43 ℃,laying hens must be ill,may also appear the situation of death,for the next step to build up the forecast model on the body temperature and temperature measuring system alarm threshold judgment to provide data support.（4）The correlation between infrared temperature,environmental variables and the real body temperature of laying hens was analyzed,and multiple linear regression,ridge regression and BP neural network models were established to predict the real body temperature of laying hens,and the advantages and disadvantages of each prediction model were judged by the average relative error of the test group.The results showed that the R~2 of BP neural network model in the training group was 0.964,and the MRE of BP neural network model in the two test groups was 0.287% and 0.761%,respectively.Meanwhile,the prediction deviation of BP neural network model was within 0.5℃,and its generalization ability was much stronger than regression model and ridge regression model.（5）Overall scheme design of non-contact infrared temperature sensing equipment for poultry.According to the characteristics of layers with many feathers,small size,space layout of breeding farm,thousands of layers and complex environment,the corresponding technical indexes were designed.Temperature measurement system hardware design,using STM32F103 chip as the main control unit,MLX90640 as the infrared temperature measurement module,VL53L0 X as the laser ranging module;Temperature measurement system software design,using Qt to develop the corresponding host computer interface,through the serial port to achieve communication with the host computer,real-time display poultry infrared temperature image related data.