Sow Body Temperature Detection And Detection System Design Based On Infrared Thermal Imaging

Pig body temperature is a key physiological characteristics for assessing its health status and preventing disease.The traditional contact temperature measurement by mercury thermometers requires much effort and will cause stress to pigs,which is difficult to meet the needs of large-scale breeding.Therefore,this paper studies the non-contact temperature measurement method of pig body temperature based on infrared thermal imaging technology.The main contents are as follows:(1)Collection and arrangement of test data.The experiment was carried out from July20,2021 to September 15,2021.The body temperature,infrared images of 11 body surface parts including eyes,ear bases,neck,shoulders,front back,back,buttocks,tail bases,vulva,buttocks,and abdomen and the environmental factors around the sows(temperature,humidity and wind speed)of 108 sows were collected,as well as pig type information(breed,heat information,gestation time,collection time).The data management software was developed,the database table structure was designed for the experiment data,and the collected experiment data,including sow body temperature,body surface temperature information,environmental factors and pig type information,were stored in the My SQL database for management.(2)Data preprocessing and feature analysis.Standardized data preprocessing was carried out on the experimental data,and in order to improve the precision of the model,the categorical variables pig breed,gestation time,estrus,and collection time were coded by one-hot encoding as potential useful factor variables of the model.The correlation between rectal temperature and different temperature types of body surface was analyzed.It was found that the correlation between maximum temperature and rectal temperature was the highest(r=0.44-0.59).The correlation between rectal temperature,maximum body surface temperature and environmental factors(temperature,humidity,wind speed)was analyzed.It was found that when the temperature,humidity and wind speed of the swine house was relatively stable,the influence of environmental factors on rectal temperature was not significant(P>0.05).There was no significant correlation between body surface temperature and ambient temperature,humidity(P>0.05),but there was a significant correlation with wind speed(r=-0.26～-0.38).(3)The regression model of rectal temperature was established and the results were compared and analyzed.The data set was divided by 5×4 nested cross validation,and the support vector machine(SVR),ridge regression(RR),random forest(RF)model and least squares support vector machine(LSSVR)model were established with the surface temperature and environmental factors as the feature and rectal temperature as the label.The results show that the LSSVR model has high accuracy and generalization,the model R~2 is 0.639,RMSE,MAE are 0.133℃,0.110℃respectively.The models established by the maximum temperature,average temperature and maximum number temperature of different temperature types were compared,and the results show that the maximum temperature is the optimal parameter to describe the temperature of the body surface.Four possible factors are added to improve the accuracy of the model,including pig breed,gestation time,estrus and collection time.The results showed that the model R~2 was increased by 4%,8%,and 10%,respectively,except for the pig breed,which did not improve the accuracy of the model.The final model R~2was 0.773,and the RMSE and MAE were 0.106°C and 0.09°C,respectively.At the same time,compared with previous studies,the experiment added two body surface parts,abdomen and tail root.The results showed that these two parts also increased the model R~2by 7%.(4)The temperature detection system was designed and implemented.Based on the body temperature prediction model of least squares support vector machine(LSSVR),a visual body temperature detection system is designed and implemented with QT technology.The registration of infrared equipment,the reading and real-time display of infrared video data flow,the preservation of infrared images and videos in different formats,and the prediction of pig body temperature were completed.It is convenient to realize the automatic monitoring of pig body temperature and provide reference for the healthy breeding of pigs.