Abstract:
Objective To address the challenge of automatic temperature detection for large-scale caged broilers, a method combining deep learning and regression analysis was proposed.
Method An infrared thermal imager was used to capture infrared images of broiler chickens, the YOLOv5s deep learning algorithm was used to train the model for the region of interest (broiler chicken head). Multiple linear regression and multiple nonlinear regression were respectively introduced to establish prediction models between the temperature of the broiler’s region of interest and the temperature under its wings, ultimately achieving the goal of automatic body temperature detection.
Result The test results showed that the precision and recall of the region of interest model based on deep learning were 93.8% and 95.8% respectively. The average relative errors of the multiple linear regression 1 and multiple nonlinear regression temperature prediction models were 0.28% and 0.27% respectively, and the maximum differences between the predicted and actual temperature values were 0.34 and 0.32 ℃ respectively.
Conclusion The nonlinear model has a higher accuracy rate in predicting the body temperature of broilers, providing technical support and a preliminary research basis for automated in-house temperature inspection equipment for chicken farming.
下载: