Design And Implementation Of Dead Broiler Identification System Based On Infrared Thermal Imaging Technology

With the continuous improvement of residents’ consumption concept and consumption level,the market demand for broilers is also increasing..The quality of broiler broilers is closely related to the health of the national diet.In the current standardized broiler production,a large number of sick and dead broilers occur every day.The realization of automatic identification of dead broilers can reduce the risk of bacterial infection caused by live broilers pecking dead broilers,ensure the safe production of broilers and reduce the work intensity of management staff.At present,the identification of dead broilers mainly depends on regular inspections by the housekeepers.This method is inefficient and long-term exposure to harsh environments is extremely detrimental to the physical and mental health of a breeder.Based on this,an automatic identification system for dead broilers has been developed.Achieve uninterrupted round-the-clock inspection of dead broilers in order to reduce the workload of broiler administrators and large-scale spread of epidemic situation in broiler houses.In this article,we take the broilers of New Hope White Feather Broiler Farm in Pingdu City,Shandong Province as the research object,and use the thermal infrared data acquisition system to collect the real-time thermal infrared images of the flock broilers,and wirelessly transfer the collected thermal infrared images of the flock broilers to the cloud server.After the received image is processed by a dead broilers recognition model based on a convolutional neural network,the dead broilers on the broilers farm can be identified in real time,and the recognition result is displayed on the visualized Android application.The main research contents of this paper are as follows.(1)Construction of broiler thermal infrared acquisition system.The navigation system was re-developed using a control algorithm with fuzzy preference behavior.positioning and path planning of the walking module were realized by the Monte Carlo algorithm and the A*algorithm.The thermal infrared acquisition module was used to collect real-time thermal infrared images of broilers.The thermal infrared images were transmitted to the cloud server through a wireless transmission module composed of a wireless bridge and a network switch.(2)Study the temperature and image changes of different dead broilers.The study compares the temperature changes of dead broilers at different ages,establishes post-mortem temperature models of different-day-old broilers,and builds thermal infrared data sets of dead broilers based on the temperature changes of different death gestures and temperature changes in different death attitudes.(3)Research on dead broilers recognition models based on YOLOv3 and FasterR-cnn algorithms.Compare the recognition effects of the YOLOv3 dead broilers detection model and FasterR-CNN dead broilers model to establish a dead broilers recognition model.Tests show that the recall and precision of the dead broilers recognition model are more than 98%,and the detection speed reached 8.8 fps.Compared with the previously proposed method,the algorithm has significantly improved the recognition.(4)Implementation of software platform for dead broilers identification.Develop a dead broilers recognition Android client on the Android Studio platform with Java as the development language.After the broiler thermal infrared data is transmitted to the server,the middleware receives,parses,and processes the data,and then stores the processing results in the MySQL database.The administrator can view the data in the database management system in real time through the Android terminal and take the corresponding action measures to reduce economic losses in broiler farms.