Treffer: BROODER HEATER WITH ARTIFICIAL INTELLIGENCE

It is known that ensuring optimal temperature regime of young poultry at the early stage of embryonic development is possible only in a narrow temperature range and instrumental control of temperatures when keeping young poultry under brooder lamps does not provide the necessary accuracy. According to the current recommendations of manufacturers of modern crosses of poultry, the adjustment of the brooder lamp suspension height is based on the data of visual control during which the behavior of young birds is evaluated. The possibility of automating the process of adjusting the height of the suspension of brooder lamps was investigated. A video analytics software and hardware complex based on the YOLOv2 neural network, a PyAi-k210 machine vision platform from the 01studio company based on a Kendryte k210 microcontroller, and an electric linear actuator were used. The neural network was trained on a dataset of 1000 images for each class with a resolution of 224x224 pixels. Maixhub.com cloud service was used for training. Image rotation and reflections were applied for automatic augmentation. The mobilenet_0.5 831Kb model was used. The training duration is 100 epochs. 4 electromagnetic relays were used to control the linear actuator. 2 power electromagnetic relays with two groups of contacts for switching the polarity of the linear actuator power supply and 2 electronic relay modules with galvanic isolation, which were controlled directly from the microcontroller contacts. The program code of the microcontroller, obtained during training of the model, was adjusted to form a control signal to the electronic relay modules. Programming was carried out in MicroPython language. As a result of the study, a test sample of the device was made, which registers the position of young birds in the brooder and automatically changes the height of brooder lamp suspension.