Method and application for estimating the density of broiler flocks based on deep learning

In the intensive poultry breeding process, precise regulation of environmental factors such as temperature and humidity is directly related to the breeding efficiency. Intelligent regulation methods are extremely important for cost reduction and efficiency improvement. The setting values of existing poultry house temperature and humidity controllers are mostly manually determined, which requires enormous labor costs. With the development of information technology, it is becoming easier to obtain breeding data. Starting from large-scale breeding data, using machine learning and other methods to intelligently estimate the optimal temperature and humidity of poultry houses is an effective way to solve the above problems. Considering that there is a strong correlation between temperature and humidity and the spatial distribution of poultry, this article starts with the density estimation of poultry groups, proposes a Transformer-based poultry density estimation method, and selects broilers as the research object. The results obtained from experiments on the broiler density estimation dataset (M2-fowl) demonstrate an average absolute error (MAE) of 20.97 and a mean squared error (MSE) of 33.15 for this method. These findings indicate the algorithm’s capability to provide accurate estimates of poultry density, establishing a groundwork for assessing the comfort levels in poultry houses and implementing intelligent adjustments to temperature and humidity.