Detection of heat-stressed chickens in poultry house based on deep network and optical flow vectors in the Fourier domain

The productivity and quality of the entire flock are negatively impacted by heat stress in chickens, which can have major repercussions, particularly in crowded farming settings where diseases are easy to spread and hard to control. This study uses deep networks and optical flow to identify heat stress in chickens. The technique focuses on identifying obvious signs of heat stress, such as panting and open-mouth breathing in chickens. There are two phases to the suggested approach: (1) using a deep network to detect open-mouth breathing in chickens; (2) using the Gunnar Farnebck algorithm to compute the optical flow vectors of the wattle, the breathing frequency is estimated in the Fourier domain for the detection of panting chickens. The proposed method was tested on the obtained dataset and demonstrated its ability to recognise heat-stressed chickens in crowded conditions, achieving an overall performance metric of 0.90 by integrating the results of both phases. The two-phase approach, which incorporates the open-mouth breathing behaviour and panting frequency, improves the efficiency and assures robust, reliable heat stress detection.