An AI-enabled radiometric ocular thermography framework for monitoring heat vulnerability and welfare in poultry

Heat is a silent but relentless adversary in the world of poultry farming. When chickens are subjected to prolonged high temperatures, their bodies struggle to cope, leading to reduced growth, lower egg production, and in severe cases, death. For farmers, the challenge lies in detecting this distress early, often before the birds show obvious signs of illness. The eyes offer a crucial clue. Because the temperature of the eye correlates closely with a bird's core body temperature, monitoring the eyes can serve as an early warning system for heat stress. However, checking the temperature of individual birds in a crowded barn is a daunting task. Doing so manually requires stopping the flock, handling each bird, and taking a reading, a process that is slow, stressful for the animals, and impossible to scale for thousands of birds.

To solve this, researchers have turned to a combination of thermal imaging and artificial intelligence. Thermal cameras can see the heat radiating from an animal's skin without ever touching it, creating a picture where warmer areas appear brighter. Yet, a camera alone cannot distinguish which part of a bird is the eye, nor can it automatically calculate the temperature of that specific spot from a chaotic scene filled with moving birds. This is where artificial intelligence steps in, acting as a digital assistant that can learn to recognize specific body parts and extract precise data from complex images. The goal is to create a system that allows farmers to simply walk through a barn, capture images, and instantly receive a report on the thermal health of their flock.

In a recent study conducted in India, an independent researcher developed and tested such a system specifically for poultry. The work focused on creating a complete framework that could automatically find chicken eyes in thermal images and measure their temperature, all while running on a mobile device. The study took place during a hot summer month in two poultry houses in Odisha. One house held adult birds, roughly twenty-six weeks old, while the adjacent house held younger juveniles, about thirteen weeks old. The researcher walked through the barns with a handheld thermal camera connected to a smartphone, capturing images of the birds as they moved about. The camera recorded both a standard visible-light photo and a raw thermal image, which contains the actual temperature data for every pixel in the frame.

The core of the project involved teaching a computer program to recognize and isolate the eyes of the chickens. The researcher first manually marked the eyes in a selection of images using specialized software, creating a set of examples for the computer to learn from. These examples were used to train a type of artificial intelligence model known as a segmentation network. Unlike a model that simply draws a box around a bird, this system was trained to draw a precise outline around just the eye, separating it from the rest of the bird's head and the surrounding environment. The researchers trained two separate versions of this model: one specifically for adult birds and another for the younger juveniles, as the size and appearance of the eyes differ between the two age groups.

Once the models were trained, they were put to the test on the hundreds of images captured in the barns. The system successfully identified and outlined the eyes of 123 adult birds and 128 juvenile birds. For each identified eye, the software extracted the temperature data from the corresponding area of the thermal image. The analysis revealed a clear and significant difference between the two groups. The average eye temperature of the juvenile birds was 0.741°C higher than that of the adult birds. This finding suggests that the younger birds were experiencing greater heat stress or were less efficient at cooling themselves compared to the older, more mature birds. The statistical analysis confirmed that this difference was not a random fluctuation but a consistent pattern observed across the data.

To ensure these results were reliable, the researcher subjected the findings to rigorous testing. They ran thousands of simulated resampling tests to see if the temperature difference held up when the data was shuffled and re-analyzed. In the vast majority of these tests, the younger birds consistently showed higher temperatures. The study also addressed the difficulty of the task, noting that the model trained on the younger birds was slightly less precise than the one for adults, likely because the smaller, less distinct features of juvenile eyes make them harder for a computer to locate in a crowded, moving flock. Despite this challenge, the system successfully processed the data and provided a clear thermal profile for the birds.

The culmination of this research is a practical tool for farmers. The researcher built a mobile web application that allows a user to upload a thermal image of a chicken, select whether it is an adult or a juvenile, and receive an immediate estimate of the bird's eye temperature. The app also provides a severity classification, helping the user understand if the temperature indicates a healthy state or a warning sign of heat stress. This approach offers a way to monitor the welfare of a flock without disturbing the animals or requiring expensive, stationary equipment. By automating the detection of heat stress, the system empowers farmers to intervene early, potentially saving birds from the more serious consequences of overheating.

While the study presents a promising solution, the researcher acknowledges certain limitations. The data was collected from two separate houses, meaning that factors other than age, such as slight differences in ventilation or micro-climates between the two buildings, could have influenced the temperature readings. Additionally, the crowded nature of the barns meant that some birds were partially hidden, which can introduce challenges for any imaging system. The researcher suggests that while thermal eye imaging is a powerful tool, it works best when combined with other observations, such as water intake or the general posture of the birds, to form a complete picture of flock health.

Ultimately, this work demonstrates that it is possible to use artificial intelligence to turn a simple thermal photo into a detailed health report for poultry. The study confirms that younger birds are indeed more vulnerable to summer heat than their older counterparts, a fact that can now be monitored automatically and continuously. By making this technology accessible through a mobile platform, the research paves the way for a new standard in livestock management, where the well-being of the flock is tracked with precision and care, ensuring that heat stress is detected and addressed before it becomes a crisis.