A combination of infrared imaging, thermal imaging and color cameras on an uncrewed drone, along with an AI system to interpret the data, can help emergency responders and search-and-rescue teams locate, identify and track people who have gone missing in the wilderness. The experimental system helps responders pinpoint where a missing person is and determine whether they are hurt or even alive.

People who get lost or hurt while exploring nature can become stranded for days. Rescue teams often use drones to look for the person or signs of their whereabouts. The small drone my colleagues and I built at my lab at Kennesaw State University flies autonomously using a grid search pattern. It sends live video and images to a ground station operated by the rescue team.

When the AI system finds a person, it analyzes images to determine whether the individual is upright or lying on the ground. It segments parts of the person’s body, identifying the person’s head and the body’s position. It then zeroes in on the forehead. It extracts forehead temperature readings, pixel by pixel, from the imaging data to estimate forehead temperature. We have two papers detailing these findings accepted for the American Institute of Aeronautics and Astronautics Aviation Forum 2026 conference.

Our AI model then assesses whether the person is conscious or unconscious and identifies abnormal temperatures that could indicate heat stress, hypothermia or other physical complications, or death – all vital information for a search-and-rescue team.

In field trials we have conducted, the system has provided consistent temperature readings of the heads of volunteers from our research team who have walked out into a variety of environments, under different conditions.

Why it matters

It is critical to get accurate and timely information on the whereabouts of a missing person. The likelihood that the person will survive decreases steeply as time passes.

An AI-enhanced drone can make search-and-rescue operations significantly more efficient than sending teams of people out into the environment to search on foot, especially in poor weather conditions or under thick foliage. Rescuers who know whether a person is conscious or unconscious can also better gear up for what they need to do to retrieve the person and administer aid. Our technology could save lives.

What other research is being done

Search-and-rescue personnel use various kinds of drones, but the machines often lack the ability to positively identify humans, especially under thick foliage, in bad weather or when the person is lying down or unconscious. The AI-based technology we have developed overcomes those challenges.

Better sensors that are very lightweight, that can function at night or in rain, and can see more clearly through thick foliage could further improve our drone and drones used by others. Researchers are devising AI-powered sound recognition for detecting screams for help, advanced thermal imaging for better nighttime vision and autonomous drones that could act as first responders.

Also under development are drones that can carry heavy payloads, such as flotation devices, fly for up to 14 hours or perform real-time mapping of the ground below.

What’s next

One of our next steps is to have multiple drones fly together and autonomously coordinate search-and-rescue operations among themselves. This will allow the technology to cover a much larger area, perhaps hundreds of square miles.

We are also designing a large drone that can carry up to 110 pounds (50 kilograms) of payload and stay aloft for an hour.

The Research Brief is a short take on interesting academic work.

Adeel Khalid receives funding from the Office of Research at Kennesaw State University.