HawkEye - Rescue Drone Module

A module that can be attached to a UAV (drone/plane) to find a human using YOLOv8 AI and a thermal camera, log their GPS location, and drop help right on the spot.

Comment

Inspiration

In search and rescue the biggest problem isnt reaching the person, its finding them fast enough. Drones can look from above but the pilot still has to manually notice someone on the screen. We wanted the drone to actually understand what its seeing.

What it does

we built an attachable module that can go on a UAV (drone or plane). The idea is simple, the camera on the system detects heat, confirms its a human, saves the location, and drops help right where they are.

How we built it

We used a Raspberry Pi as our main microcontroller and connected a thermal camera, Raspberry Pi Camera Module 2, and servos to drop a small first aid kit with a parachute. The thermal camera constantly scans, and when it detects heat we run YOLOv8m to confirm its a person. It then takes a screenshot tagged with date, time, and GPS location.

Challenges we ran into

  • the thermal camera had a low resolution initially. -had to use python when we were used to c/c++.
  • had issues running both cameras at once.
  • had initial hiccups designing CAD models.
  • GPS sensor was unreliable due to not working indoors. -Difficulty integrating to upload location data on detected humans. -overused the backup IP geolocation API.

Accomplishments that we're proud of

  • successfully got both cameras working with timestamped photos of human and thermal body scans detected.
  • successfully built and designed a UAV and drone module.
  • built a user friendly interface in software for the drone. -successfully integrated multiple sensors in python libraries in order to build advanced target detection system. -successfully used OpenCV and YOLOv8 to detect human beings and recognize when they are under the drone with HIGH level accuracy for location documation.
  • Used CAD (Onshape, Fusion360) to design a UAV and drone module.

What we learned

  • learned Vision, OpenCV and YOLOv8 on the RaspberryPi.
  • learned how to work with MongoDB.
  • learned about wifi and IP geolocation methods.
  • how to work with python in an embedded system platform. (RaspberryPi)
  • control RaspberryPi to work with UART and I2c sensors.

What's next for HawkEye - Rescue Drone Module

  • Training an AI module to read and interpret the heat map data.
  • Training the AI to also recognize animals (dogs, cats, pets) and hazards (fire, flood) using the heat map and camera data.
  • More refined drone module and UAV design.
  • Upgrading RaspberryPi to Nvidia jetson.
  • Developing friendly user interface for first responders.

Built With

Share this project:

Updates