UAV Drone: Object Tracking using Kalman Filter

In our scenario, our drone delivers packages in busy urban areas. Instead of dropping them from above with parachutes like other systems, we use a cord mechanism. This cord lowers the package once the drone reaches its destination, offering a smoother experience. But there's a risk: someone could pull or trip on the cord, potentially causing harm.

To manage this risk, our drone has cameras onboard. These cameras help us create a tracking system. It watches for people or animals near where the package will land. If it spots anyone, it waits to lower the package until it's safe. This proactive approach ensures safe deliveries without endangering anyone.

Urban areas pose unique challenges for tracking systems. Buildings and other obstacles can block the cameras' view. Plus, there are lots of moving things to keep track of. Our system needs to handle these challenges to work reliably and safely. In the video above, an obstructed object is shown where YOLOv8 fails to detect it. However, utilizing the Kalman Filter, we are still able to predict the future states of the object by leveraging the Process model and the current states.