This project is designed by UCI MAE department with the goal of creating a cannon system with the ability to discern, track and shoot at RC cars. The system should be able to hit targets in a 360 degree area from a range of 5 - 15 feet, all while in a safe manner. The three step process is autonomous requiring humans only to power the system on and reload after firing. This project addresses the current inability of off the shelf cannon systems to discern the targets it is firing at, while providing a meaningful learning experience for UCI students.
Our project utilizes computer vision through the Pi Camera Module 3 Wide and the Raspberry Pi 5. With the YOLOv3 object detection algorithm documented on OpenCV, RC cars and their velocities could identified through a digital feed. A target's position was then used with an annular gear turntable driven by a stepper motor to aim the cannon. Once locked onto the target's future position, a DC motor then fired off cannon balls through a rack and pinion system.
This project produced a system that was able to identify, track and shoot at RC cars. The system operated at an effective range of 5 - 10 feet in the standard test environment of a well light room. At further distances, RC cars could not be reliably tracked reducing the systems capability to fire leading shots. In addition, environmental conditions like camera glares or dark shadows further reduced the systems ability to track the target. The system operated with intermittent 360 area coverage through a sweeping scan of the surrounding area. All systems operated in an autonomous manner and the project's total budget didn't exceed the provided $500.