Closing the Loop from Pixels to Motion

Our Guangdong Second Prize entry in the 2023 National Undergraduate Electronic Design Contest used two servo-driven gimbals for vision-guided laser tracking. I implemented image processing, Kalman filtering, and PID control, then tuned the integrated system during the four-day competition.

OpenMV detected laser spots; an Arduino Mega2560 converted image coordinates into actuator commands. Regression-based calibration addressed geometric bias, while filtering and trajectory interpolation reduced noise-driven motion.

Performance on the Demonstration Tasks

The reported static and dynamic positioning tests achieved less than 1 cm error, with 0.5–0.6 seconds from target detection to stable tracking lock. Exposure tuning, LAB color filtering, and servo calibration helped sustain the complete detection-to-control loop under the competition conditions.

On-site tracking-system demonstration