Robust Color Classification for Autonomous Robotic Boats

Abstract

This research focuses on the development of a robust color classification system for autonomous robotic boats with applications in the context of buoy identification. The proposed methodology utilizes supervised learning techniques to classify buoy colors effectively, even under challenging and varying lighting conditions. I collect and label extensive datasets from real-world scenarios featuring a diverse range of buoy colors and lighting conditions. Employing multiclass SVM and multinomial Naive Bayes algorithms, our approach achieves high accuracy rates, demonstrating its potential for real-time applications. Future milestones include data diversification and the incorporation of lighting direction as a feature, aiming to enhance the algorithm's robustness and accuracy in practical deployment scenarios. Ultimately, this research contributes to improving the reliability of autonomous robotic boat navigation in competitions and beyond.