A robust reference hole detector for robotic drilling

Abstract To enhance the accuracy of robotic drilling in aircraft structures, numerous methods for detecting reference holes have been proposed. However, these methods are often ineffective in the presence of occlusions or high noise levels. In this paper, we propose a coarse-to-fine method for detecting multiple types of reference holes. At the coarse localization stage, a gradient-based template matching method is used to efficiently and accurately identify candidate regions. For the precise detection stage, a hybrid method is proposed to extract contours within candidate regions by integrating edge-following and iterative techniques. By leveraging template information iteratively, the proposed method enhances accuracy while substantially reducing computational costs. The experimental results demonstrate that the proposed method achieves a maximum detection error of less than 0.4 pixels and maintains a recall rate above 94% across different types of reference holes. Furthermore, in comparative experiments using reference hole images, our method consistently achieves the highest recall rate among all tested approaches, validating its effectiveness and robustness in complex industrial environments.