A High-Accuracy Slotted Hole Detector
Liming Tao, Renbo Xia, Jibin Zhao, Yinghao Li, Hangbo Zou, Fangyuan Wang · 2022
Slotted holes are often integrated into industrial parts to flexibly adjust the position of the parts. The measurement of the geometric parameters of the slotted hole is the key to ensure the mechanical properties of the parts. Identifying and detecting the edges of slotted holes on the image is the basis for the slotted hole measurement with high-accuracy. This is a challenging problem because the projection of the slotted hole on the image is not a primitive geometry. In this paper, a high-accuracy slotted hole detector is designed. First, a curvature-based algorithm is introduced to detect the dominant points on the curves of interest after edge detection with Canny detector. Then, a dominant point verification algorithm based on a new projective invariant is proposed to significantly remove incorrect dominant points on the edge of the slotted holes. Finally, a dominant point tuning algorithm based on line and ellipse fitting is developed for accurately finding the segmentations between lines and ellipse arcs to calculate slotted hole parameters. Extensive experiments on real images and synthetic images demonstrate that the proposed slotted hole detector has high detection accuracy and strong robustness.