Dual contour edge information fusion for 3D aircraft skin circular hole detection

To address the challenges of varied aircraft skin hole detection types and susceptibility to noise, this paper proposes a method based on the perspective of dual contour edge information fusion. The core method combines stereoscopic vision and structured light dual contour information consistently, focusing on extracting edge point information around the hole edge to achieve precise detection of circular holes. In this approach, a line multi-directional gradient feature detector (LMGFD) is introduced for locating the holes from plane stereoscopic image. Furthermore, we establish a three-dimensional (3D) circular hole detection method (BPCircle) based on the dual contour edge information fusion. Finally, experiments demonstrate that our proposed method achieves superior accuracy and robustness based on public benchmark dataset and our own collected standard IPCDS dataset (including two-dimensional (2D) images, 3D point clouds, and measured data of three-coordinate measuring machine). The dataset and code can be found from https://github.com/Nicho1sdqw/123.