Multi-modal machine vision-based gap detection algorithm for composite surface stitching.

This paper proposes a multi-modality-based machine vision gap detection method, aiming at the problems of insufficient feature extraction, low accuracy of point cloud segmentation, and poor edge fitting effect in traditional carbon fibre composite material gap detection methods. First, an improved sub-pixel gap edge detection method is proposed to extract more abundant gap features. Then, an adaptive unified point cloud orientation method is designed to achieve accurate segmentation of the point cloud gap centreline by enhancing the point cloud curvature feature. Finally, an innovative joint processing method based on 2D-3D vision is proposed, which can classify and fit the discrete feature points of the 2D gap by introducing the 3D midline and generating the gap edge. Experiments show that this method can accurately and reliably extract the tiny gaps in laying carbon fibre composites. This method is suitable for online layup detection of carbon fibre composites.