Hybrid Computer Vision System for Traceability Improvement in the Metal Manufacturing Sector

Abstract

[Abstract] The integration of digital technologies in the shipbuilding industry has become essential to improve efficiency and ensure precision in complex manufacturing processes. This study explores the application of computer vision techniques for the identification and traceability of minor and simple subassemblies in shipbuilding. We developed a dual approach combining 3D point cloud analysis and 2D deep learning based instance segmentation to detect and classify components. A depth camera was used to acquire 3D point clouds and high-quality 2D images. For the 3D approach, we used surface and edge matching techniques, and for the 2D approach, we fine-tuned advanced instance segmentation models such as YOLO11 and MaskRCNN2. The results show that this combined approach significantly improves the reliability of subassembly identification, which is essential for improving traceability, reducing errors, and optimizing production workflows in the shipbuilding sector.