Multi-part Deviation Assessment in Ship Assembly Structures Through Instance Segmentation

Jian Liu,Deyu Wang
DOI: https://doi.org/10.1016/j.oceaneng.2024.116941
IF: 5
2024-01-01
Ocean Engineering
Abstract:In the shipbuilding process, manufacturing inaccuracies can lead to discrepancies between the actual fabricated structures and their corresponding three-dimensional (3D) design models. This can make it difficult to guarantee complete conformity between the fabricated structures and their 3D, as-designed models. While the manufactured structures may be very similar to the as-designed models, it is difficult to effectively evaluate the deviations in the assembly of each individual part. This paper proposes a method for evaluating the assembly quality of ship structures by segmenting individual parts. Laser scanning is used to collect the surface geometric data of the ship structure, which is referred to as the as-scanned model. An abstract representation of each part within the as-scanned model and its corresponding as-designed model is then obtained by implementing a region-growing algorithm. Using the abstract representation of each part in conjunction with the two-dimensional (2D) synthetic images of the hull parts, accurate pairing and registration of each part within the ship structure is achieved, facilitating accurate evaluation of each part within the ship assembly structure. The reliability and accuracy of this method have been validated using two commonly constructed structures as testing objects.
What problem does this paper attempt to address?