Computer-Vision Based Bend Quality Inspection for Sheet Metal Parts
Author
Kantas, Christos
Term
4. semester
Publication year
2025
Submitted on
2025-06-04
Pages
66
Abstract
In modern sheet-metal production, bends must meet specifications to ensure quality and fit. 3D scanners capture these parts as point clouds—dense sets of points that describe the surface in three dimensions. This thesis examines how to assess bend quality directly from such data. We present the manufacturing system and the available scans, then introduce an automatic processing pipeline. The pipeline pre-processes the scans, detects where the bends are, and measures each bend’s total inclination angle (how far it is bent) and its arc length (the length along the curved path). The method is first tried on ideal data to show that the concept works in principle. It is then tested for robustness to noise and for its ability to produce realistic results on real scans.
I moderne pladeproduktion skal bøjninger opfylde krav for at sikre kvalitet og pasform. 3D-scannere registrerer delene som punktskyer—tætte samlinger af punkter, der beskriver overfladen i tre dimensioner. Dette speciale undersøger, hvordan bøjningers kvalitet kan vurderes direkte ud fra sådanne data. Vi beskriver fremstillingssystemet og de tilgængelige scanninger og præsenterer derefter en automatisk behandlingspipeline. Pipelinen forbehandler scanningerne, finder hvor bøjningerne er, og måler hver bøjnings samlede hældningsvinkel (hvor meget den er bøjet) og dens buelængde (længden langs den krumme bane. Metoden afprøves først på ideelle data for at vise, at den er principielt velfungerende. Derefter testes den for robusthed over for støj og for, om den giver realistiske resultater på virkelige scanninger.
[This apstract has been rewritten with the help of AI based on the project's original abstract]
Keywords