Emilio Joseph Grieco
Analysis of the Application of Computer Vision and Advanced Models (such as GPT-Vision) Based on Large Language Models (LLMs) for Damage Detection and Classification Projects.
Rel. Andrea Bottino. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Informatica (Computer Engineering), 2026
|
Preview |
PDF (Tesi_di_laurea)
- Tesi
Licenza: Creative Commons Attribution. Download (12MB) | Preview |
Abstract
Automated quality inspection plays a central role in modern manufacturing, where consistent product quality must be guaranteed within the tight cycle times of the assembly line. This thesis addresses the automatic detection of defects in the positioning of a sealing gasket on a tractor cabin frame, a step that is currently performed manually and inspected only prior to roof mounting, when a misaligned gasket can still be corrected before it becomes costly to fix. Two fundamentally different technological strategies are designed, implemented, and compared. The first is an engineered computer vision pipeline that combines a deep learning–based segmentation model (the Segment Anything Model) with clas- sical image-processing techniques: the gasket and the metal frame are segmented into binary masks, their upper boundaries are extracted as polylines, and the lo- cal deviation between them is compared against an empirically learned tolerance envelope to produce an OK/DEFECT verdict.
A dedicated automatic point-selection algorithm, based on texture and colour cues, makes the segmentation robust to the moderate acquisition shifts that occur in a production environment
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Corso di laurea
Classe di laurea
Aziende collaboratrici
URI
![]() |
Modifica (riservato agli operatori) |
