polito.it
Politecnico di Torino (logo)

Artificial Intelligence Algorithms for Electronic Component Recognition

Giovanni Cadau

Artificial Intelligence Algorithms for Electronic Component Recognition.

Rel. Daniele Apiletti. Politecnico di Torino, Corso di laurea magistrale in Data Science And Engineering, 2024

Abstract:

In the challenging scenario of Automatic Electronic Testing of Printed Circuit Boards (PCBs), a crucial role is played by the recognition of specific electronic components, such as resistors, inductors, capacitors, and others. This task usually requires an extensive amount of manual labor, necessitating the presence of a domain expert. The great heterogeneity of PCBs and the highly variable environmental conditions under which images are collected within a corporate production mechanism also demand high precision and significant time. The objective of this thesis is to demonstrate the use of artificial intelligence (AI) algorithms, particularly machine learning (ML) and deep learning (DL) ones, to perform the task of image recognition, determining the position and nature of various components within PCBs. Additionally, this thesis aims to integrate the resulting models into a corporate context, bringing efficiency benefits. Starting from an accurate and extensive collection of images, in order to ensure high data quality, and defining an exhaustive pipeline for preprocessing and augmentation, it was possible to develop ML and DL models that were integrated into the corporate management software. The aim was not only to optimize the accuracy of the models but also to address aspects particularly relevant in a corporate context, such as latency, memory usage, and energy consumption. The use of Domain Randomization and Adaptive Domain Randomization techniques, combined with the use of optimization algorithms, both gradient-based and gradient-free, allowed for the creation of robust models, capable of handling the variability and complexity of the environment in which the images were collected and might be collected in the near future, without the need to retrain a new model.

Relators: Daniele Apiletti
Academic year: 2024/25
Publication type: Electronic
Number of Pages: 144
Additional Information: Tesi secretata. Fulltext non presente
Subjects:
Corso di laurea: Corso di laurea magistrale in Data Science And Engineering
Classe di laurea: New organization > Master science > LM-32 - COMPUTER SYSTEMS ENGINEERING
Aziende collaboratrici: SEICA SpA
URI: http://webthesis.biblio.polito.it/id/eprint/33095
Modify record (reserved for operators) Modify record (reserved for operators)