Etudiant : EL BASSRI ZAYD

Filière : LF Sciences Mathématiques et Informatique

Encadrant : Pr. EN-NAHNAHI NOUREDDINE

Annèe : 2022

Résumé : Photovoltaic (PV) are an increasingly important source of renewable power that will play an essential role in fighting with the worldwide energy consumption, this enforces having a good quality of PV modules from production to operation. Furthermore, the defect segmentation of photovoltaic cells in Electroluminescence Images is becoming harsh and expensive manually, and require a big knowledge in so many fields of defect detection, so providing an automatic detection of such defects is a challenging task. In this work we propose a machine learning approach using a convolutional neural network architecture for recognizing defects in EL images. which achieves state of the art results of 83% on solar cell dataset of EL images with pretrained model, and 73% with our customized architecture. It takes only 8 ms for predicting one image. Moreover, we evaluate data augmentation operations to deal with data scarcity. Overfitting appears a significant issue, to deal with it we adopted some efficient strategies to generate our model. The impact of each strategy is presented. The proposed interface can help for automatic detection in field and industry.