Automated Detection of Solar Cell Defects with Deep Learning

Bartler, Alexander; Mauch, Lukas; Yang, Bin; Reuter, M.; Stoicescu, L.

doi:10.23919/eusipco.2018.8553025

Cited by 66 publications

(26 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Two data augmentation techniques are proposed, horizontal and vertical flipping of images and 90°, 180°, and 270° image rotations. The choice of these data augmentation techniques is due to its widespread use in the literature as can be seen in Bartler et al, Deitsch et al, Pierdicca et al, and Chollet et al…”

Section: Methodsmentioning

confidence: 99%

“…This study proposes the use of the stochastic gradient descent (SGD) and Adam optimiser to maximise classification accuracy. Like the choice in data augmentation techniques, these optimisers are frequently implemented in the literature as can be seen in Bartler et al, Deitsch et al, and Pierdicca et al…”

Section: Methodsmentioning

confidence: 99%

“…Two data augmentation techniques are proposed, horizontal and vertical flipping of images and 90 , 180 , and 270 image rotations. The choice of these data augmentation techniques is due to its widespread use in the literature as can be seen in Bartler et al, 14 Deitsch et al, 15 Pierdicca et al, 17 and Chollet et al 26 In order to optimise the MobileNet and VGG-16 models, an experimental approach is used to find an optimiser that yields the highest classification accuracy. This study proposes the use of the stochastic gradient descent (SGD) and Adam optimiser to maximise classification accuracy.…”

Section: Defect Detection and Classificationmentioning

confidence: 99%

“…In addition to electrical characterisation, studies by Farress et al, Bartler et al, and Deitsch et al have shown that electroluminescence (EL) imaging is suitable for the purposes of PV module defect detection. This is achieved using techniques such as deep learning.…”

Section: Related Workmentioning

confidence: 99%

See 3 more Smart Citations

Photovoltaic defect classification through thermal infrared imaging using a machine learning approach

Dunderdale

Brettenny

Clohessy

et al. 2019

Progress in Photovoltaics

View full text Add to dashboard Cite

This study examines a deep learning and feature‐based approach for the purpose of detecting and classifying defective photovoltaic modules using thermal infrared images in a South African setting. The VGG‐16 and MobileNet models are shown to provide good performance for the classification of defects. The scale invariant feature transform (SIFT) descriptor, combined with a random forest classifier, is used to identify defective photovoltaic modules. The implementation of this approach has potential for cost reduction in defect classification over current methods.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Defect Detection and Classificationmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

Photovoltaic defect classification through thermal infrared imaging using a machine learning approach

Dunderdale

Brettenny

Clohessy

et al. 2019

Progress in Photovoltaics

View full text Add to dashboard Cite

show abstract

“…Therefore, samples from known classes that are located near the decision boundaries are sometimes rejected as unknown classes. Nevertheless, in practice, a slightly lower closed set accuracy is tolerable [2,19].…”

Section: Comparisonmentioning

confidence: 99%

Deep Open Set Recognition Using Dynamic Intra-class Splitting

2020

Self Cite

View full text Add to dashboard Cite

This paper provides a generic deep learning method to solve open set recognition problems. In open set recognition, only samples of a limited number of known classes are given for training. During inference, an open set recognizer must not only correctly classify samples from known classes, but also reject samples from unknown classes. Due to these specific requirements, conventional deep learning models that assume a closed set environment cannot be used. Therefore, special open set approaches were taken, including variants of support vector machines and generation-based state-of-the-art methods which model unknown classes by generated samples. In contrast, our proposed method models unknown classes by atypical subsets of training samples. The subsets are obtained through intra-class splitting (ICS). Based on a recently proposed two-stage algorithm using ICS, we propose a one-stage method based on alternating between ICS and the training of a deep neural network. Finally, several experiments were conducted to compare our proposed method with conventional and other state-of-the-art methods. The proposed method based on dynamic ICS showed a comparable or better performance than all considered existing methods regarding balanced accuracy.

show abstract

Early hotspot detection in photovoltaic modules using color image descriptors: An infrared thermography study

Ali

Saleem

Masood

et al. 2021

Intl J of Energy Research

View full text Add to dashboard Cite

Summary This paper proposes a new framework for early hotspot detection in the photovoltaic (PV) panels using color image descriptors and a machine learning algorithm. In the proposed approach, the acquired thermographic images of PV panels are divided into non‐overlapping regions, and then color image descriptors are computed for the regions. The color descriptors are then used as features to train different machine learning algorithms to classify the PV panels into three classes (ie, normal, hotspot, and defective). After extensive testing and comprehensive analysis, the experimental results show that Red‐Green Scale‐Invariant Feature Transform (rgSIFT) descriptor with k‐Nearest Neighbor (k‐NN) outperforms all other images descriptors and machine learning combinations with an accuracy rate of 98.7%. The experimental results also show the effects of the size of non‐overlapping regions on the classification accuracy. It is observed that the classification accuracy decreases as size is increased or decreased around the optimal non‐overlapping region image size of 71 × 71 pixels. The proposed method has a significant role in carbon‐free cities and can easily be implemented to inspect the PV system.

show abstract

Automated Detection of Solar Cell Defects with Deep Learning

Cited by 66 publications

References 12 publications

Photovoltaic defect classification through thermal infrared imaging using a machine learning approach

Photovoltaic defect classification through thermal infrared imaging using a machine learning approach

Deep Open Set Recognition Using Dynamic Intra-class Splitting

Early hotspot detection in photovoltaic modules using color image descriptors: An infrared thermography study

Contact Info

Product

Resources

About