Fast and Robust Detection of Solar Modules in Electroluminescence Images

Vento, Mario; Percannella, Gennaro; Christlein, Vincent; Brabec, Christoph J.; Doll, Bernd; Maier, Andreas K.; Hoffmann, Mathis; Talkenberg, Florian

doi:10.1007/978-3-030-29891-3_46

Contribution to a book

FZJ-2021-01456

Fast and Robust Detection of Solar Modules in Electroluminescence Images

Vento, M. (Editor) ; Percannella, G. (Editor) ; Hoffmann, M. (Corresponding author) ; Doll, B. ; Talkenberg, F. ; Brabec, C. J.FZJ* ; Maier, A. K. ; Christlein, V.

2019
Springer International Publishing Cham
ISBN: 978-3-030-29890-6 (print), 978-3-030-29891-3 (electronic)

Computer Analysis of Images and Patterns / Vento, Mario (Editor) [https://orcid.org/0000-0002-2948-741X] ; Cham : Springer International Publishing, 2019, Chapter 46 ; ISSN: 0302-9743=1611-3349 ; ISBN: 978-3-030-29890-6=978-3-030-29891-3 ; doi:10.1007/978-3-030-29891-3 Cham : Springer International Publishing, Lecture Notes in Computer Science 11679, 519 - 531 (2019) [10.1007/978-3-030-29891-3_46]

This record in other databases:

Please use a persistent id in citations: http://hdl.handle.net/2128/27466 doi:10.1007/978-3-030-29891-3_46

Abstract: Fast, non-destructive and on-site quality control tools, mainlyhigh sensitive imaging techniques, are important to assess the reliabilityof photovoltaic plants. To minimize the risk of further damages andelectrical yield losses, electroluminescence (EL) imaging is used to detectlocal defects in an early stage, which might cause future electric losses. Foran automated defect recognition on EL measurements, a robust detectionand rectication of modules, as well as an optional segmentation intocells is required. This paper introduces a method to detect solar modulesand crossing points between solar cells in EL images. We only require1-D image statistics for the detection, resulting in an approach that iscomputationally ecient. In addition, the method is able to detect themodules under perspective distortion and in scenarios, where multiplemodules are visible in the image. We compare our method to the state ofthe art and show that it is superior in presence of perspective distortionwhile the performance on images, where the module is roughly coplanarto the detector, is similar to the reference method. Finally, we show thatwe greatly improve in terms of computational time in comparison to thereference method.

Contributing Institute(s):