Applied imagery pattern recognition for photovoltaic modules’ inspection: A review on methods, challenges and future development

We present a literature review of Applied Imagery Pattern Recognition (AIPR) for the inspection of photovoltaic (PV) modules under the main used spectra: (1) true-color RGB, (2) long-wave infrared (LWIR), and (3) electroluminescence-based short-wave infrared (SWIR). Three sequentially linked building blocks underpin this work. The first overviews reference guidelines of image acquisition and the main detectable defect patterns under each spectrum. It also provides key insights regarding the implementation of Unmanned Aerial Vehicles (UAVs) to acquire imagery, especially from a photogrammetric perspective. The second block presents various image pre-processing steps used to prepare inspection-ready datasets. These comprise radiometric correction, segmentation and edge extraction, geometric correction and cell clipping. The third surveys defect detection and classification through digital image processing and machine/deep learning techniques. We elaborate an in-depth topic discussion that, in parallel: highlights the main related challenges, provides core guidelines for AIPR-based PV inspection workflows, and suggests key research avenues for future studies. This review synthesizes the recent advances of the body knowledge. It also constitutes an insightful reference for professionals and academics within the PV operations and maintenance field who are considering the possibilities that digital imagery can offer.