Analysis and Evaluation of Pre-processing Techniques for Fault Detection in Thermal Images of Solar Panels

Abstract

AbstractMajority of solar panel thermal images require post-acquisition manipulations for optimization of contrast, brightness, and noise removal. Noise removal and contrast improvement are major part of pre-processing operations. Thermal imaging is one of the non-contact techniques used for fault detection in solar panels. Thermal images captured through thermal camera are often corrupted with noise due to various environmental conditions. Use of suitable denoising filter is an essential pre-processing step in case of thermal imaging. In this paper, various digital filters such as Gaussian, median, bilateral, mean, and Wiener filter are tested for removal of noise. The performance of these filters is evaluated using statistical measures such as mean square error (MSE), structural similarity index (SSIM), signal-to-noise ratio (SNR), and peak signal-to-noise ratio (PSNR). After filtering thermal images with suitable filter, contrast must be enhanced using good histogram equalization technique. To enhance the contrast of filtered images, various histogram equalization techniques are applied. This paper proposes use of brightness preserving dynamic fuzzy histogram equalization (BPDFHE) for solar panel thermal images by comparing the performance against histogram equalization (HE), mean preserving Bi-histogram equalization (BBHE), contrastive limited adaptive equalization (CLAHE), equal area dualistic sub-image histogram equalization (DSIHE) techniques. The qualitative attributes used for evaluation are peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and absolute mean brightness error (AMBE).KeywordsThermal imagingThermal graysFiltersHistogram equalizationMSEPSNRSSIMSNR