Evaluation of Motion-Induced Noise and Pixel-Bleeding in Electroluminescence Field Inspection of PV Modules

There is an increasing demand for accurate daylight diagnostics of larger photovoltaic (PV) plants with Electroluminescence (EL) imaging. Modulated contact-EL can remedy solar noise, but mobile platforms are required to increase inspection speed. However, any motion induces noise, even with perfect tracking of the PV modules in the images. This paper investigates the impact of motion and camera calibration on the quality of daylight contact-EL imaging since both introduce noise. Using the $SNR_{50}$ and $SNR_{Kari}{ }^{\prime}$ metrics and visual inspection, a total of 58 stationary and moving EL imaging series, 23 calibrated and 35 uncalibrated, were analyzed and investigated both with and without module tracking. $SNR_{50}$ proved an unreliable predictor of image quality, but $SNR_{Kari}{ }^{\prime}$ also revealed uncertainties when faced with camera motion. Motion severity was a superior metric due to an enhanced ability to predict non-tracked image quality, but more stable metrics must be developed. Without stabilization, image quality deteriorated rapidly at motion above 0.18 and 0.06 pixel/image with and without camera calibration, respectively. With stabilization, calibrated image series stayed at a quality level suitable for manual diagnostics, even at extreme motion. Still, the uncalibrated series did not; showing calibration vital for moving imaging inspection platforms. For reliable diagnostics and automated processing, better algorithms are needed.