Electroluminescence imaging and automatic cell classification in mass production of silicon solar cells

With increasing manufacturing volume, automation in solar cell production and quality control becomes increasingly important. In this paper we develop and demonstrate a pipeline for optimization and evaluation of automatic cell sorting algorithms based on electroluminescence imaging. We provide general applicable guidelines for optimization throughout the whole process and demonstrate their application for an exemplary case. As a result, we show that our evaluation algorithm ELEval-2 outperforms trained personnel of a solar cell manufacturer in solar cell quality rating. This claim is based on a diligent quantitative comparison of the false detection rates of algorithm and employees. We conclude that our approach provides means for rating, comparison and optimization of such algorithms. Also, we aim to initiate an open discussion about an international standard on performance assessment of electroluminescence evaluation approaches.