Micro-crack detection of solar cell based on adaptive deep features and visual saliency

Abstract

Purpose An effective machine vision-based method for micro-crack detection of solar cell can economically improve the qualified rate of solar cells. However, how to extract features which have strong generalization and data representation ability at the same time is still an open problem for machine vision-based methods. Design/methodology/approach A micro-crack detection method based on adaptive deep features and visual saliency is proposed in this paper. The proposed method can adaptively extract deep features from the input image without any supervised training. Furthermore, considering the fact that micro-cracks can obviously attract visual attention when people look at the solar cell’s surface, the visual saliency is also introduced for the micro-crack detection. Findings Comprehensive evaluations are implemented on two existing data sets, where subjective experimental results show that most of the micro-cracks can be detected, and the objective experimental results show that the method proposed in this study has better performance in detecting precision. Originality/value First, an adaptive deep features extraction scheme without any supervised training is proposed for micro-crack detection. Second, the visual saliency is introduced for micro-crack detection.