Robust Visual Detection for Vague Scratches defect in inhomogeneous surface

Scratch inspection has become a challenging problem in the detection of multicrystalline solar cells surface quality because the scratches in inhomogeneous textured surfaces are usually shallow and weak. Thus, this paper presents a novel framework for detecting vague scratches in the complex inhomogeneous surface texture of multicrystalline solar cell. Firstly, an enhanced texture energy measure is used for highlighting candidate scratch defcects information, and simultaneously suppressing inhomogeneous surface texture information. Then, a robust structural-texture decomposition method with an adaptive parameter is employed to capture the meaningful structure information from the enhanced image. Futheremore, an optimal threshold from a Otsu's threshold segmentation method is used to distinguish defect from the background. Finally, some experimental results have shown the proposed method can effectively detecte vague scratches in multicrystalline solar cell surface.