Abstract
Abstract
Crack defects on the surface of glassivation passivation parts (GPP) wafers are a significant factor leading to chip functional failure. Thus, efficient and stable surface defect detection is required to ensure product quality. However, existing detection methods do not handle weak crack defects on the wafer surface with an inhomogeneous texture background effectively. To overcome this challenge, we propose a weak crack defect detection method for GPP wafers with complex backgrounds. First, a weak crack feature enhancement method is proposed based on the Gabor filter with different orientations. Then, the crack candidates are generated using the least-squares method to transform the shape-based information into a line segment and direction independence judgment. Finally, a line segment clustering method is implemented for the final crack defect recognition. Experimental results demonstrate that the proposed defect detection method works effectively and robustly with inhomogeneous texture backgrounds and satisfies the requirement of a real-time detection system.
Funder
Anhui Provincial Key Research and Development Plan
Subject
Applied Mathematics,Instrumentation,Engineering (miscellaneous)