Dual-camera framework for detecting subtle, low-contrast defects on multi-layered cover glass in display manufacturing

Cover glass defects in smartphone displays require precise detection to ensure optical performance, yet post-lamination inspection faces three challenges: multi-layer transparency interference, low-contrast defect visibility, and anisotropic process noise. This study proposes a co-designed hardware-algorithm framework to address these limitations: the dual-camera imaging system tackles multi-layer transparency interference via multi-perspective synchronized acquisition, while two algorithms address the remaining challenges. Attention-guided FT-SWT fusion combines frequency/spatial domain analysis with attention mechanisms to isolate low-contrast defect features, and anisotropic multi-scale Gaussian filtering removes directional noise without blurring boundaries. Validated across industrial samples, the framework achieves 98.7% detection precision with real-time capability, successfully identifying micron-scale particles, adhesive voids, and texture anomalies. The methodology provides a robust quality control solution for transparent multi-layer composites in display manufacturing.