High Performance Wafer Defect Classification Model Based on Feature Fusion and RGB SEM Images

The intelligent analysis of wafer defects is crucial for enhancing yield management and improving production efficiency. This study proposes an automatic defect classification (ADC) method for RGB SEM images and integrates the features of VGG32 and UNet models through cross-fusion techniques. Experimental results demonstrate significant advantages of RGB SEM images over gray SEM images in defect classification, and the feature cross-fusion model achieves better accuracy compared to individual models.