Wafer Surface Defect Detection Based on Background Subtraction and Faster R-CNN.

Area of Science:

• Semiconductor manufacturing
• Computer vision
• Image processing

Background:

• Wafer surface defects are challenging to detect due to similarity with background.
• Automated defect detection is crucial for intelligent manufacturing.

Purpose of the Study:

• To propose a new wafer surface defect detection method.
• To improve the accuracy and reliability of defect detection.

Main Methods:

• Improved spectral analysis for image period measurement.
• Local template matching for substructure image reconstruction.
• Background subtraction via image differencing.
• Detection using an improved Faster R-CNN network.

Main Results:

• The proposed method effectively reconstructs the background image.
• Background interference is significantly reduced.
• The improved Faster R-CNN achieves higher detection accuracy.
• Mean Average Precision (mAP) increased by 5.2% compared to original Faster R-CNN.

Conclusions:

• The developed method accurately detects wafer surface defects.
• It meets the demands of intelligent manufacturing for high detection accuracy.
• This approach offers a robust solution for semiconductor quality control.