Optical Panel Inspection Using Explicit Band Gaussian Filtering Methods in Discrete Cosine Domain

Area of Science:

• Materials Science
• Electrical Engineering
• Computer Vision

Background:

• Capacitive touch panels (CTPs) are popular due to their durability and responsiveness.
• Inspecting defects on CTPs is challenging due to their multilayer structure and textured backgrounds.
• Small defects on textured surfaces are particularly difficult to detect automatically.

Purpose of the Study:

• To develop an automated method for defect inspection of capacitive touch panels (CTPs) with surface texture.
• To enhance the visibility of small defects against complex backgrounds.
• To improve the accuracy and efficiency of CTP quality control.

Main Methods:

• Utilized spectral attributes of the discrete cosine transform (DCT).
• Proposed a novel three-way double-band Gaussian filtering (3W-DBGF) method.
• Applied band filtering, threshold filtering, and Gaussian distribution filtering to process spectral data.

Main Results:

• The proposed DCT-based 3W-DBGF method achieved a flaw detection rate of 94.21%.
• The false-positive rate for normal areas was reduced to 1.97%.
• A correct classification rate of 98.04% was obtained, demonstrating high accuracy.

Conclusions:

• The 3W-DBGF method effectively weakens textured backgrounds and enhances defects in CTPs.
• Simple statistical methods can accurately set binarization thresholds for defect separation.
• The developed approach offers a robust solution for automated CTP defect inspection.