A physics-informed eigenfilter for artifact removal in ultrasonic scanning videos for structural inspection

Area of Science:

• Non-destructive testing
• Ultrasonic imaging
• Signal processing

Background:

• Artifacts in industrial ultrasound videos obscure critical flaw detection.
• Conventional filtering methods struggle with real-world signal variations.
• Eigenspace filtering shows promise for artifact removal in medical ultrasound.

Purpose of the Study:

• Adapt and refine eigenspace filtering for industrial ultrasound artifact suppression.
• Develop a robust method for flaw imaging in railroad rails using a Rolling Search Unit (RSU).
• Clarify the principles of eigenfiltering for industrial applications.

Main Methods:

• Applied eigenfiltering to industrial ultrasound data from a Synthetic Aperture Focus Technique (SAFT) system.
• Analyzed spatiotemporal autocorrelation differences between artifacts and flaws.
• Proposed and implemented a novel recursive eigenfilter with rectification and a nonnegativity constraint.

Main Results:

• Demonstrated that eigenfilter effectiveness relies on autocorrelation differences.
• The recursive eigenfilter successfully suppresses artifacts even with shifting positions.
• Experimental results show outstanding filtering performance on artificial and natural rail flaws.

Conclusions:

• The proposed recursive eigenfilter offers significant improvements for artifact suppression in industrial ultrasound.
• This method enhances the detection of weak structural features like flaws in railroad rails.
• The filtering approach is broadly applicable to scanning-based imaging systems.