A knowledge-guided, drift-aware computer vision pipeline for automated nanoscale atomic force microscope surface...

Area of Science:

• Materials Science
• Nanotechnology
• Computer Vision

Background:

• Atomic Force Microscopy (AFM) measurements are crucial for nanoscale surface characterization.
• Traditional AFM analysis is manual, observer-dependent, and lacks robustness in tracking features across multiple scans.
• Drift and layer-dependent responses complicate accurate tribological analysis.

Purpose of the Study:

• To introduce a two-stage, knowledge-guided, and drift-aware computer vision (CV) pipeline for AFM data analysis.
• To establish a robust nanoscale point-tracking method for automated analysis across consecutive loading conditions.
• To enable pixel-level computation of tribological responses, including friction forces and load-friction curves.

Main Methods:

• A two-stage CV pipeline combining SIFT-ORB feature extraction and RANSAC for Stage 1 point correspondence.
• Stage 2 employs a drift-aware chained tracking structure for preserving point identity.
• Validation on a 2D material system to characterize lateral drift and disentangle layer-dependent tribological responses.

Main Results:

• The pipeline successfully establishes nanoscale point tracking across consecutive AFM load steps.
• It accurately characterizes lateral drift and layer-dependent tribological responses in 2D materials.
• The approach produces machine-readable, reproducible, and statistically consistent tribological outputs.

Conclusions:

• The developed CV pipeline redefines AFM analysis by automating manual workflows.
• It provides the first drift-aware and knowledge-guided CV-based nanoscale characterization method.
• This approach enhances the reliability and reproducibility of tribological measurements.