Nanowire Detection in AFM Images Using Deep Learning.

Area of Science:

• Nanotechnology
• Materials Science
• Computer Vision

Background:

• Atomic force microscope (AFM) based nanomanipulation is crucial for assembling nanoscale devices.
• Efficient and automated identification of nanoparticles and features in AFM images is essential for nanomanipulation.
• Current methods may lack the precision and speed required for complex automated assembly.

Purpose of the Study:

• To develop an automated deep learning-based method for detecting flexible nanowires in AFM images.
• To improve the efficiency and reliability of nanomanipulation by enabling precise nanowire identification.
• To create a robust algorithm capable of handling variations in nanowire morphology and image quality.

Main Methods:

• Application of the You Only Look Once (YOLO) network for initial nanowire detection.
• Utilizing morphology transformation algorithms, including adaptive threshold edge detection, to refine nanowire skeletons.
• Employing a bidirectional long short-term memory with a conditional random field layer (BI-LSTM-CRF) for precise posture and position determination.

Main Results:

• The developed program automatically detects nanowires of diverse morphologies with nanometer resolution.
• Achieved over 80% reliability in detecting nanowires within the testing dataset.
• Demonstrated good robustness, with detection results minimally affected by image quality variations.

Conclusions:

• The proposed deep learning approach enables efficient and reliable automated detection of flexible nanowires.
• This method significantly advances the capabilities for automated nanomanipulation and nanoscale device fabrication.
• The algorithm's robustness ensures applicability across various experimental conditions and image qualities.