An AFM rapid detection method for micro-nano level defects on surface of optical elements based on convolutional neural network
By employing a fast AFM detection method based on convolutional neural networks, and utilizing optical microscope image stitching and AFM undersampled data to reconstruct the model, the problem of low detection efficiency of micro-nano defects on the surface of optical components is solved, achieving efficient and accurate detection of micro-nano defects.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HARBIN INST OF TECH
- Filing Date
- 2023-10-25
- Publication Date
- 2026-07-21
AI Technical Summary
Existing methods for detecting micro- and nano-defects on the surface of optical components are inefficient, especially for defects smaller than 0.5 to 20 μm, where the identification accuracy is low. Furthermore, long-term AFM scanning can lead to mechanical drift, affecting the detection accuracy.
A fast AFM detection method based on convolutional neural networks is adopted. Sub-aperture images are obtained by optical microscope and stitched together. An image reconstruction model is constructed by combining undersampled and fully sampled data from AFM. End-to-end mapping is performed using convolutional neural networks to achieve rapid detection of defects.
It improves AFM detection efficiency and reduces the impact of mechanical drift on defect morphology information, making it suitable for efficient and rapid detection of a large number of micro- and nano-defects on the surface of optical components.
Smart Images

Figure CN117452025B_ABST