Random residual neural network-based nanoscale positioning measurement

In the field of positioning measurement, complex components, stringent environment and time-consuming calibration are still the main limitations. To address the above issues, this paper presents a deep learning-based positioning methodology, which integrates image processing with the nanomanufacturing technology. The non-periodic microstructure surface with nanoscale resolution is fabricated to provide the surface pattern. A residual neural network is used for surface pattern recognition to reduce the search area. A survival probability mechanism is proposed to improve the transmission efficiency of the network layers. Template matching and sub-pixel interpolation algorithms are combined for the pattern matching. The proposed methodology defines a comprehensive framework for the development of precision positioning measurement, and the effectiveness of the method was validated by the pattern recognition accuracy and the positioning measurement performance collectively. Finally, the trained network exhibits a correct recognition accuracy of 97.6%, and the whole measurement speed was close to being real-time. Experimental results also demonstrated the advantages and competitiveness of the proposed approach compared to the laser interferometer principle.