Optical nanoscale positioning measurement with a feature-based method

Traditional nanoscale positioning measurement methods require high-precision components and time-consuming calibration. To address these problems, this paper develops a fast and robust feature-based positioning (FRFP) method for nanoscale positioning measurement. Firstly, a unique polar microstructure surface is designed, and ultra-precision machined for imaging and matching. The next step uses a box filter which is computational saving to detect the initial feature points from the surface images. After that, the image scale is built to extract the robust feature points. The filter size is changed in each scale layer instead of image size to further reduce the number of calculations. Then, the orientation assignment process is conducted for angular displacement detection. After generating the robust feature points descriptors with 64-dimensional vectors, the feature point matching is performed to determine the absolute position changes between the images. Finally, sub-pixel interpolation is also merged into the FRFP method to further improve the positioning resolution. To show the effectiveness of the FRFP method, experiments are conducted from the aspects of angular uncertainty, position uncertainty, measurement speed, and robustness respectively. All the experimental results demonstrate the efficiency and robustness of the FRFP method.