Residual Gratings Removal In Moiré Pattern Using Autoencoder
| dc.contributor.author | Nga, Eng Fon | |
| dc.date.accessioned | 2022-12-14T08:21:21Z | |
| dc.date.available | 2022-12-14T08:21:21Z | |
| dc.date.issued | 2021-06-01 | |
| dc.description.abstract | Moiré pattern is an interference pattern that is formed when two gratings with similar pitch superimposed together. It has exceptional advantage of non-invasive and contactless properties. As a result, it is widely utilized in various applications such as measurement of crack growth, wafer alignment in photolithography, identification of presence of nano components in Transmission Electron Microscope (TEM) images and formation of moiré topogram which is used to diagnose various posture asymmetry disease such as Scoliosis. Recently, Yen and Ratnam proposed Graphical Analysis Method (GAM) as an alternative method to Fast Fourier Transform (FFT) to overcome its limitation of high complexity computations. GAM suffered the loss of fine details of moiré fringes due to morphological dilation operations which eventually penalized the accuracy in spite of its high precision measurements. Hence, stacked denoising autoencoder (SDA) is proposed in this work to overcome the aforementioned problem in GAM while preserving its advantages of high level precision and simple computation. Four types of moiré pattern and their corresponding moiré fringes were used to train the SDA to increase its range of applicability. These moiré pattern include High Frequency Moiré Pattern (HFMP), Low Frequency Moiré Pattern (LFMP), One Fringe Real Moiré Pattern (OFRMP), Three Fringes Real Moiré Pattern (TFRMP) where simulated moiré pattern were obtained by using MATLAB and SOLIDWORKS while real moiré pattern were obtained directly from GAM’s dataset. Reconstructed moiré fringes were evaluated qualitatively on Region of Interest (RoI) of circular moiré fringes and quantitatively by examined on the precision of measurement based on statistical results derived from outputs of curve fitting operation. Qualitative evaluation showed that SDA successfully preserved the fine details of moiré fringes while quantitative evaluation with 95% level of confidence showed that SDA has high level precision measurement that is comparable to GAM such as low error rate of 0.1970 ± 0.2869 point, 0.0027 ± 0.0060 mm, -0.0045 ± 0.0029 mm and -0.0034 ± 0.0020 mm in eccentricity magnitude and 0.0004 ± 0.0332o, -0.9432 ± 0.3023o, -0.3023 ± 1.2872o and 1.5287 ± 0.7821o in eccentricity direction for HFMP, LFMP, OFRMP and TFRMP respectively. The applicability and practicality of SDA were proven in the demonstration of reconstructing random moiré fringes by successfully reconstructed the center of moiré fringes. | en_US |
| dc.identifier.uri | http://hdl.handle.net/123456789/16899 | |
| dc.language.iso | en | en_US |
| dc.publisher | Universiti Sains Malaysia | en_US |
| dc.title | Residual Gratings Removal In Moiré Pattern Using Autoencoder | en_US |
| dc.type | Other | en_US |
Files
License bundle
1 - 1 of 1
Loading...
- Name:
- license.txt
- Size:
- 1.71 KB
- Format:
- Item-specific license agreed upon to submission
- Description: