ANALYSIS AND CORRECTION OF NONLINEARITY ERROR INDUCED BY OPTICAL DEVICES IN DIGITAL FRINGE PROJECTION SYSTEM
Digital fringe projection technology is one of the most popular 3D imaging technologies because of its relatively easy and inexpensive set up. However, this technology comes with its problems. The use of commercial optical devices in the technology produces a response that has nonlinear characterist...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/56857 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Digital fringe projection technology is one of the most popular 3D imaging technologies because of its relatively easy and inexpensive set up. However, this technology comes with its problems. The use of commercial optical devices in the technology produces a response that has nonlinear characteristics (gamma correction) which will make the reconstruction results have low accuracy. This final project aims to study several existing and quite popular nonlinearity error compensator methods, namely the Double-3PSI (DPSI) and Hilbert-PSI (HPSI) methods, and compare them to one of the recent methods, the 3-sigma criterion method. In the process of comparing the three methods, this final project uses a qualitative approach to four objects with varying levels of surface complexity and a quantitative approach to a flat object whose accuracy is represented by the RMSE value.
Qualitatively, the three methods proved successful in eliminating the error component due to nonlinearity. However, for the HPSI method, it can't handle complex objects as well as DPSI and 3sigma. On the other hand, the HPSI method can provide improvements to the image's error due to shadow in the image acquisition process that cannot be compensated by the DPSI and 3 sigma methods.
Then, quantitatively, a comparison of the reconstruction results was carried out on flat objects with variations in frequency (3, 5, and 7 cyc/100px) as well as variations in steps (3, 5, 7 STEP). For each change in frequency, the number of steps used is the same – vice versa. As a result, the 3 sigma method showed the best performance with RMSE values (mm): 0.121, 0.129, 0.080 in each frequency variation, respectively. Then, for the variety of steps, the HPSI method consistently succeeded in providing the minimum RMSE values with values (mm): 0.136, 0.203, and 0.072.
?
|
|---|