University of Twente Student Theses

Login

Rapportage about testing of the 3D motion correction

Kas, G.R. (2017) Rapportage about testing of the 3D motion correction.

[img] PDF
647kB
Abstract:The research in this paper presents an approach of motion correction for movements of the face during three-dimensional (3D) face scanning. As CMOS (Complementary Metal Oxide Semiconductor) scanners are more commonly used nowadays than CCD (Coupled-Charge Device) scanners, because CMOS scanners have higher readout speeds. These CMOS scanners produce the so called rolling shutter effect, which is introduced due to the fact that the scanned object is moving faster than the time it takes to produce a full scan on the object. This rolling shutter effect affects face recognition algorithms, as the resulting facial scan has this rolling shutter artefact introduced. The current method that is used to compensate for the motion of a face is designed due to the symmetrical structure of the face. This correction is only available on the horizontal lines of the face due to having this symmetrical property on the axis of the horizontal lines of the face. The method introduced in this research is mainly focused on the (x,y)-plane to estimate a human balance motion during the scan, as it is expected that there will be no intentionally movements to enhance the failure of a face recognition. As resources of the rolling shutter effect are limited, a test set has to be made in which the phenomenon of the moving shutter effect during a three-dimensional (3D) face scan appears. A face is placed inside a grid of the 'camera' and moved in certain directions during a scan from the top to the bottom. Which results in new faces that consist of the phenomenon described above. These faces that consist of the motion will be compensated for with an motion correction algorithm. The deformed face is moved in front of a 'camera' such that they are moved to the right place compared to a reference frame. Movements in the x-direction are recognized and compensated for at the program. The difference in standard deviation of the x is 0,0894mm. So the original face and the reconstructed face seem to be pretty similar.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Electrical Engineering BSc (56953)
Link to this item:https://purl.utwente.nl/essays/74270
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page