Izračun položaja ljudskog tijela u prostoru

scientific master's theses

• Author: Tomislav Pribanić
• Mentor: Mario Cifrek (mentor)
• Committee member: Mario Cifrek (član povjerenstva)
• Committee member: Stanko Tonković (član povjerenstva)
• Committee member: Vladimir Medved (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; (Department of Electronic Systems and Information Processing); Zagreb
• Defense date and country: 2001-07-04, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing; Data Processing
• Universal decimal classification (UDC): 004 - Computer science and technology. Computing. Data processing

Abstract

Predmet rada bila je rekonstrukcija položaja ljudskog tijela u prostoru. Objašnjena su osnovna načela nastajanja slike tijekom procesa fotografiranja koja služe kao temelj opto-elektroničkih sustava za 3D rekonstrukciju. Proces fotografiranja je analitički opisan funkcijom modela kamere. Proračun parametara modela kamere, tzv. kalibracija kamere, je izvršen upotrebom kalibracijske strukture koja može biti trodimenzionalna, dvodimenzionalna ili jednodimenzionalna. Ta dimenzionalnost, uz (ne)linearnost funkcije modela kamere, najviše utječe na složenost kalibracije i rekonstrukcije u smislu proračuna i samog korištenja dotičnog sustava za 3D rekonstrukciju. Jasno je istaknuta prednost nelinearnog modela kamere ispred linearnog. Pokazano je da pozitivan učinak na točnost rekonstrukcije uslijed smanjenja volumena kalibracije i/ili povećanja rezolucije slike je očit prije svega za nelinearni model kamere. Nadalje, broj kalibracijskih točaka iznad kojeg nema bitnog poboljšanja točnosti rekonstrukcije, niti za linearni niti za nelinearni model, je oko 30. Utvrđeno je da optimalni kut između stereo para kamera iznosi oko 90°. Teorijski opis i praktični rezultati su potvrdili Levenberg-Marquardt metodu kao vrlo dobar izbor za rješavanje nelinearne metode najmanjih kvadrata prilikom korištenja nelinearnih modela kamera. Primjer rekonstrukcije pokretne slike istaknuo je potrebu za automatizacijom procesa kalibracije i rekonstrukcije, te uporaba dovoljnog broja kvalitetnih video kamera kako bi se onemogućilo nestajanje markera, tj. nemogućnost njihova raspoznavanja na slici. Kao alternativa korištenja kalibracijskog kaveza pokazan je i način rekonstrukcije tzv. virtualnim 3D objektom.
Utemeljen je vlastiti nelinearni model kamere za kojeg su razvijeni algoritmi u Matlabu, te koji zajedno sa napravljenim kalibracijskim kavezom čini sustav za 3D rekonstrukciju. Točnost razvijenog sustava je sumjerljiva s komercijalno dostupnim sustavima.

Abstract (english)

The objective of the work was a reconstruction of the human body position in the space. Basic principles describing points projections from 3D space to image plane, during the process of imaging, were described. It serve as foundation for opto-electronic 3D reconstruction systems. Imaging itself were presented in analytical form through camera model function. Calculation of camera model parameters, i.e. camera calibration, was done by the calibration structure which can be 3, 2 or 1 dimensional. Dimensionality itself along with (non)linearity of camera model were proven to be crucial factors determining complexity of calibration/reconstruction calculation and use of the system as such. Superiority of non-linear camera models was clearly pointed out. It has been shown that decrease in calibration volume and/or increase in picture resolution improves accuracy particularly in the case of non-linear camera model. Also sufficient number of calibration points for both linear and non-linear models is 30. It appeared that optimal angle between camera optical axis is around 90°. Practical results confirmed theoretical expectations that Levenberg-Marquardt method is very efficient mean for solving non-linear square problems present when non-linear camera model is used. An example of moving scene reconstruction revealed necessity for automatic calibration/reconstruction accompanied with high quality cameras preventing markers from disappearing and blurring pictures. As an alternative to traditional calibration cage, reconstruction by the means of 3D virtual object was exercised.
As final result of work non-linear camera model was established, very practical 3D calibration structure was constructed and corresponding Matlab routines for 3D reconstruction were developed. Thus, emerged 3D reconstruction system offers reconstruction accuracy as good as other todays present commercial reconstruction systems.

• Keywords: 3D rekonstrukcija; model kamere; marker; Levenberg-Marquardt; kalibracija; 3D virtualni objekt; izobličenje leća; DLT; postura
• Keywords (english): 3D reconstruction; camera model; Levenberg-Marquardt; calibration; 3D virtual object; lens distortion; DLT; posture
• Language: croatian
• URN:NBN: urn:nbn:hr:168:008599
• Study programme: Title: Postgraduate master programe in computing; Study programme type: university; Study level: postgraduate; Academic / professional title: magistar inženjer računarstva (magistar inženjer računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2019-04-18 13:17:27