Obtainging 3D structures from Two Uncalibrated Views

Po-Hao Hung Shang-Hong Lai
lai@cs.nthu.edu.tw lai@cs.nthu.edu.tw
NTHU NTHU

Project summary

Structure reconstruction is a popular and interesting problem. Many researches had reported for solving this kind of problems. Because image-based modeling is somewhat lack of information of scene, solving this problem sometimes need to give some constraints or under some particular environments. Full-automatic reconstruction is very hard. In this project, we also give some constraint on our model. Because we'll reconstruct up to affine, we need to apply parallel lines, we will choose those object that has parallel lines to reconstruct, such as human buildings or furniture. We can also give a cubic object in our scene if there aren't enough parallel lines on the reconstruction object. Our system is semi-automatic, we can automatic extract feature points, lines and matching them. But miss-extracting or miss-matching is avoidless, we need to correct it manually.

Project summary

Structure reconstruction is a popular and interesting problem. Many researches had reported for solving this kind of problems. Because image-based modeling is somewhat lack of information of scene, solving this problem sometimes need to give some constraints or under some particular environments. Full-automatic reconstruction is very hard. In this project, we also give some constraint on our model. Because we'll reconstruct up to affine, we need to apply parallel lines, we will choose those object that has parallel lines to reconstruct, such as human buildings or furniture. We can also give a cubic object in our scene if there aren't enough parallel lines on the reconstruction object. Our system is semi-automatic, we can automatic extract feature points, lines and matching them. But miss-extracting or miss-matching is avoidless, we need to correct it manually.

Project summary

Structure reconstruction is a popular and interesting problem. Many researches had reported for solving this kind of problems. Because image-based modeling is somewhat lack of information of scene, solving this problem sometimes need to give some constraints or under some particular environments. Full-automatic reconstruction is very hard. In this project, we also give some constraint on our model. Because we'll reconstruct up to affine, we need to apply parallel lines, we will choose those object that has parallel lines to reconstruct, such as human buildings or furniture. We can also give a cubic object in our scene if there aren't enough parallel lines on the reconstruction object. Our system is semi-automatic, we can automatic extract feature points, lines and matching them. But miss-extracting or miss-matching is avoidless, we need to correct it manually.

Related publications

Last updated on January 8, 2004.

Betpasgiris.vip restbetgiris.co betpastakip.com restbet.com betpas.com restbettakip.com nasiloynanir.co alahabibi.com hipodrombet.com malatya oto kiralama istanbul eşya depolama istanbul-depo.net eşya depolama papyonshop.com şehirler arası nakliyat evden eve nakliye ofis taşıma transfernakliyat.com.tr heceder.org 30 tl deneme bonusu aviator oyunu betexper casino siteleri kamyonet.biz.tr tuzcuoğlu nakliyat malatya temizlik shell aspx shell eşya depolama depolama sarıyer evden eve nakliyat şişli evden eve nakliyat istanbul evden eve nakliyat