A Simple Method to Recover 3-D Rigid Structure from Motion using SIFT, RANSAC and the Tomasi-Kanade Factorization
Traditionally two frames are used to estimate the 3-D structure, while recent approaches have made use of a long sequence of frames. The latter gives a better recovery of a structure because it amasses temporal information over time. Tomasi-Kanade factorization also assumes all features to be visible throughout the entire image stream. This results in a dense 2-D cloud and therefore allows full recovery of the entire 3-D structure.
This paper addresses the problem of 3-D structure reconstruction from motion by using the Tomasi-Kanade factorization method applied to a sequence of frames. Orthographic projection and rigidity is assumed and the singular value decomposition technique is used to factor the measurement matrix (W) into two matrices which correspond to the objects 3-D structure (S) and camera rotation (R) respectively. To construct W, feature correspondences are established by applying a SIFT tracker following which RANSAC is used to discard the false matches detected by the SIFT tracker. The 3-D point-cloud (S) is converted to a mesh of triangles by connecting the nearest three neighboring points. Finally, the mesh of triangles is rendered by means of the Plyview from CyberWare to achieve the final 3-D structure.
Keywords: 3-D Structure, Tomsi-Kanade Factorization Method, Singular Value Decomposition (SVD), SIFT Tracker, RANSAC.
Download Full-Text
ABOUT THE AUTHORS
Samir H. Abdul-Jauwad
Samir H. Abdul-Jauwad holds a PhD (1985) in Electrical Engineering from the University of Sheffield, England, U.K. He got his Bachelor (1973) and Master degree (1976) in Electrical Engineering from King Fahd University of Petroleum & Minerals (KFUPM), Dhahran, Saudi Arabia. He is currently a Professor of Electrical Engineering at King Fahd University of Petroleum & Minerals (KFUPM), where his teaching and research interests include Signal and Image Compression, Image Processing, Wavelets, Satellite Communications. He is a member of KFUPM Scientific Council, and also a Senior Member of IEEE.
Rehmat Ullah
Rehmat Ullah graduated from University of Engineering and Technology Peshawar, Pakistan with a BSc (2004) and MSc (2008) degrees in Computer Systems Engineering. He also obtained his second MSc (2011) from Lahore University of Management Sciences, Lahore, Pakistan, in Computer Engineering with concentration in Image and Video Coding. He is currently a lecturer in Department of Computer Systems Engineering at University of Engineering and Technology, Peshawar, Pakistan. His research interests include Image and Video Coding, Image Analysis, and Computer Vision.
Farman Ullah
Farman Ullah did his BSc (2006) in Computer Systems Engineering from University of Engineering and Technology, Peshawar, Pakistan and his MSc (2011) in Computer Engineering from Center for Advanced Studies in Engineering, Islamabad, Pakistan. He is pursuing his PhD from Korea Aerospace University, South Korea. He worked as Telemetry Engineer at Advanced Engineering Research Organization, Pakistan for four years. Currently he is working as a lecturer in Department of Electrical Engineering at COMSATS Institute of Information Technology, Attock, Pakistan.
Samir H. Abdul-Jauwad
Samir H. Abdul-Jauwad holds a PhD (1985) in Electrical Engineering from the University of Sheffield, England, U.K. He got his Bachelor (1973) and Master degree (1976) in Electrical Engineering from King Fahd University of Petroleum & Minerals (KFUPM), Dhahran, Saudi Arabia. He is currently a Professor of Electrical Engineering at King Fahd University of Petroleum & Minerals (KFUPM), where his teaching and research interests include Signal and Image Compression, Image Processing, Wavelets, Satellite Communications. He is a member of KFUPM Scientific Council, and also a Senior Member of IEEE.
Rehmat Ullah
Rehmat Ullah graduated from University of Engineering and Technology Peshawar, Pakistan with a BSc (2004) and MSc (2008) degrees in Computer Systems Engineering. He also obtained his second MSc (2011) from Lahore University of Management Sciences, Lahore, Pakistan, in Computer Engineering with concentration in Image and Video Coding. He is currently a lecturer in Department of Computer Systems Engineering at University of Engineering and Technology, Peshawar, Pakistan. His research interests include Image and Video Coding, Image Analysis, and Computer Vision.
Farman Ullah
Farman Ullah did his BSc (2006) in Computer Systems Engineering from University of Engineering and Technology, Peshawar, Pakistan and his MSc (2011) in Computer Engineering from Center for Advanced Studies in Engineering, Islamabad, Pakistan. He is pursuing his PhD from Korea Aerospace University, South Korea. He worked as Telemetry Engineer at Advanced Engineering Research Organization, Pakistan for four years. Currently he is working as a lecturer in Department of Electrical Engineering at COMSATS Institute of Information Technology, Attock, Pakistan.
