Automatic camera calibration and structure recovery of multi-planar scenes using two views
| Autor: | Santés López, Flavio Mario |
|---|---|
| Přispěvatelé: | Vigueras Gómez, Javier Flavio |
| Jazyk: | angličtina |
| Rok vydání: | 2010 |
| Předmět: |
12 [cti]
Computer Science::Computer Vision and Pattern Recognition ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Astrophysics::Instrumentation and Methods for Astrophysics 1 [cti] visión por computadora / segmentación no supervisada / localización de la cámara / reconstrucción de la escena /calibración de la cámara msc:visión por computadora / segmentación no supervisada / localización de la cámara / reconstrucción de la escena /calibración de la cámara |
| Zdroj: | Centro de Investigación en Matemáticas CIMAT Repositorio Institucional CIMAT |
| Popis: | In man-made environments, the presence of planar objects is common. These surfaces are described by walls, doors and many objects such as books, desks and other furniture. Recently, multiple view geometry has been used by the Computer Vision Community as a tool in order to describe the elements of multi-planar scenes and recover the calibration coefficients of the cameras that are used to analyze the three dimensional scene. Calibration deals with the recovering of the focal distance (intrinsic calibration) and localization parameters of a camera (extrinsic calibration). Reconstruction of a multi-planar scene consists of determining the planar equation of each planar structure. The aim of this work is to show how by using linear methods is possible to recover the calibration coefficients and describe the elements in the multi-planar scene using only two different views or images. Related work dealing with camera localization and structure recovery presents non-linear formulations that need an initialization guess in order to solve these problems. Our work introduces a multi-linear form that captures all the coefficients needed to solve the calibration and structure recovery problems. This work presents a novel approach to solve the calibration and structure recovery problems, based on linear methods that exploits geometrical and algebraic constraints induced by rigidity and planarity in the scene. Linearity allows our approach to be suitable for real-time camera calibration and scene reconstruction. Furthermore, we do not compute the two-view fundamental matrix. Therefore, we do not face stability problems commonly associated with explicit epipolar geometry computation. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|