opencv - Why do we need to move the calibration object for pinhole camera calibration? -



opencv - Why do we need to move the calibration object for pinhole camera calibration? -

is there particular reason why need multiple poses (e.g. varying z or rotation) obtain focal length , principal point photographic camera matrix? in other words, sufficient calibrate pinhole photographic camera single pose? i.e. keeping location of calibration object (let's standard checkerboard) constant?

i assume asking in context of opencv-like photographic camera calibration using images of planar target. reference algorithm used opencv z. zhang's classic paper . give-and-take in top half of page 6 shows n >= 3 images necessary calibrating 5 parameters of pinhole photographic camera matrix. imposing constraints on parameters reduces number of needed images theoretical minimum of one.

in practice need more various reasons, among them:

the need have plenty measurements overcome "noise" , "random" corner detection errors, while using practical target well-separated corners. the difference between measuring info , observing (constraining) model parameters. practical limitations of physical lenses, e.g. depth of field.

as illustration sec point, ideal target pose calibrating nonlinear lens distortion (barrel, pincushion, tangential, etc.) frontal-facing, covering whole field of view, because produces big number of well-separated , aligned corners on image, approximately same grade of blur. however, worst pose can utilize in order estimate field of view / focal length, purpose need observe important perspective foreshortening.

likewise, possible show location of principal point constrained set of images showing vanishing points of multiple pencils of parallel lines. of import because location inherently confused component parallel image plane of relative motion between photographic camera , target. vanishing points help "guide" optimizer's solution toward right one, in mutual case target translate w.r.t camera.

opencv camera computer-vision computational-geometry camera-calibration

Comments

Popular posts from this blog

formatting - SAS SQL Datepart function returning odd values -

c++ - Apple Mach-O Linker Error(Duplicate Symbols For Architecture armv7) -

php - Yii 2: Unable to find a class into the extension 'yii2-admin' -