Light plane calibration from a single view

measurements of physical dimensions. The analysis of the images is substantially simplified by the fact that the depth of an observationis uniquely defined by its position in the image. This article shows that, by defining the object coordinate system appropriately, thetransformation from object to image coordinates can be expressed as a 3 by 3 homogeneous transformation matrix, also when theobject is rotated relative to the light plane. The transformation matrix can be computed from a minimum of four calibration pointson a known, roof-shaped calibration object observed from a single view. The inverse matrix gives object coordinates as a function ofimage coordinates. The 3D camera coordinates of an observation can be obtained from its image coordinates by a homogeneous 4 by 3transformation matrix. The article shows how the elements of this matrix can be extracted from the 3 by 3 matrix. Attractive featuresof the calibration method is that the orientation of the calibration object and the light plane can be chosen freely, that only a simplecalibration object is needed, that redundant calibration points are easily included in the computation and that a single image is sufficientto determine all the calibration parameters.