Mathematics of projective versus perspective collineations in camera orientation and calibration.

This paper presents a preliminary result of ongoing research on monitoring natural disasters induce structural infrastructures damages by using off-the-shelf digital camera. It's Orientation and calibration algorithm that steamed from a more general pinhole camera model can be varied into two distinct approaches: projective collineation and perspective collineation which widely embraced by computer vision community and photogrammetric community respectively. This research investigates reliabilities of those varying methods for computing orientation and calibration of the captured images of a bridge. Some photographs of the surveyed bridge were processed using two proprietary software which each employs different collineation algorithm. Some findings reveal that the projective collineation algorithm uses more robust matrix operation but requires more parameters and consequently need more correspondent points on each captured image. On the other hand, fewer parameters and common points are required for the more rigorous perspective model although it requires a nonlinear relationship of matrix operations. Results show that the projective collineation algorithm gives higher model stability but requires more stable camera and optics. Furthermore, figures and numbers are provided to support our findings that both algorithms produce approximately equal precision of the orientation and calibration parameters. [ABSTRACT FROM AUTHOR]