Circular image block method

Jussi Heikkinen

In this research a new method is developed for close range photogrammetric measuring conditions. The novel method is designed for conditions where object space is around the object and photogrammetric measurements are to be done inside the object space.

This novel method, which takes care of the imaging geometry as well as constructing the whole model in a same coordinate system without any additional coordinate transformations, is called the Circular Image Block method.  As in all image based measurements, position and orientation of cameras respect to other camera poses must be known or have to be determined in an estimation process.  This guarantees that the 3-D model determined by intersecting image rays from multiple cameras will have correct shape and size.  In case of stereo imaging the cameras are installed to look in parallel and the distance of projection centers along their common x-axis is determined through calibration.  But there are cases like aerial photography where you only know the approximate pose and orientation of images.  The exact pose and orientation have to be determined by estimation.  In order to do that you have to measure at least three correspondent points on images, where correspondent image points mean image points, which are projections of the same 3-D point in object space.  That is the reason why subsequent images must partly overlap.  Three points are minimum for the mathematical solution as we have three pose parameters and three orientation angles to be solved.  From these calculations we get the relational pose and orientation of cameras and we can reconstruct the object in correct shape.  After this only the scale for object is still unknown.  For the scale we have to know at least one distance in object space.  If in addition we want to reconstruct the object in a particular coordinate system, we also need to know three object points in that coordinate system.