This paper presents a measurement solution that recovers 3D information of
objects from a panoramic image. In case of having GPS position only, the panoramic
posture at the shooting time can be calculated by image match and bundle adjustment.
Thus we can use intersection to obtain the objects’ coordinates. As large distortion in
panoramic is an obstacle in image match, an affine invariant feature matching algorithm
is used for image match, and a random sample consensus algorithm used to eliminate
coarse errors, giving better point match in terms of both quantity and quality. For the unique geometric feature of panoramic image, the conventional photogrammetric bundle
adjustment is improved and used to solve posture of panorama. Finally, an intersection
method is proposed to build normal equations through transformation of space straight
line equations before adjustment. Compared with the conventional method, the proposed
algorithm has higher accuracy. Experimental results show feasibility of the scheme. It can
accurately measures object coordinates in panoramic images.