Model-based autonomous interior orientation

The model-based approach to autonomous interior orientation is an entirely novel approach. It is driven by the simple structural description that one can build for a fiducial mark. In this research we have focused on recognizing and measuring the fiducial marks automatically. This is motivated by the desire to establish an automation chain in digital photogrammetry, starting with the interior orientation. We benefit from a simple fact that fiducial marks are artificial objects projected onto the film during exposure time. Most every fiducial has a simple, regular shape. This invites us to represent the fiducial marks as a structural description and to detect the structural elements in the image. The CAD design of the fiducial mark and the optical parameters of the projection lens are used to build a geometric model for the fiducial mark. Edge detection is performed to generate image primitives. The Hough Transform is executed over the edge image for identification and approximate localization. Least Squares Adjustment is used for precise localization and the affine transformation is used to compute the transformation parameters. The results of the Hough Transform, represented in the accumulator array, are analyzed via quadratic conic section fitting. The analysis of the fitted conic section renders Valuable information regarding the surface complexity in terms of noise and the surrounding structures to the fiducial mark. A fast version of the Hough Transform, for the peak detection, is implemented via histogram peak detection. The developed technique is capable to handle noisy and partially detected or missing information of the fiducial mark. The collected evidence in terms of hypothesis generation, verification, and validation allows us to define a percept sequence to reason about the recognition of the fiducial mark.