Applying computer vision techniques to perform semi-automated analytical photogrammetry

The purpose of this research is to show how common computer vision techniques can be implemented in such a way that it is possible to automate the process of analytical photogram-metry. This work develops a workflow that generates a sparse three-dimensional point cloud from a bundle of images using SIFT, RANSAC, and a sparse bundle adjustment along with basic photogrammetric methods. It then goes on to show how the output of the sparse reconstruction method can be used to generate denser three-dimensional point clouds that can be facetized and turned into high resolution three-dimensional models. This workflow was successfully tested on a five image dataset taken with RIT's WASP imaging sensor over the Van Lare wastewater treatment plant in Rochester, NY.