COLMAP is an open-source algorithm created in recent years. This algorithm creates a 3D reconstruction scene from a set of thousands of 2D images taken by monocular cameras. This reconstruction process, when run on high-quality reconstruction may be computationally heavy on most common computers. In that regard, the goal of this project is to improve the quality of the reconstruction run on lower quality options by using noise reduction methods and completing missing information on surfaces in order to achieve a high-quality reconstruction without using large computational power.
COLMAP was used to create a point cloud of the captured object. Considering the length of time required to recognize features in the data set and match them, a medium-quality reconstruction was chosen. This limits the final result of the point cloud and the mesh created by it. To improve the point cloud we used density-based spatial clustering of applications with noise while optimizing the run time of the algorithm. We were able to reduce the noise and artifacts in the final mesh significantly.
To create and fill the missing points in the plane, we separated the point cloud using known algorithms like RANSAC and K-Nearest-Neighbors and filled in the missing points while accounting for the plane’s geometry. Additionally, we used the symmetry commonly seen in buildings we are trying to reconstruct. We were able to successfully improve the results of the reconstruction and complete data that was missing from the dataset.