The 3D scanning market is predicted to rise by 10.2% annually through 2026 to a market size of nearly 11 billion dollars. Additionally, in 2018, it was reported that the U.S. Navy saved nearly $2 million on a project by pursuing 3D scanning technologies. These scanners allow the modeling and simulation community to create digital representations of landscapes, vehicles, or other legacy objects for a variety of uses such as virtual training environments. While the use of 3D scanning is growing, problems with the technology still exist, such as erroneous data capture due to overexposure or overly dense sampling. These issues result in point clouds that are unwieldy and challenging to use. Current post-processing techniques for point clouds often require a “guess and check” method of determining proper parameters for cleaning unwanted points or reducing (i.e., downsampling) the number of points in the cloud. This takes a high number of iterations, and significant time to produce a usable point cloud model.
This paper presents a workflow with the purpose of reducing point cloud noise and generating a clean model quickly and efficiently. Using open-source libraries, the potential of mathematically determining suitable parameters for operations performed on point clouds, such as noise filters, was investigated. The outcomes of this research present the relationships between common point cloud post-processing operations and standard point cloud metrics, such as principal component variance and total file size. Determining existing relationships for these parameters allows a post-processing workflow that removes the need for multiple iterations. The resulting workflow was applied to several models and was significantly faster in processing time than traditional iterative processes and produced smaller models in points and file size. Analysis of the results also showed potential for automating this process in future work to further ease these post-processing activities.
Keywords
3D SCANNING
Additional Keywords
Point Cloud, Noise Filtration, Python