||Built infrastructure point cloud data cleaning: an overview of gap filling algorithms
||Rashidi A,Brilakis I,Vela P
||Video captured from infrastructure scenes can be used to generate point cloud data (PCD) as a potential solution for acquiring spatial information of built infrastructure - however, video based PCD is incomplete and includes gaps, outliers and poor/non-reconstructed areas. This phenomenon has a negative impact on both visualization and measurement practices and is mainly caused by a number of reasons including insufficient coverage of all views while videotaping the scene, lack of sufficient features on uniform surfaces and possible errors in calibration, matching and optimization algorithms. To tackle this issue, researchers suggested various post processing algorithms for reconstructing missing surfaces and filling gaps/holes. This paper provides an overview on these algorithms summarize their properties in terms of efficiency, ability to work in complex geometry settings and running time. As the comparison study, three most common hole filling algorithms: MSL, GG and RFR were implemented and tested on a number of real built infrastructure scenes as the case studies. Number of generated 3D points for filling the gaps, proper distribution of points on covered surfaces and running time are three major comparison metrics has been taken into account. Results indicate that in general PML outperforms other algorithms on both flat and curved surfaces.
|Year of publication:
||Built infrastructure,triangulation,gap,Point Cloud Data,surface reconstruction
Rashidi A,Brilakis I,Vela P (2013).
Built infrastructure point cloud data cleaning: an overview of gap filling algorithms. CONVR 2013,