Detecting cylinders in 3D range data using model selection criteria

Abstract

In this paper, we use a model selection criterion to decide whether two cylinders should be merged as a single cylinder or they should be left separated. We compare and evaluate an extensive number of different model selection criteria for this purpose and examine which factors can affect their performance. We conclude that SSC, GIC, MCAIC and CAIC have a better performance (for this particular application) compared to the other criteria.