GEODIM: A Semantic Model-Based System for 3D Recognition of Industrial Scenes

Keeping an inventory of the facilities within a factory implies high costs in terms of time, effort, and knowledge, since it demands the detailed, orderly, and valued description of the items within the plant. One way to accomplish this task within scanned industrial scenes is through the combination of an object recognition algorithm with semantic technology. This research therefore introduces GEODIM, a semantic model-based system for recognition of 3D scenes of indoor spaces in factories. The system relies on the two aforementioned technologies to describe industrial digital scenes with logical, physical, and semantic information. GEODIM extends the functionality of traditional object recognition algorithms by incorporating semantics in order to identify and characterize recognized geometric primitives along with rules for the composition of real objects. This research also describes a real case where GEODIM processes were applied and presents its qualitative evaluation.