Semi-automatic extraction of primitive geometric entities from point clouds

Goussard, Charl Leonard (2001-12)

Thesis (MScEng)--University of Stellenbosch, 2001.

Thesis

ENGLISH ABSTRACT: This thesis describes an algorithm to extract primitive geometric entities (flat planes, spheres or cylinders, as determined by the user's inputs) from unstructured, unsegmented point clouds. The algorithm extracts whole entities or only parts thereof. The entity boundaries are computed automatically. Minimal user interaction is required to extract these entities. The algorithm is accurate and robust. The algorithm is intended for use in the reverse engineering environment. Point clouds created in this environment typically have normal error distributions. Comprehensive testing and results are shown as well as the algorithm's usefulness in the reverse engineering environment.

AFRIKAANSE OPSOMMING: Hierdie tesis beskryf 'n algoritme wat primitiewe geometriese entiteite (plat vlakke, sfere of silinders na gelang van die gebruiker se inset) pas op ongestruktureerde, ongesegmenteerde puntewolke. Die algoritme pas geslote geometriese entiteite of slegs dele daarvan. Die grense van hierdie entiteite word automaties bereken. Minimale gebruikersinteraksie word benodig om die geometriese entiteite te pas. Die algoritme is akkuraat en robuust. Die algoritme is ontwikkel vir gebruik in die truwaartse ingenieurswese omgewing. Puntewolke opgemeet in hierdie omgewing het tipies meetfoute met 'n normaal verdeling. Omvattende toetsing en resultate word getoon en daarmee ook die nut wat die algoritme vir die gebruiksomgewing inhou.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/52449
This item appears in the following collections: