Title :
A novel method for segmentation of cones and cylinders from geometrically fused depth maps
Author :
Ng, I. ; Illingworth, J. ; Jones, G.
Author_Institution :
Surrey Univ., Guildford, UK
Abstract :
The difficult problem of extraction of cylindrical and conic surfaces from range data is considered. A new method based on taking pairs of surface patches and generating and accumulating curved surface parameters is presented. Parameter clusters are identified by a hierarchical scheme using an unsupervised clustering method. The method is able to work with sparse or dense data and does not require image format data. It can therefore be applied to geometrically fused 3D data taken from multiple views. The method is shown to work well in complicated scenes which include significant occlusion. The benefits of multiple view fusion are shown by experiment
Keywords :
feature extraction; image segmentation; parameter estimation; sensor fusion; cones; curved surface parameters; cylinders; dense data; experiment; geometrically fused 3D data; geometrically fused depth maps; hierarchical scheme; image segmentation; multiple view fusion; multiple views; occlusion; parameter clusters; parameter estimation; range data; sparse data; surface patches; unsupervised clustering method;
Conference_Titel :
Image Processing and its Applications, 1995., Fifth International Conference on
Conference_Location :
Edinburgh
Print_ISBN :
0-85296-642-3
DOI :
10.1049/cp:19950718