A region-splitting approach to the segmentation of partially occluded polyhedral objects

Kuo Chin Fan, Chia Yuan Chang

Research output: Contribution to journalArticlepeer-review

Abstract

Object segmentation is a key task in the recognition of 3-D partially occluded objects. In this paper, an object segmentation algorithm is presented to extract partially occluded polyhedral objects from a scene. The visual scene is composed of a group of 3-D overlapping polyhedral objects which can be represented by straight-line drawings projected on a 2-D image plane. The goal is to decompose the visual scene into several 3-D objects by extracting a line drawing from the partially occluded scene. The criteria used for segmenting a visual scene into its composing objects include junction-type interpretation and region homogeneity. The proposed novel region-splitting approach is adopted instead of the traditional region-merging approach to do the segmentation job. The proposed approach not only segments a line drawing into separate objects but also finds out the inter-relations (occluding/occluded) of objects in terms of the occlusion graph. The concavity and convexity of each separated object can be determined from the occlusion graph. The segmented result can be further used as a basis for partial matching in recognizing partially occluded polyhedral objects.

Original languageEnglish
Pages (from-to)57-64
Number of pages8
JournalEngineering Applications of Artificial Intelligence
Volume6
Issue number1
DOIs
StatePublished - Feb 1993

Keywords

  • 3-D object recognition
  • Attribution matrix
  • occlusion graph
  • region adjacency graph

Fingerprint

Dive into the research topics of 'A region-splitting approach to the segmentation of partially occluded polyhedral objects'. Together they form a unique fingerprint.

Cite this