REN-CHYUAN LUOW. S. YangY. H. Kim2018-09-102018-09-101988-01http://scholars.lib.ntu.edu.tw/handle/123456789/342809https://www.scopus.com/inward/record.uri?eid=2-s2.0-85069175934&doi=10.1109%2fIECON.1988.660418&partnerID=40&md5=18c520001e60c1e296812144896d1c31The objective of this paper is to develop an innovative 3-D objects recognition algorithm which solves correspondence problem and reduces computational cost for recognition. Correspondency between features of unknown object and its model and complicated computation for recognition seem to be a bottleneck of current 3-D object recognition. Three-dimensional generalized Hough transformation is introduced which can be applied to general 3-D objects, and reduces both the computation time and data base memory requirements. Data base is constructed in off-line fashion based on the view-point independent characteristics from the known 3-D objects using generalized 3-D Hough transformation. In the recognition process, view-. point independent characteristics from unknown objects are obtained and used as a key for searching the data base for the best candidate. To test the algorithm, various types of objects are investigated by use of 3-D range data. © 1988 IEEE.Accumulator array; Binary search tree; Data base; Gaussian curvature; Hough transformation; Mean curvature; Normal; Recognition; Reference point; Three dimensionAgricultural robots; Binary trees; Computer vision; Industrial electronics; Object recognition; Robotics; Trees (mathematics); Accumulator array; Binary search trees; Gaussian curvatures; Hough Transformation; Mean curvature; Normal; Recognition; Reference points; Three dimensions; Metadataconference paper10.1109/IECON.1988.660418