Range Data Segmentation with Principal Vectors and Surface Types

A new method for segmenting range data including curved surfaceisproposed.Themethod isbasedonrobust principal vectors calculation using ISL-primary-axis method. First a normal vector and principal vectors are calculated for each data point. Next, maximum and minimum curvatures are calculated refemng principal vectors.The surfacetype of each point isdeterminedwith values of the curvatures. Then the surface is segmented by discontinuity of depth, normal vectors, principal vectors and/or surface types. The discontinuityinformationis storedin a special boundaryimageto reservethereliabilityof each boundary points. Each small region in the segmented image is merged into one of the neighboring regions. The border of the target region is segmented for each neighboring region and each segment is scored with length and reliability which reflects the discontinuity reason. The border segment with the highest score is erased. Theexperimentalresultforactualrangedatais shown.

• Type: Conference
• Year: 1996
• Where: MVA
• Authors: Takashi Yoshimi, Yoshihiro Kawai, Fumiaki Tomita
• Keywords: MVA 1996, MVA 2007, Normal Vectors, Principal Vectors, Principal Vectors Calculation