In　this　paper,　a　new　planning　approach　for　automatic　reconstruction　of　unknown　3D　objects　with　occlusion　is　proposed.　During　reconstruction　process,　the　contour　of　occlusion　is　predicted　reasonably　firstly.　Secondly,　the　optimal　spatial　position　of　the　vision　system　would　be　determined　where　the　vision　system　can　obtain　the　maximum　visible　area.　Finally,　the　reconstruction　processes　of　the　real　objects　are　experimented　to　validate　the　feasibility　of　the　proposed　method.　©　2010　IEEE.