Abstract: Robots are becoming increasingly important to human life.  Apart from their long well established achievements in industrial applications, one that is currently emerging is domestic robot application.  However, robotic vision using visual information is still considered unsuccessful.  Recovering 3D shape from its 2D image by means of stereo fails when the object is textured less.  Therefore, methods such as structured light stripe, or tactile sensor are commonly used in order to resolve ambiguity.  In contrast, vision remains the richest human source of information.  For this reason, this research tries to explore this visual information to recover an object's depth in order for a robot to manipulate and avoid obstacles.  A few preliminary results of stereo reconstruction with homography and Shape from Shading (SFS) are shown.

About the Speaker: Effendi is a postrgraduate research student with the department of Electrical and Computer Systems Engineering at Monash University Clayton.

Location Change:  This seminar will be held in Room 132, Building 72