Vision Based Bin Picking for Industrial Robot

Abstract

Vision based object recognition has applied to automate a production line. Bin picking in cell production method has considered intense research area that performs pick and place tasks for randomly piled objects from bins. However, it has challenged to pick an object, which has characteristic of reflecting material. In this paper, we propose vision based bin picking system using transformed shape parameter detection of object features, which are calculated with 2D image. The pose of the object is estimated with the use of maximal axis detection and center position of object. Worktable has configured to reduce highlight in the object image and distortion image has compensated using strengthen image preprocessing. The proposed method has evaluated using dual arm robot, which has built based on real bin-picking work environment.