Abstract
In this paper the authors presents a vision system that has developed to aid bin picking tasks. The intention of the proposed system is to detect and identify randomly piled workpieces in the bin. The strategy described here is based on two steps: detecting workpieces and identifying candidates. The geometric feature was applied to obtain candidates and multi-feature fusion based on SVM was employed to find pickable workpieces. Experimental results demonstrate the good performance of the proposed approach and the system is robust against noise and illumination.