Robust vision-based detection and grasping object for manipulator using SIFT keypoint detector

Abstract

The ability for a manipulator to detect and grasp an object accurately and fast is very important. Vision-based manipulator using stereo vision is proposed in this paper in order able to detect and grasp an object in a good manner. We propose a framework, fast algorithm for object detection using SIFT(Scale Invariant Features Transform) keypoint detector and FLANN (Fast Library for Approximate Nearest Neighbor) based matcher. Stereo vision is used in order the system knows the position (pose estimation) of the object. Bayesian filtering implemented in order to reduce noise from camera and robust tracking. Experimental result presented and we analyze the result.