Stereo Vision System for A Bin Picking Adept Robot

Abstract

In bin picking applications, robots are required to pick up an object from a pile of stacked or scattered objects placed in a bin. To perform such tasks, identification of the objects to be picked using a vision system is indispensable. In this paper, a stereo vision based automated bin picking system is proposed which identifies the topmost object from a pile of occluded objects and computes its location. The proposed bin picking process consists of two modules namely object segmentation module and object localization module. In the segmentation module, an 'Acclimatized Top Object Threshold' [ATOT] algorithm is proposed for segmentation of topmost object and in the localization module, the location of the object is estimated by computing the 'x', 'y', 'z' coordinates of the object midpoint using a unified stereo imaging algorithm. The validity of the algorithms is experimentally verified for object pick and place operations using the object location coordinates. The developed stereo vision system was implemented and validated for bin pick and place operations on an Adept Cobra 600 Robot.

Improving Robotic Bin-Picking Performances through Human–Robot Collaboration

“…Three-dimensional visual data acquisition is the first challenge of the bin-picking problem chain. For this purpose, various technologies can be used: the most common are 3D laser scanners [11], structured light vision systems, consisting of a projector and one or two cameras [12] and stereo vision systems [13]. In bin-picking applications, the choice of the applied sensor for object registration is strongly correlated with failures, as discussed in [14].…”

A modified eye‐in‐hand stereo visual control for grasping unknown objects via Scara robot

“…Due to the camera configuration, if the camera mounted on the robot end-effector is called eye-in-hand, and if the camera is fixed in the workspace called eye-to-hand [7]. Hema [8] et al developed an intelligent bin-picking method using stereo vision. This system was able to detect the position and distance of the height of the object.…”

Objects Centroid Correlation Using MATLAB's Neural Network Toolbox for Visually-Guided Object Manipulation

“…An automatic process in a manufacturing line always uses one or more robots to make a particular product. If the task is simple and repetitive, a robot using an inflexible program is suitable for work with conveyor systems and reliable fixtures [1]. However, if the task requires adjustable manipulation, such as picking objects from random positions and placing them to specific locations, the robot needs machine vision to aid it with object localization.…”

“…Assuming the objects lie on a horizontal surface, the object localization in a two dimensional Cartesian coordinate system (x, y) needs only a monocular camera [2,3]. If the object localization operates in a three dimensional Cartesian coordinate system (x, y, z), a stereo camera is usually used [1] while a monocular camera can be employed by using special techniques e.g. corresponding features between two successive images acquired from the camera [4].…”

“…Simple techniques for the localization of 2D objects are normally based on using image segmentation and binarization to find their centroids [1,3]. On the other hand, the second layer of a RBF network performs the same function [4].…”