Recognition and handling of clothes with different pattern by dual hand-eyes robotic system

Abstract

Recently, robots have been used in clothing industries for mass production with countless merits. However, there remain many challenges for robots in recognition, pose (position and orientation) detection operations, especially when the working object is deformable and every working object has unique shape and color. In this paper, pose detection of clothes through 3D recognition is proposed for the task in which the manipulator recognizes clothes, estimates relative pose and performs pick and place function. In proposed cloth recognition system, a variety of models of different clothes with unique shape and color are generated as BMP (bit map file) format extracted from the camera. Using the photograph model, recognition of cloth is performed by using images from the two cameras that are fixed at the end effector of the robot arm. 12 varieties of different clothes samples are used for this experiment. The pose of individual cloth is estimated by Genetic Algorithm (GA). 1000 times recognition and handling experiment has been executed, having shown the effectiveness of proposed Photo-Model-Based pose recognition system.