Automated Item Picking for fashion articles using Deep Learning
Konferenz: ISR 2020 - 52th International Symposium on Robotics
09.12.2020 - 10.12.2020 in online
Tagungsband: ISR 2020
Seiten: 8Sprache: EnglischTyp: PDF
Weisenboehler, Moritz; Wurll, Christian (Karlsruhe University of Applied Sciences, Karlsruhe, Germany)
With rising sales of e-commerce, the requirements for order picking processes increase regarding flexibility, quality and costs. The usage of Automated Item Picking (AIP) can help to face these challenges by utilizing robotic systems and computer vision. This work presents an AIP solution for the sports and fashion industry using a specialized gripper and a Deep Learning based computer vision system. A Convolutional Neural Network (CNN) is combined with depth image based post-processing algorithms to identify and localize various apparel articles and shoeboxes in a load carrier. The developed gripping system combines a vacuum and a clamp gripper for item handling. The system performance is evaluated for industrial problem constellations regarding accuracy, cycle time and reliability.