KLT Bin Detection and Pose Estimation in an Industrial Environment

Aleksei Beloshapko, Christian Knoll, Bilel Boughattas, Vladimir Korkhov

Abstract: In order for Automated Guided Vehicles (AGV’s) to handle KLT bins (Kleinladungsträger, Small Load Carrier) in a flexible way, a robust bin detection algorithm has to be developed. This paper presents a solution to the KLT bin detection and pose estimation task. The Mask R-CNN network is used to detect a KLT bin on color images, while a simple plane fitting approach is used to estimate its 5DoF position. This combination gives promising results in a typical use case scenario when the KLT bin is aligned with the camera view.