Authors: Bernd Scholz-Reiter, Hendrik Thamer
Unloading of standard sea containers is a time- and cost-expensive process step within global supply chains. Until now, existing solutions for automatic unloading are limited to goods of cubical shape. The main challenge in developing a robotic system for universal logistic goods is object recognition of goods that differ in size, shape and orientation. A common approach for object recognition systems is the comparison of sensor data to a model database. This paper presents a simulation system of sensor images for universal logistic goods. The objective of simulating sensor images is generating a model database and create training data for classification. Additionally, sensor images of complete packaging scenarios can be simulated for evaluating the object recognition algorithm with ideal sensor data that does not contain measurement noise or other inaccuracies. The simulation system is explained in detail by modelling of cubical objects.