3D OBJECT CATEGORIZATION OF LOGISTIC GOODS FOR AUTOMATED HANDLING

摘要

The automated handling of universal logistic goodsrnthrough robotic systems requires suitable and reliablernmethods for categorizing different logistic goods. Theyrnmust be able to detect the pose of different types and sizesrnof logistic goods in order to identify possible grippingrnpoints or for selecting a suitable gripping system for therndetected object type. For this purpose, Time-of-Flight orrnStructured Light sensors can deliver a dense 3Drnrepresentation of the investigated scenario. This paperrnpresents a 3D object categorization system for logisticrngoods based on synthetically generated model data. Werngenerate the model data by using a sensor simulationrnframework for different TOF-sensor types. Thernframework creates point clouds of self-defined geometricrnmodels of logistic goods or CAD data. Afterwards, we usernthese synthetic point clouds for generating a suitablernmodel database offline. In order to evaluate our approach,rnwe describe the synthetic point clouds by global pointrnfeature description techniques to distinguish betweenrndifferent types of logistic goods. Finally, we evaluate ourrnconcept with real sensor data from different logisticrngoods.