Detection of Parcel Boxes for Pallet Unloading Using a 3D Time-of-Flight Industrial Sensor
Riccardo Monica, Jacopo Aleotti, Dario Lodi Rizzini
Abstract
This work presents a 3D vision system for automatic detection of cardboard parcel boxes of known size, located on the top layer of a pallet of known height. An industrial Time-of-Flight (ToF) sensor is adopted that can operate in different illumination conditions thanks to the use of an infrared light source. The perception system is intended for application in industrial warehousing for end-of-line operations, like robot depalletizing. The proposed method does not assume any predefined layout of parcels and, therefore, it can work even with an incomplete layer of misaligned cardboard boxes. The developed solution first extracts all possible object hypotheses, then an optimization problem is solved, based on a genetic algorithm, to exclude conflicts. Experiments have been performed on a real dataset including complex configurations of tightly packed parcels.