Challenge: For precise depallatisation, the product centres position within ± 1 mm and ± 2 rotation accuracy has to be sent to the robot. The dimension of the pallet is 1200×800 mm, while the pallets are stored in multiple unloading layers with 1000 mm in height. The available detection time is ~11 second. The metal surface of the product partially reflects light. The unloading pattern for each product is different and known in advance, and the product position deviates from the ideal one ~± 50 mm with random rotation. Interfering ambient light is also present. The maximum product dimension is ~250 mm. The system has to successfully detect a partially filled unloading layer and appropriately optimize the time of product unloading.
Solution: Prior to unloading, the robot acquires the image of the product on the controller trigger signal, using a camera and customized lighting. The image is processed and the information about the coordinates as well as the rotation of the product is sent through a standardized industrial communication protocol in ~100 ms. Customized illumination guarantees homogenous light conditions for image acquisition. Together with the camera, the light sources constitute a light imaging module, the weight of which is optimized for robot applications. In addition to the controlled conditions, the developed vision algorithm is robust for a wide range of image quality. This guarantees error-free functioning in the most difficult industrial conditions. A special calibration algorithm eliminates the non-linearities occurring due to the imaging system perspective error. The system successfully recognizes and ignores the empty parts of the unloading layer.