The IFOY nominees' list (2): Pallet assessment has traditionally been an additional task at goods receiving, which was resource-intensive and prone to errors. Sick's PACS (Pallet Classification System), based on deep learning, automates this process.
The system was featured in the "Specials of the Year" category at the intralogistics competition. The modular system is based on Sick's "Appspace," an innovative approach to developing high-performance apps with sensors from Waldkirch, and "dStudio." The latter is a web-based tool for classifying images using artificial neural networks, which can be used with the manufacturer's sensors. The modular system can also be used for other image processing tasks.
Automated recognition simplifies the process of automatically assigning deposits to different pallet types and can be integrated as a compact system with a small footprint in many locations.
The system's hardware consists of one or more color cameras for image capture, a light barrier arrangement for triggering, and a controller for processing the data and executing the trained neural network.
The software tools allow image recording, training, classification, and execution of the trained network, even without in-depth knowledge of programming or machine learning. Optionally, additional sensors can be integrated to perform further tasks.
Image recognition was not considered groundbreaking at the IFOY test camp. Systems already exist that also register the condition of pallets, reject damaged load carriers, and send them for repair. The unique aspect of Sick's solution lies in its integration into a "neural network" where identification, classification, and detailed processes are continuously trained to refine the recognition capabilities.
Unlike conventional image processing solutions, the use of deep learning technology here requires no detailed programming knowledge, as the system learns from concrete examples. This makes pallet identification comparatively simple for the customer. Where the use of trained neural networks usually requires in-depth knowledge of machine learning, Sick has developed "dStudio," a training software that includes a guided process flow.
Because this is a process used thousands of times, significant cost savings, resource conservation, and process quality can be achieved. The technologies employed are not only suitable for pallet identification but can also offer significant advantages in other areas. Thanks to its space-saving design, the system can be integrated even in tight assembly spaces. The use of standard sensors allows for comparatively low maintenance and upkeep of the system.
This year, 14 devices and solutions from 12 manufacturers made it to the final round of the tenth competition. The nominees underwent a three-stage audit during the IFOY test camp in March and were tested by the jury. Voting is anonymous. The results will remain secret until the Award Night on June 30th, which will take place at BMW Welt in Munich.
www.sick.com
