Sand, gravel, coal, deicing salt, diamonds, grain, sugar, coffee or grapes and waste — a lot of everyday goods are more or less grainy. To classify this bulk material by quality and size, it must be sorted in a sophisticated process. Scientists of the Karlsruhe Institute of Technology (KIT) and the Fraunhofer Institute for Optronics, Systems Technology and Image Evaluation (IOSB) have developed a system that is able to sort bulk material much faster, cheaper and more accurately than previous techniques.
Bulk materials are processed in nearly all industries, including building construction, logistics, chemistry and agriculture. Given the number of plants, any improvement of the sorting steps necessary everywhere could save billions.
"Classical belt-type sorting systems scan the material to be sorted by means of a camera; parts not to be treated are blown out by means of pressurized air nozzles," explains Georg Maier of the IOSB.
There is a problem however: The cameras now in use scan the objects only on a short belt section, which allows only rough classification. As a consequence, several runs are necessary to achieve a satisfactory result.
The team developed a sorting system in which cameras operating from different perspectives capture a more precise view of the bulk material. In this way, objects of different categories can be distinguished more effectively. In addition, algorithms based on the images predict how objects are going to move on the belt. In this way, it is possible to sort out foreign bodies much more precisely.
The advantages of the new sorting system are explained by Benjamin Noack of the ISAS by the example of spheres and hemispheres: "When seen from the top, they look alike. While hemispheres normally remain on the belt, spheres are restless, thus additionally aggravating the sorting process. However, when seen from the side, spheres and hemispheres can be distinguished." In addition, the system was now able to predict the behavior of these objects and then adjust its operation accordingly and, in this way, also achieve better classification, says Noack.
The scientists also found a way to increase the accuracy of existing sorting systems quickly and at low cost. "This requires but a software update," said Hanebeck.