Vision assisted machine for recycling applications


A recurrent problem in recycling is the separation of valuable materials from a mixture of elements. In order to recover raw materials with the highest purity, physical and/or chemical methods are used. In the case of electrical cables and electronic scrap, state of the art techniques based on differences in the physical properties of each mixture component have been applied with reasonable success. However, these techniques have limitations and require further chemical processing in a refinery. The aim of this project is to develop an automatic and flexible machine to remove small particles (metallic and non-metallic particles as small as 0.5 mm²) from a mixture of very similar materials, resulting in a final product with high purity. This output material, which can be sold at high price on the market, will require little or no additional refining. The idea is to identify the particles by differences in color or other features, using a new and innovative approach combining computer vision and feeding/extraction techniques. The economic benefits from the new process will be twofold. Production costs will be reduced and the quality of the resultant product will be increased. The main innovations are a new machine concept for automatic separation of materials from impurities, new dedicated color vision hardware and software for sorting purposes (including a new illumination concept) and new modules for feeding and extraction systems. The pictures below show the design of the machine and a picture of the prototype.


  • Monitoring and control of building automation systems
  • Efficient real-time capable fuzzy video editing
  • Fuzzy control and monitoring in embedded systems
  • Fuzzy prediction of system behaviour in non-chaotic systems


Illumination is a very important aspect of a machine vision system, as better illumination results in improved camera images. The final design of the lighting module is the result of the analysis of all the possible aspects that can influence the retrieval of the scene. The better the image, the better the processing results will be.

In the VISREC project, a good lighting system provides a good contrast between copper, lead and the background color – the conveyor belt. Some factors must be avoided, such as non-uniform light distribution in the scene and reflections (see pictures below). The presence of some of these factors makes convenient and generalized image processing difficult; if the illumination is not adequate, some impurities are not identified.

Fuzzy control

For many years, research has been done in the field of predictive maintenance and monitoring systems. As an example, a renowned company in Germany working in the area of automatic milling machines recognized the value of predictive and condition-based maintenance. They provide special hardware to predict breakdowns in gears or vibrating parts. This system can be used on parts with a long life cycle for which the breakdown behavior is well analyzed, and thus represents a predictive maintenance system. This is just one possibility to maintain a machine with well-known parts.

The VISREC project represents the development of a new kind of maintenance system based on fuzzy logic and fuzzy control. With fuzzy control it is possible to describe behavior in a linguistic manner without knowledge of a detailed interrelationship or mathematical model. This description enables a user to easily insert his expertise into the controller. The characteristic diagram below is build from fuzzy rules with our new developed maintenance fuzzy controller.

Monitoring and predictive maintenance system

Presently, fuzzy controllers are used to control processes, but it is also possible to use a fuzzy controller as a monitoring and predictive maintenance system. This system is based on the following setup. A fuzzy controller acting as the controller requlates a process (machine or similar process), but the controller is in turn observed by another fuzzy controller acting as monitor. With this monitoring it is possible to detect and correct malfunctions the controller could not detect. This monitor is observed by another Fuzzy controller acting as a maintainer. If the monitor is not able to correct the controller it warns the human process user to maintain the process.

As an example, the images below show an office (left picture) that should be maintained continuously at a minimum brightness. As an additional constraint, the lamps need about 15 minutes to achieve full brightness. Thus, it is too late to turn them on when the minimum brightness threshold has been reached an we must effectively predict the brightness in the office. The brightness will change throughout the day due to movement of the sun or clouds that obscure the sun. If all eight lamps in the office are turned on, sufficient light is produced even if it is completely dark outside. The temperature inside the office is also measured and controlled. Not just the room is monitored; if for example certain lamps are defective the controller detects this malfunction and sends a message to a user to repair those lamps.

The image on the right shows the control panel for the user, with which the user can see every measured value or be informed about any malfunction. This visual feedback increases the efficiency in maintaining the office.




Student Jobs / Open Positions