Qualicision® - Image
Verbal description of different tasks by means of qualitative fuzzy rules using linguistic concepts like: big, medium, small, etc.
Rule example: "IF high peaks and small plateaus, THEN high correctness."
In order to implement automated visual evaluations of manufactured items at any stage of a production process, adequate image-analysis algorithms have to be used. For quality control purposes, optically registered information usually undergoes qualitative evaluation.
A human controller would describe the criteria, by which he qualifies a test item, with qualitative or diffuse verbal expressions as the following: "If there is a high number of failures in the area subject to examination, then there is low product quality." In order to evaluate visual information, a decision inference is therefore needed, which allows a qualitative image description of any quality control task as well as a clear knowledge display.
Qualicision® - Image is a rule-based fuzzy decision module, which, in combination with the CVB concept, performs intuitive image analysis. By means of rules formulated with linguistic terms as "big, medium, small" describing specific quality aspects of any criterion, even complex quality control tasks can be mapped in a computer, using a relatively small number of rules. Another advantage is the possibility to model fluid transitions between specific forms of any criterion so that an even evaluation of criterion specificity changes can be assured.
Qualicision® - Image consists of two software components:
With the application program Qualicision® - Image, the user is able to create a fuzzy project in which the variables, terms and rules necessary for decision-making can be edited with graphic support. A program-integrated fuzzy decision component controls the inference behaviour. The user defines specific quality criteria in form of linguistic variables, obtaining as a result of the decision-making process a relevant statement, indicated in form of crisp values.
The application program as well as the dynamic link library of the decision component are implemented under Microsoft-C++ and can be used for real-time applications.
How to work with Qualicision® - Image under the CVB-Concept
Reference Application Datev e.G.
Datev e.G.'s shipping department often has to send different documents to the same addressee. Usually, all shipments have been sent parallel, so the addressee might have received several envelopes or parcels on the same day.
In order to avoid such multiple deliveries to one single addressee - for obvious reasons of cost reduction - via special handling techniques several independent shipments are commissioned in one cardboard tray, which is transported through a conveyor system and finally shipped. As the trays are made of cardboard, gradually they become worn and deformed, which leads to differences in shape and size. Additionally, the trays are bigger than their content, so they can be easily filled. As the whole logistic process is automated, during transport in the conveyor system, tray positions may appear with larger variations, stacked documents can be moved and moved documents damaged. Sheets might get folded or even tap the document's barcode on the address-sheet.
F/L/S Fuzzy Logik Systeme developed a Fuzzy Online Position Control System, which is able to recognize a wide range of different possible tray filling errors. The system immediately evaluates the kind of error and, if necessary, sorts the tray out of the process. Thus, stacks that have to be piled up manually or trays that are too worn will cause no further problems in the commissioning process. This is done by direct communication between the online system and the stored program control (SPC) of the conveyor system.
Though the process has to meet very high real time requirements (high speed recognition), the recognition quality granted by Fuzzy software is equal to human decision quality.