For applications where just a decision about a material class per pixel is needed, the Case Discrimination method is most suitable.
Function
The discrimination of molecular information is obtained as different cases (classes). The resulting image holds, per pixel, a classification ID, which belongs to (trained) cases. Molecular information, which is not similar enough to match a trained case, belongs to an additional case - the unknown case. By this principle, the detection of unknown objects gets feasible. This is needed for the separation of (unknown) impurities from (known) goods.
User Interface
The data set plastics is loaded (the data set is selected in the project browser to the left). Per plastic group a spectra set is defined and are shown in the Selected Spectra graph as well as their originating pixel positions by colored markers in the Input view.
The model method Case Discrimination is selected in the ribbon menu. The control panel of this method is shown on the lower right. The result of the method is shown in the Output view.
Per spectra set a case is defined of the same color as the spectra set. The resulting image is composed of all defined cases and in addition by the unknown case in black color.
Left hand sided, information on the loaded data set are available like the Statistics panel as well as descriptive parameters attached to the data set.
Method Parameters
Method parameters are shown in the panel to the lower left of the perspective.
Case panels
Per case a rectangular panel is shown. A case panel consists of:
Case color selector
Colored box to the left of each case. Set the color you would like to see in the output here. Per default the spectra sets color defines the case color. Please note, the color code is the classification ID.
Case name editor
Specify the name you would like. Per default the spectra set name is used.
Allowed distance slider
Define the tolerated distance a spectra can have to meet a case here.
When a spectrum is not similar enough to any case (i.e. when the allowed distance of all cases is exceeded), the pixel is defined as unknown case and will be visualized in black color.
Reference spectra selector
Set all spectra sets which a case should belong to.
Distance image
Click on the image to enlarge it. The pixel values correspond to the distance of the case to the spectra in the scene. Small values (black) indicate good similarity while large values (white) indicate a large distance to the case.
Delete case button
Click to delete a case.
Add Case (+ icon)
Add a case manually.
Auto
For each defined spectra set, a case is defined automatically.
Work Flow
- Select spectra from the scene to describe all objects of interest
- Click on Auto button
- Study the retrieved output image
- Fine tune the result by setting the "allowed distance" by means of the slider per case
- Try another preprocessing configuration
- Save the model for later usage (e.g. to set up the live streaming)
Mentioned in:
Found 1 search result(s) for "case discrimination method".
... spectra sets. Preprocessing 1st derivative was applied to the data beforehand. Contrary to the CCImethods, the case discrimination method results in a class ID per taught material (i.e. per spectra set ... ...
Jul 07, 2020
© 2019 by Perception Park GmbH
The content of this page and any attached files are confidential and intended solely for the addressee(s). Any publication, transmission or other use of the information by a person or entity other than the intended addressee is prohibited. If you receive this in error please contact Perception Park and delete copied material. Perception Park GmbH, Wartingergasse 42, A-8010 Graz; Austria; FN 400381x