Acquisition
A method of both spectral spot and area scanning in the VIS/NIR (400nm-1000nm) range was developed. The application of special spectral head allows parallel acquisition of the spectral data, and dramatically reduces acquisition time.
Image processing
The method offers the filtration of 2D images at a selected wavelength as well as presentation of a whole absorption and reflectance spectrum . The standard graphic operations on layers and masks could be used to select individual characters and parts of the studied document.
Analysis and Visualization
Several dedicated algorithms for spectral analysis of questioned documents were developed, including:
- automatic segmentation by maximal spectral difference,
- automatic and semi-automatic classification based on correlation method and error minimization
- Principal Component Analysis (PCA) to reduce the feature space,
- pattern recognition and machine learning, e.g. one class clasiffier.
The implemented visualization enables the presentation of the spectral data for the single point or the selected area of the document, as well as 2D and 3D images filtered by static or gradient wavelength selection.
These algorithms allow detection of subtle differences in the VIS/NIR spectra recorded for the traces of writing materials in the document, which creates an excellent tool for the identification of alterations or additions.
