High production difficulty of master-matrix, that is used for stamping the SH complex diffractive surface structures, causes the holograms produced by this matrix to be highly unique elements. Thus the operative verification whether the chosen SH were produced by its original master-matrix or not can be very important procedure for counterstand the falsification of SH. One of the methods to provide such a procedure can be based on observation of SH diffractive structure local specificities.
In our previous publications we introduced the Optical-Electronic Scanner (OES) that was developed for automatic inspection of local SH specificities by numeric analysis of coherent diffraction responses of small surface elements [Zlokazov et. al. (2013), Zlokazov et.al. (2014)]. Capturing the diffracted light intensity by matrix photodetector and comparing the captured image responses with the same sample of the same local element of template SH in different points allows to obtain originality of the hologram under investigation. Specific problem of this approach is the requirement of precise positioning of control laser beam head to provide image matching with low error rate. Two ways were realized to decrease the affect of positioning error: 1) realisation of control head positioning in connection to SH global design and 2) application of the knowledge about possible distortions and variability in template image data that will be used in digital processing of surface diffraction responses captured by coherent control system of OES. In first case the aiming channel were developed to recognise SH design images and to provide the shift of optical head in front of the desired element of SH. In second case, using coherent control channel, we gathered data from different training examples of template SH with different shift error of optical head positioning. In both cases application of image processing methods based on invariant pattern recognition algorithms allowed to realize the SH identity procedure in OES with high precision.
2. Image recognition using distortion invariant correlation methods
Correlation pattern recognition (CPR) is widely used technique in digital image processing. In the basis of these methods lies the calculation and analysis of correlation dependence between input data and template object functions. Advantage of CPR methods application is high precision in positioning of target object on image field and possibility of stable target recognition in presence of distortions due to application of distortion-invariant composite correlation filters (CCF) [Kumar (1992)]. CCF is a digitally synthesized two-dimensional data massive which elements can be complex or real values. The CCF calculation algorithms imply the application of training images set which represent the template object under different distortions determined a priori. One of the most investigated and efficiently applied CCF are filters with optimization of output correlation field parameters which are mathematically efficient and showed promising recognition capabilities of objects represented by grayscale images [Zlokazov et.al. (2012)].