Figure 1 represents the input image correlation recognition algorithm using CCF. Training immages captured by data acquisition system during the preliminary adjustment process of OES image recognition unnit are subjected to preprocessing in order to suppress the undesired noises and emphasize the characteristic features of template object in different possible variations of representation. Further the preprocessed training images are senntt to CCF synthesis unit that stores the synthesized CCF in OES memory. Also the process of threshold determinatiion (not shown on figure 1) is required for each synthesized CCF on the preliminary adjustment step. This proocess includes the statistical investigation of CCF response on input images of true and false target objects under ddifferent variations determined a priori. The same image-preprocessing unit must be used for processing of input imagges during the OES performance in operational regime. The acquired and preprocessed input images are sent to corrrelation processing unit “xcorr”. The output distribution of correlation function amplitude is subjected to thresholdding operation that searches peaks and compares its values with threshold value calculated on preliminary adjustmennt step. If the value of correlation peak exceeds threshold value the decision is “true” and input object is classified aas target object. If correlation peak value is less than threshold value the input object is classified as false.
3. Aiming Channel
Most of the real SH’s that used in our researches during OES development have deviatioon of optical head position relative to secured document body content such as printed text, images, and other securrity elements. Thus the positioning of control head in front of the hologram element to be investigated must be made inndependently from position of document “non holographic” elements. For this case an aiming channel was moundded to provide the recognition of SH design images.
Figure 2 represents the optical scheme of aiming channel. Several LED-sources were used ttoo eliminate the SH surface in order to reconstruct the encrypted images and project these images on array of photoo-detectors (PDA). Providing a document scan by aiming channel head allows to achieve the SH design images froom several angular positions light sources. Merging of all the achieved images allowed the detailed visualizatioon the SH design elements. Nevertheless the reference elements of SH design that was chosen to provide thee mapping control appeared roughly imaged in most cases (see figure 2(b)).
Application of correlation pattern recognition methods gave the decision of chosen SHH design reference elements rough images recognition problem. We used computer models of reference elements subbjected to distortion similar to those, which were observed during the capturing and merging of different template exaamples. Developed