Phase Only Filter Pattern Classifier: Multi Dimensional Tracking, Classification and Training
The only independent variable in electronic computingis time. Optical computing on the other hand, hasinherently two degrees of freedom, the two variablesthat define a point in a plane. Optical systemsalways process information in parallel. Such a simpleoptical element as a lens is capable of performingsuch a difficult task as a Fourier transform.Therefore, when it comes to pattern classification,optical computing is an attractive option. The PhaseOnly Filter has been showed to be a powerful tool fortracking objects in a two-dimensional plane. In thisresearch, a special tracking technique is developedto overcome weaknesses of the POF under noisycircumstances. The POF is generally implemented in the 2-D plane.However, the POF has neither been trained as apattern classifier for one-dimensional data, nor as amulti-dimensional data classifier. Methods aredeveloped in this work to apply the POF tomulti-dimensional pattern classification. Moreover,POF equivalent neural network techniques are devisedand implemented for pattern classification. Two levelneural network is developed for the case ofmulti-class classification, and a method of trainingis developed.