Image Processing : Artificial Immune System (AIS) For Digits Classification

Azah Kamilah, Muda and Siti Mariyam, Shamsudin (2004) Image Processing : Artificial Immune System (AIS) For Digits Classification. Project Report. UTeM, Melaka, Malaysia. (Submitted)

PDF (Full Text) Image_Processing_-_Artificial_Immune_System_(AIS)_For_Digits_Classification.pdf - Submitted Version Restricted to Registered users only Download (2MB)

Abstract

In this study, a biological based solution on Digits classification is explored with a computational implementation of negative selection (NS) of immune system (IS) with genetic algorithm. Immune system possesses several properties such as self/nonself discrimination, immunological memory, positive/negative se lection, immunological network, clonal selection, and learning which performs complex tasks. In the research, we use negative selection that has the property of self/nonself distinction to detect the foreign antigens. NS algorithm attempts to construct a set of distinguish detectors in the following way : a) define the self data; b) generate a random candidate detector; and c) match each candidate generated with self data. The uniqueness of this research is on the feature extraction phase, in which, the digits are extracted using moment invariants. These invariants are converted using the mathematical formulation of f : 9t ~ Lm into symbolic representation. In this study, a few samples of handwritten digits are used, in which, each digit is represented with 8 bits binary string. The matching procedure is done by searching an equal length of each string and is done by using the XOR operator. The length of the mismatch in the bits is counted and the matching score is found out using the equation as stated below: matching where 8 score = count + L 2 1 1=1 count is the length of the total number of mismatches in the string.

Actions (login required)

View Item