Image Feature Selection Method Based on Immune Encoding Mechanism

For two-classes image pattern recognition problem about object and background, based on the existing feature selection methods, a novel image feature selection method named immune antibody construction algorithm (IACA) is proposed using immune antibody encoding mechanism. In the case of sample parameter estimation, this method dose not only take into account individual feature´s sensitivity to object and background measured by entropy, but also define the inclusion and complementary formulas about multi-features in set theory perspective. Guided by the minimum energy principle, image immune antibody construction rules and corresponding algorithm are given to find an optimized feature subset as object immune antibody. Furthermore, the dimension of the subset can be automatically determined without prior setting. The computing result has proved to be the optimal feature subset. Finally, data testing result shows that this proposed algorithm has a lower computational complexity and error recognition rate than other methods, which has verified the superiority and the advanced nature of the method.