Optimal zoning design by genetic algorithms

In pattern recognition, zoning is one of the most effective methods for extracting distinctive characteristics from patterns. So far, many zoning methods have been proposed, based on standard partitioning criteria of the pattern image. In this paper, a new technique is presented for zoning design. Zoning is considered as the result of an optimization problem and a genetic algorithm is used to find the optimal zoning that minimizes the value of the cost function associated to the classification. For this purpose, a new description of zonings by Voronoi diagrams is used, which is found to be well suited for the genetic technique. The experimental tests, carried out in the field of handwritten numeral and character recognition, show that the proposed technique leads to zonings superior to traditional zoning methods