Optimizing the performance of probabilistic neural networks using PSO in the task of traffic sign recognition

Abstract

This paper presents a fast version of probabilistic neural network model for the recognition of traffic signs. The model incorporates the J-means algorithm to select the pattern layer centers and Particle Swarm Optimization (PSO) to optimize the spread parameter, enhancing its performance. In order to cope with the degradations, the Combined Blur-Affine Invariants (CBAIs) are adopted to extract the features of traffic sign symbols without any restorations which usually need a great amount of computations. The experimental results indicate that the fast version of PNN optimized using PSO is not only parsimonious but also has better generalization performance. © 2008 Springer-Verlag Berlin Heidelberg.