Classifier combination for vehicle silhouettes recognition

This article investigates how the combination of neural classifiers in committees can improve the performance achieved in a pattern recognition task: vehicle silhouettes recognition. For such different approaches for classifier combination are evaluated. The performance achieved by these approaches are compared to those achieved by the individual neural classifiers. Three neural network models and five different combination methods are investigated. Three neural models are used in the experiments: multilayer perceptron (MLP) network, the radial basis function (RBF) network, (3) and the cascade correlation network. These models are used due to the different approaches employed by them to build decision boundaries. The article discusses the individual neural classifiers used in this work.