A neural network based cascaded classifier for face detection in color images with complex background

Abstract

When utilizing neural networks as a classifier in face detection systems there are two important problems which should be solved: 1. High computations between the network layers and 2. Adjusting the topology of the network. The proposed system in this paper uses a genetic algorithm to directly solve the second problem and a fuzzy inference engine as a pre-classifier to indirectly deal with the first problem. After computing a small number of reliable and easy to extract features from skin like regions, in the pre-classification step, a set of flexible rules are applied by a fuzzy inference engine. The accepted regions are fed into a neural network for final decision making. Using this combination of classifiers has established an acceptable tradeoff between the computation and the missed faces while the rate of correct detection is acceptably high. © 2008 Springer-Verlag Berlin Heidelberg.