ESTIMATING THE CROWDING LEVEL WITH A NEURO-FUZZY CLASSIFIER

This paper introduces a neuro-fuzzy system for the estimation of the c rowding level in a scene. Monitoring the number of people present in a given indoor environment is a requirement in a variety of surveillanc e applications. In the present work, crowding has to be estimated from the image processing of visual scenes collected via a TV camera. A su itable preprocessing of the images, along with an ad hoc feature extra ction process, is discussed. Estimation of the crowding level in the f eature space is described in terms of a fuzzy decision rule, which rel ies on the membership of input patterns to a set of partially overlapp ing crowding classes, comprehensive of doubt classifications and outli ers. A society of neural networks, either multilayer perceptrons or hy per radial basis functions, is trained to model individual class-membe rship functions. integration of the neural nets within the fuzzy decis ion rule results in an overall neuro-fuzzy classifier. Important topic s concerning the generalization ability, the robustness, the adaptivit y and the performance evaluation of the system are explored. Experimen ts with real-world data were accomplished, comparing the present appro ach with statistical pattern recognition techniques, namely linear dis criminant analysis and nearest neighbor. Experimental results validate the neuro-fuzzy approach to a large extent. The system is currently w orking successfully as a part of a monitoring system in the Dinegro un derground station in Genoa, Italy. (C) 1997 SPIE and IS&T.