Estimating the crowding level with a neuro-fuzzy classifier

摘要: This paper introduces a neuro-fuzzy system for the estimation of crowding level in scene. Monitoring number people present given indoor environment is requirement variety surveillance applications. In work, has to be estimated from image processing visual scenes collected via TV camera. A suitable preprocessing images, along with an ad hoc feature extraction process, discussed. Estimation space described terms fuzzy decision rule, which relies on membership input patterns set partially overlapping classes, comprehensive doubt classifications and outliers. society neural networks, either multilayer perceptrons or hyper radial basis functions, trained model individual class-membership functions. Integration nets within rule results overall classifier. Important topics concerning generalization ability, robustness, adaptivity performance evaluation are explored. Experiments with real-world data were accomplished, comparing approach statistical pattern recognition techniques, namely linear discriminant analysis nearest neighbor. Experimental validate large extent. The currently working successfully as part monitoring Dinegro underground station Genoa, Italy.