Wavelet-neural network models for automatic freeway incident detection

Freeway incidents are non-recurrent and pseudo-random events that disrupt the normal flow of traffic and create a bottleneck in the freeway network. Reliable and fast automatic freeway incident detection is essential for emergency relief and traffic control and management. Earlier solutions have not produced practically useful results primarily because the complexity of the problem does not lend itself to accurate mathematical and knowledge-based representations. In this research, we present new multi-paradigm intelligent systems solutions for the freeway incident detection problem employing advanced signal processing, pattern recognition, and classification techniques. The methodology integrates effectively wavelet, fuzzy, and neural network computing techniques to improve reliability and robustness of the detection.; The fuzzy-wavelet radial-basis function neural network (RBFNN) model uses lane occupancy and speed time-series data from the upstream detector station. A wavelet-based de-noising technique is employed to eliminate undesirable fluctuations in the data. Fuzzy c-mean clustering is used to extract significant information from the observed data and to reduce its dimensionality. A RBFNN is developed to classify the de-noised and clustered observed data. The performance of the model is evaluated and compared with the benchmark California algorithm #8 using both real and simulated data. Based on the evaluation criteria of detection rate, false alarm rate, detection time, and algorithm portability, the model outperformed the California algorithm consistently under various roadway geometry and traffic flow scenarios.; The wavelet energy model uses lane occupancy and flow rate time-series data from the downstream detector station. Wavelet analysis is used to de-noise, cluster, and enhance the observed traffic data, which is then classified by a RBFNN. An energy representation of the traffic pattern in the wavelet domain is found to best characterize incident and incident-free traffic conditions. False alarm during recurrent congestion and compression waves is eliminated by normalization of a sufficiently long time-series pattern. The model is tested under various urban and rural freeway scenarios, producing excellent detection and false alarms characteristics. Moreover, it detected most incidents within 2 minutes of their occurrence. An important characteristic of these models is that they are portable and do not require expensive re-calibrations for optimal network wide performance.