An Automated Morphological System for Vehicle-Pedestrian Classification and Inference of Traffic Congestion Metrics

For transportation research and planning, estimating traffic congestion is a crucial issue. Traffic management systems are swiftly integrating methods for traffic congestion detection as essential components. As a first stage in the process of alleviating the problem of traffic congestion, they provide useful data to traffic management systems that aid them in anticipating and avoiding the harmful consequences of traffic congestion. The congestion of traffic can be measured using a variety of metrics. In this paper, a Deep Neural Network (DNN)-based automated system is proposed, where the traffic congestion metrics calculation takes place simultaneously with the data capture. The proposed model is capable of categorizing vehicles and pedestrians along with the computation of the essential congestion metrics. The existing related systems, considering the measurement of the different levels of congestion, have manual data capture, data entry, and data processing, and then manually compute the required traffic indices according to the established indices. Few well-known existing metrics, such as Traffic Volume, Delay Rate, Travel Time Index, and Peak Hour Factor have been computed. With the aid of the inferred data, the determination of the road condition and traffic quality at a particular junction is possible.