Description:
<jats:p>With rapid urbanization, road network inefficiencies have become a primary impediment for developing sustainable smart cities. Associated challenges range from traffic congestion, ambient pollution, accidents, and lost productivity. In this regard, intelligent transportation system (ITS) based solutions are being proposed to design and develop smart urban mobility. However, real-world traffic flow parameters (such as vehicular count, flow, classification, speed, road capacity, time/distance headway, temporal/spatial densities, heatmaps, and trajectories) are fundamental building blocks for designing and developing ITS solutions. In existing literature, varying solutions have been proposed for traffic flow characterization. These can be categorized as either intrusive or non-intrusive sensors. However, these solutions have serious limitations, especially under heterogeneous traffic flow behavior. To overcome these limitations, compute vision-based solutions have emerged as optimum traffic flow characterization solutions. In this work, all compute vision-based edge computing solutions in existing literature have been reported and analyzed. Through comparative analysis, it is concluded that edge computing solutions are the optimum choice as compared to Internet-of-Video-Things based solutions. However, edge computing solutions are seriously constrained because of the compute resources of single board computers. To overcome this limitation, mathematical traffic flow characterization equations have been reported for calculating eight additional traffic flow parameters. These mathematical equations can optimize the already proposed edge computing solution’s performance by 400%.</jats:p>