Comprehensive Analysis to Detect Optimal Vehicle Position for Roadside Traffic Surveillance Using Lightweight Contour-Based CNN

In the realm of transport development, the fusion of modern technology and vehicle surveillance in roadside areas becomes indispensable. Traditional surveillance demands continuous monitoring through closed-circuit television cameras. It results in a huge amount of data, which requires high computation. This study delves into the challenges of real-time processing of vehicle surveillance within smart cities with quality data. In addition to a specific focus on monitoring the roadside traffic region despite technological advancements, including target variability, lighting conditions, and occlusion, the manuscript introduces a lightweight contour-based convolutional neural network to address these challenges. The proposed work aims to gain the maximum features from the vehicle via detecting the optimal position and incorporating a Region-Proposal-Network, Region-of-Interest-Align and pooling, Non-Maximum-Suppression, Structural-Similarity-Index, and Peak-Signal-to-Noise-Ratio. The proposed work extracts hierarchical information from a custom video dataset and demonstrates superior performance with an accuracy rate of 97.36% and a minimum loss of 0.0816 in an elapsed time of 1s 159ms. Furthermore, it achieves a validation loss of 0.1506, and a validation accuracy of 96.46%. Additionally, manuscripts illustrate different datasets and models through a systematic literature review. Moreover, the manuscript also illustrates the Smart-City framework and Integrated Traffic Management System architecture.