Detection and Monitoring of Unauthorized Heavy Vehicles in Residential Areas using Deep Learning

Timely identification and monitoring of unauthorized heavy vehicles in residential areas is motivated by a pressing need to address the escalating challenges associated with urban congestion, safety concerns, and environmental impact. The influx of heavy vehicles into residential zones not only poses risks to the well-being of residents but also contributes significantly to noise pollution and infrastructure deterioration. Further, illegal sand mining operations along riverbanks in urban areas in several developing countries are also known to route their transport vehicles through deserted residential roads during late hours. This article presents a novel approach to the real-time detection of unauthorized heavy vehicles, specifically trucks and buses, in residential areas. A key contribution of this work is the development of a comprehensive dataset specific to heavy vehicles, catering to diverse vehicle types. This dataset is utilized for training and validating four pretrained Convolutional Neural Networks (CNNs): VGG16, Resnet, MobileNet and Inceptionv3, to compare their performance in accurately detecting and classifying heavy vehicles. A comparative analysis of various pretrained CNN architectures that evaluates their effectiveness in heavy vehicle identification is presented, considering various performance metrics for gauging the models' reliability. The results of the study show the effectiveness of the MobileNet deep learning CNN architecture for accurately identifying and monitoring unauthorized heavy vehicles in real-time. The comprehensive heavy vehicle dataset and the comparative analysis of pretrained CNNs contribute valuable insights to the field and can motivate further development of intelligent systems that integrate the proposed model with intelligent systems that provide real-time information about unauthorized vehicular traffic to the appropriate law enforcement authorities.