The growing challenges of urbanization and increasing traffic volumes pose significant risks to global mobility, environmental sustainability, and urban health. To address these issues, this research proposes an innovative aerial surveillance framework that leverages advanced machine learning techniques and intelligent systems for vehicle monitoring. The process begins with robust preprocessing of aerial imagery using histogram equalization, followed by vehicle segmentation via Fully Convolutional Networks (FCN) and accurate detection through Faster R-CNN, ensuring adaptability to complex and varying traffic scenarios. A hybrid feature extraction approach integrates Gabor Filters and Local Binary Patterns (LBP) with deep features from EfficientNet, providing a comprehensive and nuanced representation of vehicle characteristics. These features are then optimized through a Genetic Algorithm to maximize detection accuracy and efficiency. The final classification, achieved with a Neuro-Fuzzy Classifier, reaches an impressive accuracy of 92.38%, validated on the VAID dataset, a benchmark in aerial traffic analysis. This framework not only improves vehicle detection but also offers real-time, scalable solutions for urban mobility, providing a vital tool for mitigating traffic congestion and enhancing the sustainability of future cities. This research lays the groundwork for next-generation intelligent traffic monitoring systems, capable of addressing the evolving challenges of modern urban transportation.
Smart Traffic Monitoring with EfficientNet and Neuro-Fuzzy Classifier via Aerial Surveillance
2024-12-30
1127774 byte
Conference paper
Electronic Resource
English
Estimating the crowding level with a neuro-fuzzy classifier
British Library Online Contents | 1997
|Neuro-fuzzy techniques for traffic control
Tema Archive | 1998
|Neuro-Fuzzy Techniques for Traffic Control
British Library Conference Proceedings | 1997
|