Real-Time Traffic Intensity Estimation and Management using Deep Learning Techniques
Priya Dasarwar,Vaishali Patil,Suryakant Murhekar,Yash Kamble,Tabassum H Khan,Snehlata Dongre
DOI: https://doi.org/10.1109/NMITCON62075.2024.10698813
2024-08-09
Abstract:In India, traffic flow estimation is crucial for effective transportation management and urban planning. This research proposes a comprehensive approach leveraging deep learning techniques, particularly the YOLOv8 object detection model, to estimate road traffic flow in urban environments. The model is trained on the widely used COCO dataset, ensuring robust performance in detecting various objects, including vehicles, pedestrians, and cyclists. Subsequently, the model is fine-tuned on a custom dataset comprising annotated traffic images, enabling it to adapt specifically to the complexities of urban traffic scenarios. The research methodology encompasses several key steps. Firstly, the pretrained YOLOv8 model is utilized to perform inference on sample images, demonstrating its capability to accurately detect vehicles within the urban environment. Visualization and analysis of the detection results showcase the model’s effectiveness in identifying vehicles with high confidence levels, laying a foundation for further experimentation. Following the initial assessment, the model undergoes fine-tuning on the custom dataset, optimizing its performance for the task of traffic flow estimation. Training parameters such as epochs, image size, and learning rate are carefully tuned to achieve optimal results. Additionally, metrics such as box loss, classification loss, and distribution focal loss are monitored during training to assess model convergence and performance. Upon completing the training phase, the model’s performance is evaluated using a validation set, providing insights into its generalization ability and effectiveness in real-world scenarios. The evaluation metrics, including precision, recall, and F1 score, offer quantitative measures of the model’s performance, further validating its utility for traffic flow estimation tasks. Furthermore, the research extends beyond static image analysis to real-time traffic intensity estimation using video data. By integrating the trained model into a video processing pipeline, the system dynamically analyzes traffic density in urban environments, enabling timely interventions and traffic management strategies. The proposed system incorporates region-based traffic analysis, allowing for lane-specific traffic intensity estimation and identification of congested areas.
Engineering,Environmental Science,Computer Science