Accurate vehicle classification and speed estimation are essential for effective traffic monitoring and management in urban environments. This study presents a YOLOv5-based deep learning model integrated with an attenuation layer to enhance detection precision across diverse vehicle categories. The system classifies stereo vision-based footage into seven major groups. A stereo camera setup captures live traffic scenarios, allowing for depth estimation to determine object positions and velocities. The attenuation layer refines feature extraction by reducing background noise, thereby increasing reliability in dense urban conditions. The YOLOv5 model achieved a high detection precision of 99%, validating its effectiveness in multi-class vehicle recognition. The speed estimation method demonstrated high accuracy, with a low margin of error of ±0.05, confirming its suitability for real-time applications. Integration of the attenuation layer significantly improved noise resistance and overall model robustness in complex scenes. Thus, the proposed method enhances both detection accuracy and speed estimation capabilities, supporting advanced intelligent transportation systems for smarter urban traffic management.
Shah et al. (Tue,) studied this question.