Vehicle detection and classification for traffic management and autonomous systems using YOLOv10

With the continuous development of Intelligent Transportation Systems (ITS), real-time vehicle detection and classification have become critical tasks for urban traffic management and autonomous driving. However, existing detection methods face challenges such as small target detection, severe occlusion, and changing traffic conditions. The main purpose of this study is to address these challenges and improve vehicle detection in ITS environments. We propose a novel detection framework, YDFNet, that integrates the YOLOv10 algorithm for fast feature extraction, the BiFPN (Bidirectional Feature Pyramid Network) for multi-scale feature fusion, and the DETR (Detection Transformer) architecture for global feature modeling. Our approach leverages the advantages of each method to enhance detection accuracy and efficiency, especially in complex traffic scenarios. Experimental results on the UA-DETRAC and COCO datasets demonstrate that YDFNet outperforms existing methods in terms of detection accuracy, small target detection, and inference speed. The novelty of our work lies in the effective combination of YOLOv10, BiFPN, and DETR, which improves the model’s robustness to dynamic environments while maintaining real-time performance. This research provides a new solution for efficient and accurate vehicle detection in ITS and lays the foundation for future work on multimodal data fusion and model optimization.