An Improved YOLOv8-Based Vehicle Detection Algorithm for Complex Traffic Scenes

Yunjuan Cheng; Botao Liu

doi:10.54691/785f7j20

Authors

Yunjuan Cheng
Botao Liu

DOI:

https://doi.org/10.54691/785f7j20

Keywords:

YOLOv8; Vehicle Detection; Attention Mechanism; Lightweight; Adaptive Loss.

Abstract

To address illumination changes, occlusion, small objects, and the limited compute of edge devices in intelligent transportation scenarios, this paper proposes an improved detection framework built upon YOLOv8. A Centralized Feature Module (CFM) is introduced into the backbone to fuse global semantics with corner-level details and enhance small-object representation. A Dynamic Detection Head (DDH) is further designed that cascades scale, spatial, and task attentions to adaptively allocate features. In addition, an Adaptive Modulation Loss (ADLF) dynamically adjusts the IoU threshold and loss weights, improving robustness to occlusion and ambiguous boundaries. Experiments on UA-DETRAC and SODA10M demonstrate that, while maintaining real-time performance (≥220 FPS), the proposed method improves mAP50 from 90.9% to 93.1% on UA-DETRAC and from 69.7% to 72.3% on SODA10M, markedly reducing missed detections for small objects and in cluttered backgrounds. The model contains ~10.6M parameters and ~27.8G FLOPs-lower than the baseline-showcasing strong potential for lightweight edge deployment on in-vehicle and roadside devices.

Downloads

Download data is not yet available.

References

[1] GIRSHICK R. Fast R-CNN[C]// Proceedings of the IEEE International Conference on Computer Vision. Piscataway, NJ: IEEE, 2015: 1440–1448.

[2] TERVEN J, CÓRDOVA-ESPARZA D M, ROMERO-GONZÁLEZ J A. A comprehensive review of YOLO architectures in computer vision: From YOLOv1 to YOLO-NAS[J]. Machine Learning and Knowledge Extraction, 2023, 5(4): 1680–1716.

[3] LIU W, ANGUELOV D, ERHAN D, et al. SSD: Single Shot Multibox Detector[C]// European Conference on Computer Vision. Cham: Springer, 2016: 21–37.

[4] GIRSHICK R. Fast R-CNN[C]// Proceedings of the IEEE International Conference on Computer Vision. Piscataway, NJ: IEEE, 2015: 1440–1448.

[5] VIOLA P, JONES M J. Robust real-time face detection[J]. International Journal of Computer Vision, 2004, 57(2): 137–154.

[6] LOWE D G. Distinctive image features from scale-invariant keypoints[J]. International Journal of Computer Vision, 2004, 60(2): 91–110.

[7] FELZENSZWALB P F, GIRSHICK R B, MCALLESTER D, et al. Object detection with discriminatively trained part-based models[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2009, 32(9): 1627–1645.

[8] HASELHOFF A, KUMMERT A. A vehicle detection system based on Haar and triangle features[C]// IEEE Intelligent Vehicles Symposium. Piscataway, NJ: IEEE, 2009: 261–266.

[9] GAO T, LIU Z, GAO W, et al. Moving vehicle tracking based on SIFT active particle choosing[C]// International Conference on Neural Information Processing. Berlin, Heidelberg: Springer, 2008: 695–702.