• Laser Journal
  • Vol. 46, Issue 1, 84 (2025)
REN Anhu, LI Yufei, and CHEN Yang
Author Affiliations
  • School of Electronic Information Engineering, Xi'an University of Technology, Xi'an 710021, China
  • show less
    DOI: 10.14016/j.cnki.jgzz.2025.01.084 Cite this Article
    REN Anhu, LI Yufei, CHEN Yang. Improved detection of unusual highway traffic events for YOLOv8[J]. Laser Journal, 2025, 46(1): 84 Copy Citation Text show less

    Abstract

    An improved YOLOv8n target detection algorithm is proposed to address the low detection accuracy, the excessive amount of detection model parameters affecting the detection speed, the omission and misdetection of small target detection, and the satisfaction of real-time detection of highways that exists in the existing UAV aerial detection images. The CBAM attention mechanism is added to the neck network structure to increase the detail information of small targets and improve the feature extraction accuracy; the original backbone network is replaced with MobileNetV3 network structure, and the overall network is improved by lightweighting, so as to improve the detection efficiency and detection accuracy; and the Focal-EIoU Loss is used to replace the original CIoU loss function, and the performance optimisation of the loss function is carried out to improve the detection model. function to optimise the performance and improve the generalisation ability of the detection model. The experimental validation of the model on the dataset shows that the improved MCF_v8n detection model outperforms the original model, with a reduction of about 23% in the number of parameters, a decrease of about 30% in the computation volume, and an improvement of 5.0 and 2.4 percentage points in mAP@0.5 and mAP@0.50 ∶ 0.95, respectively, which demonstrates good detection performance overall.