Scholarlink

<p>To tackle the challenge of road monitoring for pedestrian safety and vehicle detection, this paper presents an enhanced pedestrian-vehicle object detection algorithm based on YOLOv8. Initially, the loss function within the model is substituted with the WIoU loss function. Second, the SA mechanism is introduced into the Neck layer, and the improved module is integrated into both the Neck and Backbone layers to enhance feature extraction efficiency and performance. Finally, the ASFF module is incorporated into the Head layer to resolve issues such as scale misalignment and feature information loss during multi-scale feature fusion. The results of the improved method on the dataset show that the Recall, Accuracy, and mAP have increased by 5.2, 1.1, and 2.8 percentage points respectively. Moreover, the mAP for detecting pedestrians and cyclists has been improved by 3.2 and 5.3 percentage points respectively. This demonstrates that the improved method is suitable for pedestrian and vehicle detection, ensuring pedestrian safety.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on pedestrian and vehicle detection method based on improved YOLOv8 model

  • 期刊论文

  • 发表日期：

    2025年9月22日

查看全文
  • Xuefei Ma,
  • Min Zhou,
  • Haifeng Zhu,
  • Rahim Khan,
  • Zilong Yang,
  • Yingjian Wu,
  • Modi Qi

摘要

To tackle the challenge of road monitoring for pedestrian safety and vehicle detection, this paper presents an enhanced pedestrian-vehicle object detection algorithm based on YOLOv8. Initially, the loss function within the model is substituted with the WIoU loss function. Second, the SA mechanism is introduced into the Neck layer, and the improved module is integrated into both the Neck and Backbone layers to enhance feature extraction efficiency and performance. Finally, the ASFF module is incorporated into the Head layer to resolve issues such as scale misalignment and feature information loss during multi-scale feature fusion. The results of the improved method on the dataset show that the Recall, Accuracy, and mAP have increased by 5.2, 1.1, and 2.8 percentage points respectively. Moreover, the mAP for detecting pedestrians and cyclists has been improved by 3.2 and 5.3 percentage points respectively. This demonstrates that the improved method is suitable for pedestrian and vehicle detection, ensuring pedestrian safety.