| 基于改进YOLOv8的夜间车辆检测研究 |
 点此下载全文 |
| 引用本文：郭俊，郭金霆，张世龙.基于改进YOLOv8的夜间车辆检测研究[J].计算技术与自动化,2026,(1):108-113 |
| 摘要点击次数: 274 |
| 全文下载次数: 219 |
|
|
| 中文摘要:针对夜间光照不足、小目标易遮挡等导致检测性能下降的问题，提出了改进模型DFF-YOLOv8。该模型在YOLOv8基础上引入Dysample动态采样模块，自适应调整特征采样率以增强弱光小目标感知；融合ASFF特征融合机制，实现多尺度信息高效整合；并采用Focaler IoU损失函数动态优化回归精度。基于BDD100K夜间数据集实验结果显示，DFF-YOLOv8在mAP@50、mAP@50:95、P、R上分别提升4.8%、3.8%、1.9%、4.1%，显著改善夜间低光环境下的检测性能。 |
| 中文关键词:夜间光照不足 动态采样 特征融合 损失函数 YOLOv8 |
| |
| Research on Nighttime Vehicle Detection Based on Improved YOLOv8 |
|
|
| Abstract:To address the decline in detection performance caused by insufficient illumination and occlusion of small targets at night, this paper proposes an improved model DFF-YOLOv8. Based on YOLOv8, the model introduces the Dysample dynamic sampling module, which adaptively adjusts the feature sampling rate to enhance the perception of small targets under low light. The ASFF feature fusion mechanism is integrated to achieve efficient multi-scale information aggregation, while the Focaler IoU loss function is adopted to dynamically optimize regression accuracy. Experimental results on the BDD100K nighttime dataset show that DFF-YOLOv8 improves mAP@50, mAP@50:95, precision, and recall by 4.8%, 3.8%, 1.9%, and 4.1%, respectively, significantly enhancing detection performance in low-light nighttime environments. |
| keywords:insufficient nighttime illumination dynamic sampling feature fusion loss function YOLOv8 |
| 查看全文 查看/发表评论 下载pdf阅读器 |
|
|
|
