| 基于YOLOv8n_DSC的复杂道路场景目标检测算法 |
 点此下载全文 |
| 引用本文：张强1 ，王可庆2，薛国强1.基于YOLOv8n_DSC的复杂道路场景目标检测算法[J].计算技术与自动化,2026,(2):1-9 |
| 摘要点击次数: 117 |
| 全文下载次数: 100 |
|
|
| 中文摘要:针对现有目标检测算法存在精度差、计算量大的问题，尤其是在复杂道路场景下对被遮挡目标的检测精度较低的情况，提出了一种基于YOLOv8的道路目标检测改进算法YOLOv8n_DSC。首先，为了在目标被遮挡或重叠的情况下更准确捕捉目标形状和特征，在YOLOv8的骨干网络中设计了基于动态蛇形卷积的C2f_DSC模块来代替C2f模块；其次，为了给模型提供更精确的目标定位，抑制不相关的目标特征信息，使局部特征和全局特征更好地融合，在颈部网络中嵌入双层路由注意力模块；最后，为了解决模型计算量较大的问题，使用共享卷积层的思想改进了检测头结构，加快模型的运算推理速度。实验表明，本文提出的YOLOv8n_DSC算法模型在数据集BDD100K上的mAP@50比原模型提升了3.4%，mAP@50:95提升了2.2%，计算量减少了约25%。 |
| 中文关键词:YOLOv8 目标检测 动态蛇形卷积 检测头 |
| |
| Target Detection Algorithm of Complex Road Scene Based on YOLOv8n_DSC |
|
|
| Abstract:Aiming at the problems of poor accuracy and large amount of computation in existing target detection algorithms, especially the low detection accuracy of occluding targets in complex road scenes, this paper proposes an improved road target detection algorithm YOLOv8n_DSC based on YOLOv8. Firstly, in order to capture the shape and features of the target more accurately when the target is blocked or overlapped, the C2f_DSC module based on dynamic snake convolution is designed to replace the C2f module in the backbone network of YOLOv8. Secondly, in order to provide the model with more accurate target location, suppress the irrelevant target feature information, and make the local feature and global feature fusion better, the bi-level routing attention module is embedded in the neck network. Finally, in order to solve the problem that the model requires a large amount of computation, the detection head structure is improved by using the idea of shared convolution layer to speed up the calculation and reasoning speed of the model. Experiments show that on the data set BDD100K, the mAP@50 and mAP@50:95 of the YOLOv8n_DSC algorithm model proposed in this paper increase by 3.4% and 2.2%, respectively, compared with the original model, and the computational load is reduced by about 25%. |
| keywords:YOLOv8 target detection dynamic snake convolution detecting head |
| 查看全文 查看/发表评论 下载pdf阅读器 |
|
|
|
