基于改进YOLOv8的道路坑洼检测算法

为解决现有目标检测算法在道路坑洼检测方面精准性较低的问题,该研究提出一种基于改进YOLOv8的道路坑洼检测算法。首先,在YOLOv8主干网络中引入三分支注意力(Triplet Attention,TA)模块,强调在计算注意力权重时捕捉跨维度交互的重要性,以提供更丰富的特征表示,并在计算上更加高效,有助于更准确地定位和识别检测对象;该研究针对坑洼道路提出一种新的轻量化检测头——Flex_Detect,使用双分支卷积并且动态调整锚框,确保模型在不同尺度的特征图上能够有效地进行目标检测,有助于提高模型对不同尺寸目标的适应性,提高模型在目标检测任务上的性能和泛化能力。实验结果提出的YOLOv8_Efficient在公开数据集上的平均精度相较于原始YOLOv8n提升2.5%,相较于YOLOv5n提升4.1%。

In order to solve the problem of low accuracy of existing object detection algorithms in road pothole detection,a road pothole detection algorithm based on improved YOLOv8 is proposed.First of all,the Triplet Attention(TA) module is introduced into the YOLOv8 backbone network to emphasize the importance of capturing cross-dimensional interactions when calculating attention weights,so as to provide richer feature representations and be more efficient in calculation,which is helpful to locate and identify detection objects more accurately.In this study,a new lightweight detection head,Flex_Detect,is proposed for potholed roads,which uses double-branch convolution and dynamically adjusts the anchor frame to ensure that the model can effectively detect targets on feature maps of different scales,which is helpful to improve the adaptability of the model to targets of different sizes,and improve the performance and generalization ability of the model in object detection tasks.The experimental results show that the average accuracy of YOLOv8_Efficient on open data sets is 2.5% higher than that of the original YOLOv8n and 4.1% higher than that of YOLOv5n.