【摘要】 路面裂缝检测是保障道路安全、实现道路损伤及时修复的一项重要任务。针对现有路面检测存在的检测精度低、定位不准等问题，提出一种基于YOLOv8改进的路面裂纹检测算法YOLOv8-pavement。首先在模型训练时对数据集进行离线数据增强，提高模型的泛化能力，其次在骨干网络末端添加Focal Modulation(FM)模块来捕捉图像中长距离依赖和上下文信息以适应裂纹对象的大跨度和细长特征。最后在颈部网络中使用CSPStage(CS)模块，提高特征表达性能，减小模型的参数量和计算量。实验证明，与初始YOLOv8n模型相比mAP50提高了1.2个百分点，而模型的参数量和计算量分别降低3个百分点和4.9个百分点，该算法具有良好的检测效果。更多还原

【Abstract】 Pavement crack detection is an important task to ensure road safety and realize timely repair of road damage.-at the problems of low detection accuracy and inaccurate positioning in existing pavement detection, an improved pavement crack detection algorithm YOLOV8-pavement based on YOLOv8 was proposed. Firstly, off-line data enhancement is carried out on the data set during model training to improve the generalization ability of the model. Secondly, Focal Modulation(FM)module is added at the end of the backbone network to capture the long distance dependence and context information in the image to adapt to the large span and slender features of the cracked object. Finally, CSPStage(CS)module is used in the neck network to improve the performance of feature expression and reduce the number of parameters and computation. Experiments show that compared with the original yolov8n model, mAP50 increases by 1.2 percentage points, while the parameter number and calculation amount of the model decrease by 3 percentage points and 4.9 percentage points respectively. The algorithm has a good detection effect.更多还原