为实现复杂环境下道路路面裂缝的高效检测,提出基于改进YOLOv8的路面裂缝检测方法.为减少模型参数量,采用Ghost Conv代替传统卷积,构建新的C2f模块,并改进损失函数,在回归损失中加入归一化Wasserstein距离损失来弥补交并比(Intersection over Union,IoU)损失的缺点.为克服样本不平衡带来的问题,将分类损失修改为变焦损失函数.测试结果表明,该算法具有良好的检测效果,可以准确检测图片中的所有裂缝.
Road Crack Detection Method Based on Improved YOLOv8
To achieve efficient detection of road surface cracks in complex environments,a road surface crack detection method based on improved YOLOv8 is proposed.To reduce the number of model parameters,Ghost Conv is used instead of traditional convolution to construct a new C2f module,and the loss function is improved.Normalized Wasserstein distance loss is added to the regression loss to compensate for the shortcomings of Intersection over Union(IoU)loss.To overcome the problem of imbalanced samples,the classification loss is modified to a zoom loss function.The test results show that the algorithm has good detection performance and can accurately detect all cracks in the image.
万方数据
维普
