基于改进YOLOv8算法的路面病害数字化研究

Digital Study of Road Damage on Improved YOLOv8 Algorithm

为解决基于机器视觉检测路面病害的速度与精度不高、算法优化效果不显著的问题,在YOLOv8算法中引入CBAM模块,并开展消融试验分别引入SE模块和ECA模块,使用开源的RDD 2020数据集训练模型,经图像预处理得到二进制图像。与YOLOv8算法相比,CBAM-YOLOv8算法的精确率、召回率、mAP@0.5、mAP@0.5:0.95均有所提高,而SE-YOLOv8算法和ECA-YOLOv8算法的检测精度较YOLOv8有所降低。进一步通过RDD 2022数据集验证了CBAM-YOLOv8算法在路面病害检测中的优越性。CBAM-YOLOv8算法的应用可以进一步提高路面病害检测的速度和精度,具有显著的优化效果。

In order to solve the problem that the speed and accuracy of road damage detection based on machine vision are not high,and the optimization effect of algorithm is not significant,CBAM module is introduced into YOLOv8 algorithm,and ablation experiments are carried out to introduce SE and ECA modules separately.The IEEE big data of RDD 2020 is used to train the model,and binary images are obtained through image preprocessing.Compared with YOLOv8 algorithm,the accuracy rate,recall rate,mAP@0.5,mAP@0.5:0.95 of CBAM-YOLOv8 algorithm have been improved,while the detection accuracy of SE-YOLOv8 algorithm and ECA-YOLOv8 algorithm is lower than that of YOLOv8 algorithm.Furthermore,the superiority of CBAM-YOLOv8 algorithm in road detection is verified through RDD 2022 dataset.The application of CBAM-YOLOv8 algorithm can further improve the speed and accuracy of road damage detection.