摘要
围岩节理作为最常见的病害之一,时刻影响着隧道的结构稳定性及安全性.传统方法主要依赖人工识别,该方法不仅效率低下,而且由于人为主观判断的介入,识别结果也会受到影响.针对围岩节理检测精度低、效率低下等问题,基于 YOLOv8 算法提出了一种新的 SP-YOLOv8 围岩节理检测分割算法,对 1 430 张光明隧道和明月山隧道围岩节理图像进行训练识别,这些图像包含了多种复杂地质特征和不同的环境条件.结果显示,该识别算法科学可行,Box(P),BOX(R),Mask(P),Mask(R)性能评估指标高达 95.07%,92.69%,93.81%,91.39%.该方法能够实现对围岩节理快速、精准识别,为围岩节理识别领域的研究和实践提供了有力的技术支持.
Surrounding rock joints are one of the most common defects that constantly affect the structural stability and safety of tunnels.Traditional methods primarily rely on manual identification,which is not only inefficient but also prone to errors due to the introduction of human subjectivity.To address the problems of low accuracy and inefficiency in surrounding rock joint detection,a novel SP-YOLOv8 surrounding rock joint detection and segmentation algorithm based on the YOLOv8 algorithm is proposed.It has been trained on 1 430 images of surrounding rock joints from the Guangming Tunnel and the Mingyueshan Tunnel,which encompass a variety of complex geological features and different environmental conditions.The results of the training and recognition process demonstrate that this algorithm is scientifically sound and feasible,with performance metrics for Box(P),BOX(R),Mask(P),and Mask(R)reaching 95.07%,92.69%,93.81%,and 91.39%,respectively.This method enables rapid and accurate identification of surrounding rock joints,providing robust technical support for research and practical applications in the field of surrounding rock joint identification.