基于MSR多尺度细节增强与双阈值约束的隧道裂缝提取方法  

Tunnel Crack Extraction Method Based on MSR Multi-Scale Detail Enhancement and Dual-Threshold Constraint

作　　者：黄剑雄 刘宜胜[1] 

机构地区：[1]浙江理工大学机械工程学院,浙江 杭州

出　　处：《建模与仿真》2023年第4期3587-3594,共8页Modeling and Simulation

摘　　要：由于传统图像处理算法难以处理在地铁隧道复杂照明环境下的裂缝缺陷图像,提出一种基于多尺度细节增强与双阈值约束的隧道裂缝提取方法。首先,对图像进行预处理,依据Retinex理论得到图像的反射分量的同时获得图像高频成分并进行融合,以突出图像中的裂缝信息。然后通过一种在全局大津阈值约束下的改进局部阈值法将图像分割成背景成分与伪裂缝成分。最后,通过裂缝结构特征分析设置形状滤波完成裂缝信息的提取。实验结果表明,该方法相比于其他方法在裂纹检测中表现出更好的性能。Due to the difficulty of traditional image processing algorithms in handling crack defect images in complex lighting environments in subway tunnels, a tunnel crack extraction method based on mul-ti-scale detail enhancement and double-threshold constraints is proposed. Firstly, the image is pre-processed, and the reflection component of the image is obtained based on the Retinex theory, while the high-frequency components of the image are fused to highlight the crack information in the image. Then, the image is segmented into background components and pseudo crack compo-nents using an improved local threshold method under a global Otsu threshold constraint. Finally, the crack information is extracted by setting a shape filter through crack structure feature analysis. Experimental results show that compared with other methods, this method performs better in crack detection.

关 键 词：RETINEX理论 细节增强 图像分割 图像处理算法 裂纹检测 照明环境 隧道裂缝 高频成分 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]