基于条件随机场的钢箱梁正交异性板疲劳微裂纹检测

Fatigue crack detection for orthotropic plate of steel box girder based on conditional random fields

钢箱梁正交异性板在桥梁建设中应用广泛,对此类结构疲劳损伤进行无损检测并发现病害特征,有利于及时的进行现场维修加固。该文使用高倍数显微相机进行疲劳微裂纹的检测,提出基于条件随机场(CRFs)的金属疲劳微裂纹检测算法,该方法使用单个像素的表观特征来进行裂纹判别,同时也考虑其他像素标注值间的影响,从而很好地抑制离散噪音点。通过多种表观特征和机器学习方式自动识别出区分性最强的特征从而加以选择使用。实验结果表明,基于 CRFs 模型的裂纹检测方法对于试件 6~16 万次的图像裂纹宽度测量与人工测量值非常接近,这为钢箱梁正交异性板疲劳损伤快速检测提供可靠的分析手段。

Orthotropic plates steel box girder are widely used in bridge construction nowadays, it is very vital to efficiently inspect the fatigue damage of the orthotropic plates for reinforce maintenance. The paper utilizes a high resolution microscope camera to detect orthotropic plate fatigue cracks based on conditional random fields (CRFs). The proposed method not only takes the appearance features of the independent pixels into account but also the correlation between neighboring points’ labels, which suppresses the noise very well. The most distinctive features are Automatically identified through multiple visual features and machine learning methods The experiment shows that the proposed method, which is selected and used. The experiment shows that the method of crack detection based on CRFs model is very close to the measurements of the specimen for 60-160 thousand times, which provides a reliable analytical method for the rapid detection of fatigue damage of orthotropic plates.