基于自适应阈值和连通域的隧道裂缝提取

Tunnel Crack Extraction Based on Adaptive Threshold and Connected Domain

为了解决传统人工隧道裂缝检测存在诸如效率低、主观性大、安全性差等弊端,利用裂缝具有方向性和连续性等特征,提出了一种基于自适应阈值和连通域标记的隧道裂缝提取方法.首先,根据裂缝的方向性特征,设计了一种阿拉伯数字算法(Algorithmic)对裂缝进行粗提取,其中对于该算法公式中的阈值选取进行了自适应阈值化,利用改进的阈值迭代法完成系统自动获取最佳阈值,无需人工干预;然后,根据裂缝的连续性特征,采用数学形态中的连通域标记法对裂缝进行细提取,其中通过控制连通域面积大小,实现裂缝粗提取后的去噪处理,通过膨胀和腐蚀操作实现裂缝粗提取后的修复处理;最后,选取了共计165张不同类型的裂缝图像作为实验样本,在MATLAB上进行仿真实验.从实验数据可以看出,自适应阈值和连通域标记的提取方法其提取精度可高达94.2%,平均运行时间仅35.4 s,误识率和拒识率已控制在2.7%和1.1%,相较于传统的图像处理方法有着显著的提高,充分展现出良好的应用前景.

To solve the problems of traditional tunnel crack detection, such as low efficiency, subjectivity and poor safety, among others, a method of tunnel crack extraction based on adaptive threshold and connected domain marking is proposed using orientation and continuity characteristics. First, according to the orientation characteristics of cracks, an Arabia digital algorithm （Algorithmic） is designed to roughly extract cracks. Adap- tive threshold is chosen for threshold selection in the formula, and the improved threshold iteration method is used to automatically obtain the best threshold without manual intervention. Then, according to the continuity characteristics of cracks, the connected region labelling method in mathematical morphology is used to extract the cracks. By controlling the area of the connected domain, the denoising processing after the rough extraction of the crack is realized, and the repair treatment after the rough extraction of the crack is realized by expansion and corrosion operations. Finally, a total of 165 different types of crack images were selected as experimental samples, and simulation experiments were carried out on MATLAB. From the experimental data, we observe that the extraction precision of the adaptive threshold and the connected domain label extraction method can be as high as 94.2%, the average running time is only 35.4 s, the error recognition rate and the rejection rate have beencontrolled at 2.7% and 1.1%, respectively. Compared with traditional image processing methods, our proposed method demonstrates remarkable improvement and shows promise for future applications.