Abstract: In order to realize the rapid and automatic detection of tunnel slab staggering and ensure the safety of tunnel structure, a fast-detecting method for tunnel slab staggering based on oblique line structure light was proposed. Firstly, based on the principle of oblique laser triangulation and imaging, the theoretical formula for calculating the amount of misalignment was derived. And the random sampling consistency (RANSAC) algorithm was used for linear function fitting approximation to convert the angle parameters that were difficult to measure into function constants. Then, the region of interest (ROI) of the image was extracted by combining the target detection algorithm Yolo-v3 and the image gradient features. The sub-pixel level of the center of the light strip was extracted by gray square weighted center of gravity method. And the pixel displacement was obtained according to the distance formula between straight lines. Finally, the laboratory test data and actual engineering data were verified. The research results indicate that the algorithm Yolo-v3 can locate the tunnel slab staggering region, with an accuracy rate of 97.35% and a computational efficiency of 26 images per second. Meanwhile, compared with the extreme value method, the Steger method, and the gray center of gravity method, the proposed algorithm is more accurate in extracting the center of the light bar. Moreover, the absolute error of 87% of slab staggering detection results and actual measurement results is within 0.6mm, and the maximum absolute error is no more than 1.5mm. The proposed method can meet the engineering application.

Key words: tunnel engineering; slab staggering detection; laser triangulation method; structured light; gray square weighted center of gravity method

摘要： 为实现隧道错台快速、自动检测，保障隧道结构安全，提出基于斜射式线结构光的错台快速检测方法。首先，基于斜射式激光三角法原理和成像原理推导出错台量计算的理论公式，并且采用随机抽样一致性(RANSAC)算法进行线性函数拟合逼近，将不易测量的角度参数转换为函数常量；然后，结合目标检测算法Yolo-v3和图像梯度特征提取图像感兴趣区域(ROI)，采用灰度平方加权重心法对光条中心点进行亚像素级提取，依据直线间距离公式获得像素位移量；最后，对室内试验数据和实际工程数据进行验证。研究结果表明：利用Yolo-v3算法实现错台区域定位，准确率可达97.35%，计算效率每秒26张图像，同时与极值法、Steger法、灰度重心法相比，提出的算法在光条中心点提取方面更加精准，并且87%的错台检测结果与实际测量结果绝对误差在0.6 mm以内，最大绝对误差不大于1.5 mm，满足工程应用。

关键词: 隧道工程；错台检测；激光三角法；结构光；灰度平方加权重心法