桑文刚, 朱有志, 孙俊峰,等.基于Scaramuzza畸变校正模型的全景相机地铁隧道管片提取技术[J].河南理工大学学报（自然科学版）,2026,45(4):22-29.

SANG W G, ZHU Y Z, SUN J F,et al.Segment extraction method for metro tunnel linings based on the Scaramuzza distortion correction model[J].Journal of Henan Polytechnic University(Natural Science) ,2026,45(4):22-29.

基于Scaramuzza畸变校正模型的全景相机地铁隧道管片提取技术

桑文刚, 朱有志, 孙俊峰, 尹艺博

山东建筑大学 测绘地理信息学院，山东 济南250101

摘要: 目的 为解决传统隧道管片监测技术成本高、数据处理复杂等问题，突破全景相机广角镜头畸变对管片特征提取可靠性的制约，满足地铁隧道管片常态化精细化监测需求，保障隧道运营安全，开展基于Scaramuzza畸变校正模型的全景相机地铁隧道管片提取技术研究。  方法 聚焦管片高精度监测和几何特征精准提取目标，基于Scaramuzza模型构建成像几何映射关系，建立齐次约束方程，系统求解相机内外参数，并通过逐步增阶验证法确定最优多项式阶数，实现标定效果优化，并以重投影误差作为标定精度的核心评价指标。经镜头畸变校正后，进一步构建隧道三维模型并实现管片分环展开，从线性特征恢复、几何形态还原和精度指标验证3个维度系统论证技术有效性。  结果 结果显示，Scaramuzza模型标定平均重投影误差优化至0.33像素，比Fisheye模型优化约37.9%，校正后标定板矩形最大周长偏差1.93 mm、最大角度偏差0.04°，线性特征与几何形态还原效果优异；三维模型连接点位置不确定性平均降低33.24%，数据质量显著提升。  结论 该方法兼具非接触、全覆盖、强鲁棒性优势，还可兼顾监测精度、效率与成本的平衡，为类似场景的精细化监测提供了思路，具有一定的工程实践价值。

关键词:地铁隧道;管片监测;全景相机;畸变校正;三维模型

doi:10.16186/j.cnki.1673-9787.2026020009

基金项目:国家自然科学基金资助项目（42374049）；山东省自然科学基金资助项目（ZR2022MD103）；山东省重点研发计划项目（2019GSF111052）

收稿日期:2026/02/10

修回日期:2026/04/24

出版日期:2026/06/17

Segment extraction method for metro tunnel linings based on the Scaramuzza distortion correction model

Sang Wengang, Zhu Youzhi, Sun Junfeng, Yin Yibo

School of Surveying and Geo-Informatics， Shandong Jianzhu University， Ji’nan 250101， Shandong， China

Abstract: Objectives To overcome the limitations of high cost and complex data processing in traditional tunnel segment monitoring technologies， as well as the influence of wide-angle lens distortion of panoramic cameras on the reliability of segment feature extraction， this study investigates a metro tunnel segment extraction method based on the Scaramuzza distortion correction model using panoramic cameras to meet the requirements of routine and refined monitoring of metro tunnel segments and ensure tunnel operation safety.  Methods Focusing on high-precision segment monitoring and accurate extraction of geometric features， an imaging geometric mapping relationship is established based on the Scaramuzza model， and homogeneous constraint equations are constructed to estimate the intrinsic and extrinsic camera parameters. The optimal polynomial order is determined using a stepwise order-increasing verification method to optimize calibration performance， with reprojection error adopted as the core evaluation index. After lens distortion correction， a 3D tunnel model is constructed and ring-wise unfolding of tunnel segments is achieved. The effectiveness of the proposed method is validated from three aspects： linear feature restoration， geometric shape recovery， and accuracy evaluation.  Results Experimental results show that the average reprojection error of calibration based on the Scaramuzza model is reduced to 0.33 pixels， representing an improvement of approximately 37.9% compared with the Fisheye model. After distortion correction， the maximum perimeter deviation and maximum angular deviation of the calibration checkerboard are 1.93 mm and 0.04°， respectively. In addition， the average positional uncertainty of connection points in the 3D model is reduced by 33.24%， indicating improved data quality. Conclusions The proposed method features non-contact measurement， full-coverage detection， and strong robustness， while achieving a balance among monitoring accuracy， efficiency， and cost. It provides a feasible approach for refined monitoring of metro tunnel segments and similar application scenarios and demonstrates practical engineering value.

Key words:metro tunnel;segment monitoring;panoramic camera;distortion correction;3D model