排水管道是城市水循环系统的静脉血管，用于收集各种污水或雨水并将其输送到污水处理厂或河流，对水资源的可持续利用和防止城市内涝起到关键作用。高效地对排水管道进行缺陷检测和量化，对维护其正常功能起到决定性作用。为了自动化地识别和量化管道缺陷，本文首先建立了一种适用于排水管道的三维重建方法，然后构建了一种基于管道三维模型进行结构性缺陷检测、识别和量化的算法。基于该算法开发了一套排水管道结构性缺陷识别与量化系统。该系统可读取管道三维模型并识别和量化模型中存在的管道缺陷，同时提供可交互的自由浏览管道模型和缺陷的渲染界面，显著提升了排水管道的检测效率。主要内容如下：（1）为了获得管道三维模型，本文建立了一种排水管道三维重建方法。使用高像素鱼眼镜头采集管道视频数据并抽取固定间隔的管道图像；使用COLMAP开源框架恢复相机位姿并重建管道的稀疏点云；利用OpenMVS开源框架完成稠密点云重建、网格重建、纹理重建3个步骤，获得逼近真实管道的管道三维模型；最后利用管径数据对管道三维模型进行尺度标定。对比尺度标定后的管道模型各项测量值与实际管道测量值，相对误差平均为2.3%，最大为6.2%。（2）为了自动化检测和量化管道缺陷，本文针对城市排水管道中的4种主要结构性缺陷（错口、脱节、起伏和变形），提出了一种基于管道三维模型的管道结构性缺陷检测、识别和量化算法。该算法通过引入ROSA广义旋转对称轴计算高鲁棒性的管道中心线，并利用管道中心线以及模型面片法向量自动检测和识别管道模型中的4种主要结构性缺陷并进行量化和缺陷等级划分。在一段84 m长、具有24处结构性缺陷的测试管道模型中，算法的缺陷召回率为95.8%，准确率为80.8%。（3）基于管道三维重建结果和管道结构性缺陷识别与量化算法，本研究使用C++和OpenGL开发了一套排水管道结构性缺陷识别与量化系统。该系统可读取管道三维模型并自动完成管道缺陷的检测、识别和量化，最后输出可自由浏览管道模型的渲染界面，以及可调整渲染参数和包含详细缺陷信息的操作界面。利用该系统能够快速、准确地识别和量化管道结构性缺陷，直观、整体地观察管道状况，可以极大提高管道检测的效率。
Drainage pipeline is the vein of urban water circulation system, which is used to collect various sewage or rainwater and transport it to sewage treatment plants or rivers. It plays a key role in the sustainable use of water resources and the prevention of urban waterlogging. Efficient defect detection and quantification of drainage pipelines play a decisive role in maintaining their normal function. In order to automatically identify and quantify pipeline defects, this article first establishes a 3D reconstruction method suitable for drainage pipelines, and then constructs an algorithm for structural defect detection, recognition, and quantification based on pipeline 3D models. A system for identifying and quantifying structural defects in drainage pipelines has been developed based on this algorithm. This system can read the three-dimensional model of pipelines and identify and quantify pipeline defects in the model. At the same time, it provides an interactive and free browsing interface for pipeline models and defect rendering, significantly improving the detection efficiency of drainage pipelines. The main content is as follows:(1) In order to obtain a three-dimensional model of the pipeline, this article establishes a three-dimensional reconstruction method for drainage pipelines. High pixel fisheye lens lens is used to collect pipeline video data and extract pipeline images at fixed intervals; Using the COLMAP open-source framework to restore camera pose and reconstruct sparse point clouds of pipelines; Using the OpenMVS open-source framework to complete three steps of dense point cloud reconstruction, mesh reconstruction, and texture reconstruction, obtaining a 3D model of the pipeline that approximates the real pipeline; Finally, scale the three-dimensional model of the pipeline using pipe diameter data. Comparing the measured values of the calibrated pipeline model with the actual pipeline measurements, the average relative error is 2.3%, with a maximum of 6.2%.(2) In order to automate the detection and quantification of pipeline defects, this article proposes a pipeline structural defect detection, recognition, and quantification algorithm based on a three-dimensional pipeline model for four main structural defects (staggered, disjointed, undulating, and deformed) in urban drainage pipelines. This algorithm calculates a highly robust pipeline centerline by introducing ROSA generalized rotational symmetry axis, and automatically detects and identifies four main structural defects in the pipeline model using the pipeline centerline and model surface normal vectors, and quantifies and classifies defect levels. In an 84 meter long test pipeline model with 24 structural defects, the algorithm has a defect recall rate of 95.8% and an accuracy rate of 80.8%.(3) Based on the results of three-dimensional reconstruction of pipelines and algorithms for identifying and quantifying structural defects in pipelines, this study developed a drainage pipeline structural defect identification and quantification system using C++and OpenGL. This system can read the three-dimensional model of pipelines and automatically detect, identify, and quantify pipeline defects. Finally, it outputs a rendering interface that allows for free browsing of pipeline models, as well as an operation interface that can adjust rendering parameters and contain detailed defect information. By utilizing this system, structural defects in pipelines can be quickly and accurately identified and quantified, and pipeline conditions can be observed intuitively and comprehensively, greatly improving the efficiency of pipeline inspection.
