地铁隧道表面病害视觉检测技术应用综述

Survey on Visual Inspection Technique for Surface Damages of Subway Tunnel Engineering

针对地铁隧道表面病害快速检测与识别需求,分别从硬件组成、软件架构与识别算法等三方面对地铁隧道表面病害视觉检测技术展开综述分析。首先梳理了隧道病害视觉检测技术涉及的硬件参数,并分析确定出符合地铁隧道病害快速巡检需求的“线阵相机+线激光源”组成框架;其次针对隧道衬砌图像采集方面,数据与位移同步、图像增强与拼接处理为隧道病害视觉检测与识别样本数据生成的关键;在算法识别性能方面,相较于图像信号处理与浅层模式识别方法,深度学习网络算法可取得更好的隧道病害图像识别效果。论文最后指出隧道表面病害视觉检测与识别技术仍面临的瓶颈与挑战,并对其未来可能的发展方向给出总结与展望。

According to the requirements of rapid detection and recognition of subway tunnel surface diseases,based on the von Neumann architecture,this paper summarized and analyzed the visual inspection technique for surface damages of subway tunnel engineering from three aspects:hardware composition,software architecture and recognition algorithm.Firstly,the hardware parameters involved in vision inspection technology of tunnel disease was reviewed.The composition framework of linear array camera and linear laser source,which met the requirements of rapid inspection of subway tunnel diseases,was determined.For dealing with the image data of tunnel lining,it was proposed that the synchronization of data and displacement,image enhancement and splicing processing steps were keys to the generation of training data for visual detection and recognition of tunnel diseases.In terms of the performances of recognition algorithms,compared with image signal processing and shallow pattern recognition methods,it was determined that the deep learning networks with higher algorithm complexity can achieve better recognition results for tunnel disease images.This paper pointed out that the relevant bottlenecks and challenging for the visual detection and identification technology of tunnel surface diseases,and gave a summary and some prospects of future development research.