基于目标提取与SVM的排洪隧洞裂缝视觉检测算法研究  

Research on Flood-discharging Tunnel Crack Visual Detection Algorithm Based on Target Extraction and SVM

作　　者：许霄煜 郭振武 周唯逸 王斌锐 XU Xiaoyu;GUO Zhenwu;ZHOU Weiyi;WANG Binrui(College of Mechanical and Electrical Engineering,China JiLiang University,Hangzhou 310000,China;Yueqing Power Supply Company,Zhejiang Electric Power Corporation,Leqing 325600,China)

机构地区：[1]中国计量大学机电工程学院,杭州310000 [2]国网浙江省电力有限公司乐清市供电公司,浙江乐清325600

出　　处：《有色金属工程》2022年第6期101-111,共11页Nonferrous Metals Engineering

基　　金：国家重点研发计划项目(2018YFB2101004)。

摘　　要：尾矿库排洪隧洞裂缝会对隧洞的正常运行造成安全隐患,对其进行自动检测具有重要意义。针对排洪隧洞场景特点,研制了隧洞衬砌表面图像采集系统,设计了预处理算法与两级滤波算法完成目标提取,其中利用背景滤波减少了处理时间。通过支持向量机分类判断目标区域是否为真实的裂缝,提高了裂缝检测的精度。在真实的隧洞环境中对整套系统进行测试和验证,对裂缝图像的识别率达到90%,证实了硬件系统与算法的有效性。The cracks in the flood-discharging tunnel of tailings pond will cause potential safety hazards to the normal operation of the tunnel,and the automatic detection is of great significance.Aiming at scene characteristics inside the flood discharge tunnel,an image acquisition system of the tunnel lining surface was developed,and a preprocessing algorithm and a multi-level filtering algorithm were designed to extract the target,in which background filtering is used to reduce the processing time.The target was classified by the support vector machine to determine whether it was a real crack,which improved the accuracy of crack detection.The system was tested and verified in a real tunnel environment,and the recognition rate of crack image reached 90%,which confirmed the effectiveness of the hardware system and algorithm.

关 键 词：机器视觉 图像处理 裂缝检测算法 目标提取 支持向量机 

分 类 号：TD926.4[矿业工程—选矿] TP3[自动化与计算机技术—计算机科学与技术]