基于改进SSD的城市地下排水管道缺陷识别算法  

Defects of urban underground drainage pipes based on improved SSD recognition algorithm

作　　者：王治国 曹爽 管海燕 周弈 WANG Zhiguo;CAO Shuang;GUAN Haiyan;ZHOU Yi(School of Remote Sensing&Geomatics Engineering,Nanjing University of Information Science and Technology,Nanjing 210044,China;Technology Innovation Center for Integrated Applications in Remote Sensing and Navigation,Ministry of Natural Resources,Nanjing 210044,China;Jiangsu Engineering Center for Collaborative Navigation/Positioning and Smart Applications,Nanjing 210044,China;Zhejiang East China Surveying and Mapping and Engineering Safety Technology Co.,Ltd.,Hangzhou 310000,China)

机构地区：[1]南京信息工程大学遥感与测绘工程学院,南京210044 [2]自然资源部遥感导航一体化应用工程技术创新中心,南京210044 [3]江苏省协同精密导航定位与智能应用工程研究中心,南京210044 [4]浙江华东测绘与工程安全技术有限公司,杭州310000

出　　处：《测绘工程》2024年第5期7-13,共7页Engineering of Surveying and Mapping

基　　金：国家自然科学基金资助项目(41971298)。

摘　　要：城市地下排水管道的维护是保证排水管道使用功能的基础工作,对城市地下排水管道进行缺陷检查则是进行维护工作的必要前提。为了提高城市地下排水管道缺陷识别效率,实现管道缺陷识别智能化,提出一种基于改进SSD的城市地下排水管道缺陷识别算法。针对传统SSD算法特征提取不足的问题,设计了一种注意力机制模块,以提高城市地下排水管道缺陷特征提取能力;针对传统SSD算法感受野不足的问题,引入感受野扩增模块并调整其参数和结构,以捕获不同尺度的感受野;最后在自建城市地下排水管道缺陷数据集上进行实验。结果表明,该方法较传统SSD算法在城市地下排水管道缺陷识别上的平均准确率提高约7.89%,因此本方法对于城市地下排水管道缺陷识别有一定应用价值。The maintenance of urban underground drainage pipes is the basic work to ensure the operation function of drainage pipes,and the inspection of urban underground drainage pipes defect is the necessary prerequisite for maintenance.In order to improve the efficiency of urban underground drainage pipes defect identification and realize the intelligent identification of pipeline defect,a defect identification algorithm of urban underground drainage pipes based on improved SSD(Single Shot MultiBox Detector)algorithm was proposed.Aiming at the problem of insufficient feature extraction of traditional SSD algorithm,an attention mechanism module is designed to improve the feature extraction ability of urban underground drainage pipes.To solve the problem of insufficient receptive fields in traditional SSD algorithm,a receptive field amplification module is introduced and its parameters and structure are adjusted to capture different scale receptive fields.Finally,experiments are carried out on the self-built data set of urban underground drainage pipes.The results show that the average accuracy of this method is about 7.89%higher than that of the traditional SSD algorithm in the identification of urban underground drainage pipes.Therefore,this method has certain application value for in defect identification of urban underground drainage pipes.

关 键 词：地下排水管道 单次多框检测器 缺陷识别 注意力机制 感受野 

分 类 号：TU992[建筑科学—市政工程]