基于文本信息的地下排水管道缺陷分类方法  

Classification Method of Underground Drainage Pipe Defects Based on Textual Information

作　　者：张杰[1,2] 张婷雨 马龙 叶天昊 何敏 杜海鹏[5] ZHANG Jie;ZHANG Tingyu;MA Long;YE Tianhao;HE Min;DU Haipeng(School of Computer Science and Engineering,Xi'an University of Technology,Xi'an 710048,China;Shaanxi Provincial Key Laboratory of Network Computing and Security Technology,Xi'an 710048,China;Northwest Engineering Corporation Limited,Xi'an 710065,China;School of Civil Engineering and Architecture,Xi'an University of Technology,Xi'an 710048,China;School of Continuing Education,Xi'an Jiaotong University,Xi'an 710049,China)

机构地区：[1]西安理工大学计算机科学与工程学院,西安710048 [2]陕西省网络计算与安全技术重点实验室,西安710048 [3]中国电建西北勘测设计研究有限公司,西安710065 [4]西安理工大学土木建筑工程学院,西安710048 [5]西安交通大学继续教育学院,西安710049

出　　处：《给水排水》2025年第3期162-168,共7页Water & Wastewater Engineering

基　　金：国家重点研发计划(2022YFB2602200)。

摘　　要：城市地下排水管网的有效维护和管理对于保障城市基础设施的正常运行至关重要。然而,现有的排水管网缺陷排查方法存在人为主观判断强、视觉分类准确度不高等问题,严重制约了智慧城市建设进程。为解决这一难题,提出一种基于管网检测文本数据分析的地下排水管道缺陷分类方法。首先针对管网检测数据中的非结构化信息进行tf-idf文本离散化处理,将文本信息转化为可量化的特征向量。然后采用决策树方法对结构化信息进行处理,实现对管道缺陷的精准分类。在不同城市实际供水管网监测数据集上进行验证,结果表明,与多种分类方法相比,所提分类方法在各项性能指标上均为最优。非结构化信息的引入显著提高了现有分类方法的准确性,为管网缺陷分类提供了一种高效可靠的解决方案,为智慧城市建设中的基础设施管理提供了有力支持。Effective maintenance and management of urban underground drainage networks are crucial for ensuring the proper functioning of urban infrastructure.However,existing methods for detecting defects in drainage networks suffer from strong subjective judgments and low accuracy in visual classification,which severely hinder the progress of smart city development.To address this challenge,we propose a classification method for underground drainage pipe defects based on the analysis of textual data from network inspections.First,the unstructured information from the inspection data is discretized using the TF-IDF algorithm,converting the textual information into quantifiable feature vectors.Then,a decision tree method is applied to the structured data to achieve precise classification of pipe defects.Validation using actual water supply network monitoring datasets from various cities shows that,compared to multiple other classification methods,the proposed method performs optimally in all metrics.The incorporation of unstructured information significantly enhances the accuracy of existing classification methods,providing an efficient and reliable solution for defect classification in drainage networks and offering strong support for infrastructure management in smart city development.

关 键 词：非结构化信息 文本分类 决策树 管网缺陷 

分 类 号：TU990[建筑科学—市政工程]