基于YOLO和自然语言处理的施工方案智能审核方法

doi:10.11996/JG.j.2095-302X.2025051113

图学学报 ›› 2025, Vol. 46 ›› Issue (5): 1113-1122.DOI: 10.11996/JG.j.2095-302X.2025051113

基于YOLO和自然语言处理的施工方案智能审核方法

钱增志¹(), 孙玉龙¹, 张捷¹, 夏侯遐迩²(), 周大兴¹, 康伟德¹

¹ 中铁建设集团有限公司，北京 100040
² 东南大学土木工程学院，江苏南京 211189

收稿日期:2025-01-26 接受日期:2025-04-10 出版日期:2025-10-30 发布日期:2025-09-10
通讯作者:夏侯遐迩(1988-)，男，副教授，博士。主要研究方向为智能建造与运维(安全)管理等。E-mail：xhcmre@seu.edu.cn
第一作者:钱增志(1973-)，男，正高级工程师，硕士。主要研究方向为建筑工程与智能建造。E-mail：qianzengzhi.ztjs@crcc.cn
基金资助:
国家重点研发计划项目(2023YFC3804302);国家自然科学基金(72101054)

A construction plan intelligent review method based on YOLO and natural language processing

QIAN Zengzhi¹(), SUN Yulong¹, ZHANG Jie¹, XIAHOU Xiaer²(), ZHOU Daxing¹, KANG Weide¹

¹ China Railway Construction Group Co., Ltd., Beijing 100040, China
² School of Civil Engineering, Southeast University, Nanjing Jiangsu 211189, China

Received:2025-01-26 Accepted:2025-04-10 Published:2025-10-30 Online:2025-09-10
First author：QIAN Zengzhi (1973-), senior engineer, master. His main research interests cover construction engineering and intelligent construction. E-mail：qianzengzhi.ztjs@crcc.cn
Supported by:
National Key Research and Development Program of China(2023YFC3804302);National Natural Science Foundation of China(72101054)

摘要/Abstract

摘要：

建筑施工方案的人工审核存在重复性高、耗时费力且占用专家资源等问题。为提升审核效率，促进企业智能建造发展，提出一种施工方案智能审核方法，通过审核规则编制、向量模型构建及图像文字识别技术结合，实现多类型方案智能审核。审核规则基于集团技术文件和历史审核意见样本，经历史高频审核意见筛选与专家研判后，采用正则表达式技术进行逐条编制。研究构建了基于语义相似度比对的审核模式，将方案文本内容嵌入向量空间，通过计算向量余弦相似度实现语义比对，从而提升审核灵活性与容错性。同时，引入基于YOLO的图像文字识别技术，处理文档图片中的文字内容，确保审核全覆盖。实验结果显示，系统平均审核准确率达90.4%，审核时间效率较人工提升87.9%，且能处理多种格式文本输入，具备良好鲁棒性，显著提高审核工作效率，对企业数字化转型与智能建造技术普及具有重要推动作用。目前，搭载此审核技术的平台已在集团多家分公司与项目应用测试，能生成准确的审核报告，实现审核效率显著提升。

关键词: 施工方案, 智能审核, 自然语言处理, 语义相似度, 图像文字识别

Abstract:

The manual review of construction plans in the building industry suffers from high repetitiveness, substantial time consumption, and extensive expert resource usage. To improve review efficiency and promote intelligent construction development, an intelligent construction plan review method was proposed, integrating review rule compilation, vector model construction, and image-text recognition to achieve intelligent review of multiple types of plans. The review rules were based on group technical documents and historical review samples, filtered through high-frequency historical review comments and expert judgment, then compiled item by item using regular expression technology. A review model based on semantic similarity comparison was constructed, embedding plan text content into vector space and implementing semantic comparison through vector cosine similarity calculation, thereby enhancing review flexibility and fault tolerance. Additionally, YOLO-based image text recognition technology was incorporated to process textual content in document images, ensuring comprehensive review coverage. Experimental results showed an average review accuracy of 90.4% and an 87.9% improvement in time efficiency compared to manual review. The system can process multiple text format inputs with robust performance, significantly improving review work efficiency and playing an important role in promoting enterprise digital transformation and the popularization of intelligent construction technology. Currently, the platform equipped with this review technology was tested in multiple branches and projects of the group, generating accurate review reports and delivering significant improvement in review efficiency.

Key words: construction plans, intelligent review, natural language processing, semantic similarity, image text recognition

中图分类号:

钱增志, 孙玉龙, 张捷, 夏侯遐迩, 周大兴, 康伟德. 基于YOLO和自然语言处理的施工方案智能审核方法[J]. 图学学报, 2025, 46(5): 1113-1122.

QIAN Zengzhi, SUN Yulong, ZHANG Jie, XIAHOU Xiaer, ZHOU Daxing, KANG Weide. A construction plan intelligent review method based on YOLO and natural language processing[J]. Journal of Graphics, 2025, 46(5): 1113-1122.

图/表 17

参考文献 25

[1]	肖智铎, 高俊华, 王长园. 基于语义识别和机器视觉的智能审核方法研究[J]. 中国汽车, 2023(6): 52-57, 64.
	XIAO Z D, GAO J H, WANG C Y. Research on intelligent audit method based on semantic recognition and machine vision[J]. China Auto, 2023(6): 52-57, 64 (in Chinese).
[2]	陈彦. 基于自然语言处理的合同智能审核系统的研究与实现[D]. 绵阳: 西南科技大学, 2023.
	CHEN Y. Research and implementation of contract intelligent review system based on natural language processing[D]. Mianyang: Southwest University of Science and Technology, 2023 (in Chinese).
[3]	柳藴栖. 机场无线电导航原始资料智能审核研究[D]. 广汉: 中国民用航空飞行学院, 2023.
	LIU Y X. Research on intelligent review of airport radio navigation source material[D]. Guanghan: Civil Aviation Flight University of China, 2023 (in Chinese).
[4]	翟社平, 李兆兆, 段宏宇, 等. 基于词法、句法和语义的句子相似度计算方法[J]. 东南大学学报(自然科学版), 2019, 49(6): 1094-1100.
	ZHAI S P, LI Z Z, DUAN H Y, et al. Sentence similarity calculation method based on lexical, syntactic and semantic[J]. Journal of Southeast University (Natural Science Edition), 2019, 49(6): 1094-1100 (in Chinese).
[5]	赵晓康, 牛振兴, 张久鹏, 等. 基于YOLOv5-s的机场道面异物目标检测[J]. 东南大学学报(自然科学版), 2024, 54(5): 1239-1250.
	ZHAO X K, NIU Z X, ZHANG J P, et al. Foreign object debris detection on airport pavement based on YOLOv5-s[J]. Journal of Southeast University (Natural Science Edition), 2024, 54(5): 1239-1250 (in Chinese).
[6]	穆磊, 王佳, 李继宝, 等. 基于BIM和知识图谱的消防智能审图研究[J]. 消防科学与技术, 2019, 38(12): 1765-1768.
	MU L, WANG J, LI J B, et al. On intelligent review of fire drawings based on building information modeling and knowledge graph[J]. Fire Science and Technology, 2019, 38(12): 1765-1768 (in Chinese).
[7]	林佳瑞, 郭建锋. 基于BIM的合规性自动审查[J]. 清华大学学报(自然科学版), 2020, 60(10): 873-879.
	LIN J R, GUO J F. BIM-based automatic compliance checking[J]. Journal of Tsinghua University (Science & Technology), 2020, 60(10): 873-879 (in Chinese).
[8]	LI H, YANG R Z, XU S S, et al. Intelligent checking method for construction schemes via fusion of knowledge graph and large language models[J]. Buildings, 2024, 14(8): 2502.
[9]	李梦晨. 基于人工智能的互联网内容审核模式研究[D]. 北京: 北京邮电大学, 2022.
	LI M C. Research on internet content audit mode based on artificial intelligence[D]. Beijing: Beijing University of Posts and Telecommunications, 2022 (in Chinese).
[10]	董志龙. 工程设计文件的施工图纸审核工作[J]. 建筑工人, 2017, 38(10): 6-11.
	DONG Z L. Construction drawing review work of engineering design documents[J]. Builders’ Monthly, 2017, 38(10): 6-11 (in Chinese).
[11]	刘小波, 谢芊, 李留英. 应用正则表达式在ASP. NET中实现优化的输入验证方法[J]. 现代图书情报技术, 2005(10): 80-83.
	LIU X B, XIE Q, LI L Y. Research of optimized input checking method based on ASP. NET with regular expression[J]. Data Analysis and Knowledge Discovery, 2005(10): 80-83 (in Chinese).
[12]	刘江波. 地铁盾构施工规范知识图谱构建及应用研究[D]. 西安: 长安大学, 2023.
	LIU J B. Knowledge graph building and application research of metro shield construction specifications[D]. Xi’an: Chang’an University, 2023 (in Chinese).
[13]	ZHANG J S, EL-GOHARY N. Automated regulatory information extraction from building codes: leveraging syntactic and semantic information[EB/OL]. [2024-11-26]. https://doi.org/10.1061/9780784412329.063.
[14]	FRANCISCO L, LAGARE T, JAIN A, et al. Design rule checking with a CNN based feature extractor[C]// 2020 2nd Workshop on Machine Learning for CAD. New York: ACM, 2020: 9-14.
[15]	DING Y X, MA J, LUO X W. Applications of natural language processing in construction[J]. Automation in Construction, 2022, 136: 104169.
[16]	林佳瑞, 陈柯吟, 郑哲, 等. 建筑工程标准规范智能解译关键技术及应用[J]. 工程力学, 2025, 42(2): 1-14.
	LIN J R, CHEN K Y, ZHENG Z, et al. Key technologies and applications of intelligent interpretation of building engineering standards[J]. Engineering Mechanics, 2025, 42(2): 1-14 (in Chinese).
[17]	赵兴涛, 王斌君, 刘舒. 正则表达式在文档自动识别中的应用[J]. 中国人民公安大学学报(自然科学版), 2005, 11(4): 33-35.
	ZHAO X T, WANG B J, LIU S. Applcation of regular expression into automatic recognition of electric document[J]. Journal of People’s Public Security University of China (Science and Technology), 2005, 11(4): 33-35 (in Chinese).
[18]	王新, 徐照. 基于依存句法分析的建筑设计规范条文自动结构化方法[J]. 土木工程与管理学报, 2023, 40(4): 133-140.
	WANG X, XU Z. Automatic structural method of architectural design rules based on dependency parsing[J]. Journal of Civil Engineering and Management, 2023, 40(4): 133-140 (in Chinese).
[19]	张清慧, 陈谊, 武彩霞. 基于词表示模型的领域文献数据可视分析方法[J]. 图学学报, 2022, 43(4): 685-694.
	ZHANG Q H, CHEN Y, WU C X. A visual analysis approach for domain literature data based on word representation model[J]. Journal of Graphics, 2022, 43(4): 685-694 (in Chinese).
[20]	杨文娟, 王文明, 王全玉, 等. 基于感知哈希和视觉词袋模型的图像检索方法[J]. 图学学报, 2019, 40(3): 519-524. DOI
	YANG W J, WANG W M, WANG Q Y, et al. Image retrieval method based on perceptual hash algorithm and bag of visual words[J]. Journal of Graphics, 2019, 40(3): 519-524 (in Chinese).
[21]	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2016: 779-788.
[22]	FENG R Y, MIAO Y, ZHENG J X. A YOLO-based intelligent detection algorithm for risk assess-ment of construction sites[J]. Journal of Intelligent Construction, 2024, 2(4): 1-18.
[23]	EUM I, KIM J, WANG S, et al. Heavy equipment detection on construction sites using you only look once (YOLO-version 10) with transformer architectures[J]. Applied Sciences, 2025, 15(5): 2320.
[24]	胡军伟, 秦奕青, 张伟. 正则表达式在Web信息抽取中的应用[J]. 北京信息科技大学学报, 2011, 26(6): 86-89.
	HU J W, QIN Y Q, ZHANG W. Regular expression and its applications to web information extraction[J]. Journal of Beijing Information Science and Technology University, 2011, 26(6): 86-89 (in Chinese).
[25]	王煜, 邓晖, 李晓瑶, 等. 自然语言处理技术在建筑工程中的应用研究综述[J]. 图学学报, 2020, 41(4): 501-511. DOI
	WANG Y, DENG H, LI X Y, et al. A review of natural language processing application in construction engineering[J]. Journal of Graphics, 2020, 41(4): 501-511 (in Chinese). DOI

字符	描述	实例
\	转义符	\。可以匹配。
^	标记字符串的起始位置	^ab匹配以ab开始字符串
$	标记字符串的结束位置	ab$匹配以ab结尾字符串
*	匹配*前字符零次或多次	ab*可以匹配a和abbbb···
+	匹配+前字符一次或多次	ab+可以匹配ab和abbbb···
?	匹配?前字符零次或一次	ab?可以匹配a和ab
\|	匹配 \| 前或后的字符	a\|b可以匹配a，b或ab
.	匹配除\n和\r之外字符	ab.可以匹配abc，ab8···
{n}	匹配之前的表达式n次	ab{2}可以匹配abb
{n,m}	匹配之前的表达式最少n次，最多m次	ab{1,3}可以匹配ab， abb，abbb

字符	描述	实例
\	转义符	\。可以匹配。
^	标记字符串的起始位置	^ab匹配以ab开始字符串
$	标记字符串的结束位置	ab$匹配以ab结尾字符串
*	匹配*前字符零次或多次	ab*可以匹配a和abbbb···
+	匹配+前字符一次或多次	ab+可以匹配ab和abbbb···
?	匹配?前字符零次或一次	ab?可以匹配a和ab
\|	匹配 \| 前或后的字符	a\|b可以匹配a，b或ab
.	匹配除\n和\r之外字符	ab.可以匹配abc，ab8···
{n}	匹配之前的表达式n次	ab{2}可以匹配abb
{n,m}	匹配之前的表达式最少n次，最多m次	ab{1,3}可以匹配ab， abb，abbb

名称	具体环境
操作系统	Windows，Linux
开发环境	IDEA，PyCharm
开发语言	Java，Python
系统处理器	11th Gen Intel(R) Core(TM) i7
数据库	MySQL
施工方案数据来源	XXX集团施工文件资源管理系统

名称	具体环境
操作系统	Windows，Linux
开发环境	IDEA，PyCharm
开发语言	Java，Python
系统处理器	11th Gen Intel(R) Core(TM) i7
数据库	MySQL
施工方案数据来源	XXX集团施工文件资源管理系统

高支模专项方案审核类型	章节位置	审核内容	备注
章节目录完整性	目录	1.工程概况 2.编制依据 3.施工计划 4.施工工艺技术 5.施工保证措施 6.施工管理及作业人员配备和分工 7.验收要求 8.应急处置措施 9.计算书及相关图纸	审核标题内容及其顺序
高支模工程材料进场计划	3	审核是否有“高支模工程材料进场计划表”；对高支模材料的种类应对照后续施工工艺技术章节所采用的模架方案相一致	形式为表格，表格内容需包含：物资名称(材料名称)、规格型号、数量、进场时间
······	······	······	······

基于YOLO和自然语言处理的施工方案智能审核方法

A construction plan intelligent review method based on YOLO and natural language processing

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 17

参考文献 25

相关文章 1

编辑推荐

Metrics

本文评价

审核点需求	审核方案中的章节是否存在：质量标准及质量保证措施
相应正则表达式匹配模式	([\d\|[\u4e00-\u9fa5]])[.][、][．][\s][\u4e00-\u9fa5]质量[标准][要求]及[质量]*保证措施
可匹配的标题序列	“2. 质量标准及保证措施” (该标题以“阿拉伯数字+点+空格”开头，内容与需求有差异但总体正确，审核通过)
	“9 质量要求及质量保证措施” (该标题以“阿拉伯数字+空格”开头，内容与需求有差异但总体正确，审核通过)
	“7质量要求及保证措施” (该标题以“阿拉伯数字”开头，内容与需求有差异但总体正确，审核通过)
	“三、混凝土质量要求及质量保证措施” (该标题以“汉字数字+顿号”开头，内容存在“混凝土”干扰项，与需求有差异但总体正确，审核通过)
	“附件二：钢筋质量要求及保证措施” (该标题以“非编号+冒号”开头，内容存在“钢筋”干扰项，与需求有差异但总体正确，审核通过)

方案类型	审核条目数量/条	样本数量/份	正确审核条目均值/条	准确率/ %
高大(超限)模板	56	312	51.4	91.7
混凝土工程	41	422	36.8	89.7
模板工程	45	166	40.6	90.2
钢筋工程	48	200	42.6	88.8
二次结构	34	66	31.3	92.1