基于领域知识增强的电力安全违章视觉情报即时检索与分析

doi:10.3976/j.issn.1002-4026.2025165

山东科学

• 情报分析与数据处理 •

基于领域知识增强的电力安全违章视觉情报即时检索与分析

龙春福¹，赵子龙¹，赵海¹，徐文斌¹，陈佳佳¹，唐昊煜^2*

收稿日期：2025-11-28 修回日期：2025-12-05

基金项目: 浙江省电网公司科学技术项目(CF058809002023001)；

作者简介：龙春福(1979—)，男，高级工程师，研究方向为电力安全、视频理解.E-mail: guowanglongchunfu@gmail.com

*通信作者，唐昊煜(1994—)，男，博士，助理研究员，主要研究方向为视频理解、多媒体计算、电网智慧巡检.E-mail: tanghao258@sdu.edu.cn

1. 乐清市电力实业有限公司，浙江温州 325600； 2. 山东大学软件学院，山东济南 250101

收稿日期:2025-11-28 接受日期:2025-12-05 上线日期:2026-10-09
通信作者: 唐昊煜 E-mail:tanghao258@sdu.edu.cn
作者简介:龙春福(1979—)，男，高级工程师，研究方向为电力安全、视频理解.E-mail: guowanglongchunfu@gmail.com
基金资助:
浙江省电网公司科学技术项目(CF058809002023001)

Domain knowledge–driven real-time visual intelligence retrieval and analysis for power safety violations

LONG Chunfu¹, Zhao Zilong¹, ZHAO Hai¹, XU Wenbin¹, CHEN Jiajia¹, TANG Haoyu^2*

1. Yueqing Electric Power Industry Co., Ltd., Wenzhou, 325600, China; 2. School of Software, Shandong University, Jinan, China, 250101

Received:2025-11-28 Accepted:2025-12-05 Online:2026-10-09
Contact: TANG Haoyu E-mail:tanghao258@sdu.edu.cn

摘要/Abstract

摘要： 为解决非结构化视频与结构化安全规则之间的语义鸿沟和实时监测中的时序知识不完备问题，本框架以《电力安全工作规程》为核心，首先构建了涵盖六元组模式的电力作业情报知识库及其图谱，并通过显性知识映射与语义规范化模块，将用户口语化查询转化为机器可计算的规范化显性规则知识作为先验约束。在此基础上，知识驱动的视觉特征筛选机制利用知识引导聚焦于视频中的关键违章主体与行为，实现对高冗余背景信息的有效抑制与数据清洗；时序协同推理架构，通过知识迁移策略，将离线环境下的时序演化规律转化为隐性知识，解决了实时监测中因未来信息缺失导致的检索滞后问题。实验在ActivityNet Captions及自建电力视觉违章行为数据集上开展，结果表明，该框架显著提升违章情报检索精度和效率，有效实现了非结构化数据向高价值结构化情报的转化，为电力安全预警服务提供了更具时效性和可解释性的解决方案。

关键词: 领域知识增强, 视觉情报检索, 知识图谱, 实时预警, 电力安全, 知识迁移

Abstract: To bridge the semantic gap between unstructured video data and structured safety rules, and to address incomplete temporal knowledge in real-time monitoring, this framework takes the Electrical Safety Work Regulations as its core. First, it constructs a power operation intelligence knowledge base and an associated knowledge graph using a six-tuple representation. Through an explicit knowledge mapping and semantic normalization module, users’ colloquial queries are transformed into machine-computable, standardized explicit rule knowledge that serves as prior constraints. Building on this foundation, a knowledge-driven visual feature selection mechanism leverages domain knowledge to focus on key violation subjects and behaviors in video streams, thereby effectively suppressing highly redundant background information and enabling data cleaning. Furthermore, a temporal collaborative reasoning framework transforms temporal evolution patterns learned in offline environments into tacit knowledge via knowledge transfer strategies, mitigating retrieval latency caused by the absence of future information in real-time monitoring. Experiments conducted on the ActivityNet Captions dataset and a self-constructed power visual violation behavior dataset demonstrate that the proposed framework significantly improves both the accuracy and efficiency of violation intelligence retrieval. Overall, the framework enables the transformation of unstructured video data into high-value structured intelligence and provides a more timely and interpretable solution for power safety early warning services.

Key words: domain knowledge enhancement, visual intelligence retrieval, knowledge graphs, real-time early warning, power safety, knowledge transfer

中图分类号:

TP311.13

龙春福, 赵子龙, 赵海, 徐文斌, 陈佳佳, 唐昊煜. 基于领域知识增强的电力安全违章视觉情报即时检索与分析[J]. 山东科学, 0, (): 1-.

LONG Chunfu, Zhao Zilong, ZHAO Hai, XU Wenbin, CHEN Jiajia, TANG Haoyu. Domain knowledge–driven real-time visual intelligence retrieval and analysis for power safety violations[J]. Shandong Science, 0, (): 1-.

参考文献

[1] 张森达, 崔昕, 曲延刚, 等. 电力领域知识图谱的构建设计[J]. 智能城市, 2025, 11(10): 66-69. DOI:10.19301/j.cnki.zncs.2025.10.015.

[2] 任晓龙, 陈曦, 司恒斌, 等. 基于Neo4j图数据库的电力系统与综合能源系统知识图谱研究[J]. 电力科学与技术学报, 2025, 40(3): 211-221. DOI:10.19781/j.issn.1673-9140.2025.03.023.

[3] Xu Mingze, Gao Mingfei, Chen Yiting, et al. Temporal recurrent networks for online action detection[C]//2019 IEEE/CVF International Conference on Computer Vision (ICCV). Seoul, Korea. IEEE, 2019: 5531-5540. DOI:10.1109/ICCV.2019.00563.

[4] Wang Shaokun, Gan Tian, Liu Yuan, et al. Discover micro-influencers for brandsviabetter understanding[J]. IEEE Transactions on Multimedia, 2022, 24: 2595-2605. DOI:10.1109/TMM.2021.3087038.

[5] Gao Jiyang, Sun Chen, Yang Zhenheng, et al. TALL: Temporal activity localizationvialanguage query[C]//2017 IEEE International Conference on Computer Vision (ICCV). Venice, Italy. IEEE, 2017: 5277-5285. DOI:10.1109/ICCV.2017.563.

[6] Yuan Yitian, Mei Tao, Zhu Wenwu. To find where you talk: Temporal sentence localization in video with attention based location regression[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2019, 33(1): 9159-9166. DOI:10.1609/aaai.v33i01.33019159.

[7] Meng Lingwen, Wang Yulin, Ban Guobang, et al. A multi-source embedding-based named entity recognition model for knowledge graph and its application to on-site operation violations in power grid systems[J]. Electronics, 2025, 14(13): 2511. DOI:10.3390/electronics14132511.

[8] Xiao Na, Peng Bai, Li Xin, et al. Research on the construction and implementation of power grid fault handling knowledge graphs[J]. Energy Reports, 2023, 9: 182-189. DOI:10.1016/j.egyr.2023.02.073.

[9] Liu Meng, Qu Leigang, Nie Liqiang, et al. Iterative local-global collaboration learning towards one-shot video person re-identification[J]. IEEE Transactions on Image Processing, 2020, 29: 9360-9372. DOI:10.1109/TIP.2020.3026625.

[10] Zhao Yang, Lin Xuan, Xu Wenqiang, et al. AntPivot: Livestream highlight detectionviahierarchical attention mechanism[PP/OL]. arXiv (2022-06-10)[2025-11-01]. https://doi.org/10.48550/arXiv.2206.04888.

[11] Xu Mingze, Xiong Yuanjun, Chen Hao, et al. Long short-term transformer for online action detection[PP/OL]. V3. arXiv (2021-12-22)[2025-11-01]. https://doi.org/10.48550/arXiv.2107.03377.

[12] Wang Xiang, Zhang Shiwei, Qing Zhiwu, et al. OadTR: Online action detection with transformers[C]//2021 IEEE/CVF International Conference on Computer Vision (ICCV). Montreal, QC, Canada. IEEE, 2021: 7545-7555. DOI:10.1109/ICCV48922.2021.00747.

[13] Wu Jianlong, Sun Wei, Gan Tian, et al. Neighbor-guided consistent and contrastive learning for semi-supervised action recognition[J]. IEEE Transactions on Image Processing, 2023, 32: 2215-2227. DOI:10.1109/TIP.2023.3265261.

[14] Cai Jianghui, Zhang Min, Yang Haifeng, et al. A novel graph-attention based multimodal fusion network for joint classification of hyperspectral image and LiDAR data[J]. Expert Systems with Applications, 2024, 249: 123587. DOI:10.1016/j.eswa.2024.123587.

[15] Krishna R, Hata K, Ren F, et al. Dense-captioning events in videos[C]//2017 IEEE International Conference on Computer Vision (ICCV). Venice, Italy. IEEE, 2017: 706-715. DOI:10.1109/ICCV.2017.83.

[16] Chen Junwen, Mittal G, Yu Ye, et al. GateHUB: Gated history unit with background suppression for online action detection[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans, LA, USA. IEEE, 2022: 19893-19902. DOI:10.1109/CVPR52688.2022.01930.

[17] CAO Y, WANG L, ZHOU J, et al. Iterative refinement of pseudo labels for weakly supervised temporal action localization[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2023: 18427-18437.

[18] Gan Tian, Wang Xiao, Sun Yan, et al. Temporal sentence grounding in streaming videos[C]//Proceedings of the 31st ACM International Conference on Multimedia. Ottawa ON Canada. ACM, 2023: 4637-4646. DOI:10.1145/3581783.3612120.

基于领域知识增强的电力安全违章视觉情报即时检索与分析

Domain knowledge–driven real-time visual intelligence retrieval and analysis for power safety violations

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 2

Metrics

本文评价

推荐阅读 0

[1]	王朋宇, 姜树明, 魏志强, 余军, 张萌萌. 基于大语言模型的教育事业统计指标体系知识图谱构建[J]. 山东科学, 2026, 39(4): 132-142.
[2]	郭梦萦, 贾辛欣, 黄赟杰, 刘颖莹. 基于CiteSpace的国内科技强省与科技强国研究热点及趋势分析[J]. 山东科学, 2026, 39(3): 129-138.