基于多模态大模型与目标检测的输电线路外破隐患识别算法的对比研究

迟钰坤

欢迎来到能源电力期刊网！

English Version 读者登录期刊登录

您当前的位置：

首页 >

文章列表页 >

基于多模态大模型与目标检测的输电线路外破隐患识别算法的对比研究

更新时间：2025-11-14

- 基于多模态大模型与目标检测的输电线路外破隐患识别算法的对比研究
- A Comparative Study of Transmission Line External Damage Hazard Identification Algorithms Based on Multimodal Large Models and Object Detection
- 电力大数据 2025年第7期
- 作者机构：
  
  山东鲁软数字科技有限公司
- 作者简介：
- 基金信息：
- DOI：
  中图分类号：
- 纸质出版：2025
- 稿件说明：
移动端阅览
迟钰坤. 基于多模态大模型与目标检测的输电线路外破隐患识别算法的对比研究[J]. 电力大数据, 2025,(7).

chiyukun. A Comparative Study of Transmission Line External Damage Hazard Identification Algorithms Based on Multimodal Large Models and Object Detection[J]. 2025, (7).
迟钰坤. 基于多模态大模型与目标检测的输电线路外破隐患识别算法的对比研究[J]. 电力大数据, 2025,(7). DOI：

chiyukun. A Comparative Study of Transmission Line External Damage Hazard Identification Algorithms Based on Multimodal Large Models and Object Detection[J]. 2025, (7). DOI：

摘要

为解决传统输电线路外破隐患检测中存在的准确率低、时效性差的问题，该文探索并提出了两类识别算法，分别基于多模态大模型和目标检测技术。相较于传统算法，该文提出的两类算法在性能上实现了显著提升。其中，基于多模态大模型Qwen2.5-VL的输电线路外破隐患识别算法展现出更强的泛化能力，能够有效识别多种类型的外破隐患，特别是在非常见隐患类型的检测方面表现突出。而基于YOLOv11目标检测框架的隐患识别算法则在检测效率与精度方面优势明显，在包含10000张输电线路可视化装置拍摄图像的测试集上，该算法实现了90.76%的准确率和91.01%的召回率。该文提出的基于目标检测的输电线路外破隐患识别算法，不仅具备识别准确率高、速度快的特点，还能在服务器及边缘设备侧实现快速部署，满足大规模推广应用的条件，对保障输电线路安全稳定运行具有重要的实践意义。

Abstract

To address the issues of low accuracy and poor timeliness in traditional transmission line external damage hazard detection

this paper explores and proposes two types of identification algorithms based on multimodal large models and object detection techniques

respectively. Compared to conventional algorithms

the proposed algorithms achieve significant performance improvements. Among them

the transmission line external damage hazard identification algorithm based on the multimodal large model Qwen2.5-VL demonstrates stronger generalization capabilities

effectively identifying various types of external damage hazards

particularly excelling in the detection of uncommon hazard types. In contrast

the hazard identification algorithm based on the YOLOv11 object detection framework exhibits notable advantages in detection efficiency and precision. On a test set comprising 10

000 images captured by visual devices installed on transmission lines

this algorithm achieves an accuracy of 90.76% and a recall rate of 91.01%. The proposed object detection-based algorithm for identifying external damage hazards in transmission lines not only features high identification accuracy and speed but also supports rapid deployment on both servers and edge devices

meeting the requirements for large-scale application. This holds significant practical importance for ensuring the safe and stable operation of transmission lines.

关键词

Keywords

references

黄新波.基于图像感知的输电线路智能巡检综述[J].高电压技术,2024,50(5):1826-1841,I0001- I0005.

吕志宁.输电线路常见故障分析与检测方法综述[J].自动化与仪器仪表,2020,(1):161-164,168.

董卓元,高永亮,袁斌,等.一种改进型YOLOv4输电线路防外破检测方法[J].电网与清洁能源,2023, 39(6):17-25.

高超,王真,吴奇伟,等.基于特定区域目标检测的输电通道有效外破隐患检测方法[J].电网与清洁能源,2025,41(1):44-49,60.

方刚,平学良,王铭民,等.基于边缘信息融合的输电线路防外破监测系统研究与应用[J].电子器件, 2023,46(1):210-217.

魏贤哲,卢武,赵文彬等.基于改进Mask R-CNN的输电线路防外破目标检测方法研究[J].电力系统保护与控制,2021,49(23):155-162.

吴精乙,景峻,贺熠凡等.基于多模态大模型的高速公路场景交通异常事件分析方法[J].图学学报,2024,45(6):1266-1276.

迟钰坤,王倩倩,焦之明,等.基于YOLO-V5的输电线路可视化中视觉分析关键技术的研究及应用[J]. 电力大数据,2022,25(11):20-28.

吕越,薛念明,邓昊,等.基于金字塔分割注意力机制的输电线路设备检测方法研究[J].电力大数据, 2025,28(3):47-57.

韦炎炎,毛天一,李柏昂,等.视觉模型及多模态大模型推进图像复原增强研究进展[J].中国图象图形学报,2025,30(5):1197-1219.

DOMONKOS VARGA.Comparative evaluation of Multimodal Large Language Models for No-Reference image quality assessment with authentic distortions: A study of OpenAI and Claude.AI Models[J].Big Data and Cognitive Computing,2025,9(5):132-132.

BO ZHANG, HUI MA , JIAN DING, et al. Distilling implicit multimodal knowledge into large language models for zero-resource dialogue generation[J].Information Fusion,2025, 118:102985-102985.

WEI ZHONG, YIFAN LIU, YAN LIU, et al. Performance of ChatGPT-4o and four Open-Source Large Language Models in generating diagnoses based on China''s rare disease catalog: comparative study[J]. Journal of Medical Internet Research, 2025,27,e69929.

郭园方,余梓彤,刘艾杉,等.多模态大模型安全研究进展[J].中国图象图形学报,2025,30(06): 2051-2081.

张立明,冉政,张容.多模态模型嵌入知识生成的逻辑机理与路径选择[J].图书与情报,2024(4): 81-89.

董磊,吴福居,史健勇,等.基于大语言模型的施工安全多模态知识图谱的构建与应用[J].计算机工程与应用,2025,61(9):325-333.

秦赛梅,文琼,段依恋,等.对比通义千问2.5与GPT-4o模型生成的甲状腺超声结构化报告[J].中国医学影像技术,2025,41(3):409-413.

吴雪,宋晓茹,高嵩,等.基于深度学习的目标检测算法综述[J].传感器与微系统,2021,40(2):4-7, 18.

HUI YAO, YANHAO LIU, XIN LI, et al. A???????? detection method for pavement cracks combining object detection and attention mechanism[J].IEEE Transactions on Intelligent Transportation Systems, 2022,23(11):22179-22189.

郭磊,王邱龙,薛伟,等.基于改进YOLOv5的小目标检测算法[J].电子科技大学学报,2022,51(2): 251-258.

胡峻峰,李柏聪,朱昊,等.改进YOLOv8的轻量化无人机目标检测算法[J].计算机工程与应用,2024,60(8):182-191.

YONGCAN YU, JIANHU ZHAO, QUANHUA GONG, et al. Real-time underwater maritime object detection in side-scan sonar images based on Transformer-YOLOv5[J]. Remote Sensing,2021,13(18):3555.

王杨,曹铁勇,杨吉斌,等.基于YOLO v5算法的迷彩伪装目标检测技术研究[J].计算机科学,2021, 48(10): 226-232.

?李彬,李生林.改进YOLOv11n的无人机小目标检测算法[J].计算机工程与应用,2025,61(7): 96-104.

?柯巍,朱权洁,陈长茂,等.基于改进YOLOV11的卷烟仓储人员不安全行为分类及识别[J].中国安全科学学报,2025,35(3):36-44.

李梦超,文雪风,姜攀.TrackDef-YOLO:一种改进的YOLOv11模型用于铁轨表面缺陷检测[J].人工智能与机器人研究,2025,14(1):217-223.

涂育智,王法翔,吴春霖.融合多注意力机制的轻量级无人机航拍小目标检测模型[J].计算机工程与应用,2025,61(11):93-104.