本文已被:浏览 171次 下载 998次
投稿时间:2023-05-05 修订日期:2023-06-15
投稿时间:2023-05-05 修订日期:2023-06-15
中文摘要: 本文提出了一种基于RGB-W的电力通信网危险源检测方法,该方法将短距离无线定位信息做映射处理,得到一张与图像对应的置信值掩膜。该掩膜将通过旁路输入到不同的目标检测框架中(包括一阶检测框架和二阶检测框架),用以提高危险源检测结果的精确度,并减少误检和漏检。对于这两类检测方法,我们使用基于无线电定位的置信值修正来提高检测性能。针对电力通信网的危险源检测任务,本文搜集了特定场景下的数据集用以训练检测与测试检测方法。本文提出的危险源检测方法使用了平均精度指标、平均误检数以及正确检测占比作为指标进行评估。实验结果证明本文提出的危险源检测方法具有检测精度高的优点,并能够减少错误检测框的数量。
Abstract:Hazard detection is an important computer vision area in electric power communication network operation and maintenance. In this paper, we propose a hazard detection method in electric power communication network.This method turns the short-range wireless localization information into a confidence mask corresponding to the image size. The mask will be input into different object detection frameworks, including the one-stage detector and the two-stage detector, to improve the accuracy of the detections and alleviate the problems of false positives and false negatives. For both detectors, we use the confidence score revision based on the wireless localization to improve the detection performance. In addition, datasets containing specific scenarios are collected to train and test the proposed method of hazard detection task in electric power communication network. Finally, this paper adopts various metrics to evaluate the effectiveness of the proposed method. The experiments demonstrate the feasibility and superiority of the proposed method in both theory and practice.
keywords: object detection wireless localization deep learning one-stage detector two-stage detector confidence mask electriconfidence maskcation network
文章编号: 中图分类号: 文献标志码:
基金项目:
作者 | 单位 | |
邱成润* | 国网无锡供电公司 | 940890508@qq.com |
管伟 | 国网盐城供电公司 | |
邱奇 | 国网无锡供电公司 | |
秦奕 | 国网无锡供电公司 | |
李双岑 | 国网无锡供电公司 | |
贾婷 | 国网无锡供电公司 | |
周琪 | 国网无锡供电公司 |
引用文本: