现代电子技术2025，Vol.48Issue(19)：84-91,8.DOI:10.16652/j.issn.1004-373x.2025.19.014
基于双流特征融合的配电房异常入侵检测研究
Research on distribution room abnormal intrusion detection based on two-stream feature fusion
摘要
Abstract
An anomaly intrusion detection method based on two-stream and two-stage feature fusion is proposed to avoid missed and false detections caused by changes in perspective,motion blur,occlusion,and other factors in the application of conventional intrusion detection methods in power distribution rooms.A lightweight two-stream network channel is constructed for feature extraction of appearance images and optical flow images.The two-stream and two-stage fusion strategy achieves effective fusion of two kinds of features.Experimental verification was conducted on the monitoring dataset.And then,the proposed method was applied to scene detection of power distribution rooms,and its detection accuracy was better than that of the YOLOv8n benchmark model.The experimental results show that the model has high generalization ability,can reduce duplicate data collection and annotation and optimize model training and deployment.Therefore,it has practical application value.关键词
配电站房视频分析/目标检测/注意力机制/双流网络/自适应权重/复杂场景/YOLOv8n/CBAMKey words
video analysis of distribution room/object detection/attention mechanism/two-stream network/adaptive weight/complex scene/YOLOv8n/CBAM分类
信息技术与安全科学引用本文复制引用
郑洋斌,侯北平,邵方坤,曹志文..基于双流特征融合的配电房异常入侵检测研究[J].现代电子技术,2025,48(19):84-91,8.基金项目
浙江省"尖兵""领雁"研发攻关计划项目(2022C04012) （2022C04012）
