重庆理工大学学报2024，Vol.38Issue(17)：150-157,8.DOI:10.3969/j.issn.1674-8425(z).2024.09.019
一种SSD-RTF输电线路小目标检测算法研究
Research on a small target detection algorithm for SSD-RTF transmission lines
摘要
Abstract
To address the problems of the traditional transmission line UAV inspection image detection method,such as weak detection ability,high error detection and missing rate,and insufficient shallow network semantic information,this paper proposes a small target detection algorithm model of SSD-RTF transmission line.In the shallow network layer of VGG-16 in the original backbone of SSD algorithm,a visual mechanism is added to enlarge the receptive field and a three-pronged trunk feature fusion module is introduced to extract multi-scale features of the feature map to increase the robustness of the feature map.FusionNet shallow feature modules are integrated to increase the extraction capability of small targets.Attention mechanism is employed to improve the learning efficiency of key information and thus further enhance the efficiency of target detection.Improved non-maximum suppression improves the representation capability of the network.Our experimental results of the improved SSD-RTF algorithm on an independently-built power line data set show the accuracy and real-time detection of small targets improve to a certain extent,the overall mAP is up by 7.5％,and the wrong and missing detection decreases.关键词
目标检测/浅层融合/三叉乾特征融合/视觉机制/非极大值抑制Key words
target detection/shallow fusion/trigeminal stem feature fusion/visual mechanism/non-maximum suppression分类
信息技术与安全科学引用本文复制引用
唐心亮,李少杰,王建超,王震洲..一种SSD-RTF输电线路小目标检测算法研究[J].重庆理工大学学报,2024,38(17):150-157,8.基金项目
河北省高等学校科学技术研究项目(ZD2020318) （ZD2020318）
河北省教育厅青年基金项目(QN2023185) （QN2023185）
