基于渐进特征融合及多尺度空洞注意力的遮挡鸟巢检测

doi:10.3785/j.issn.1008-973X.2025.03.011

浙江大学学报(工学版)

2025, Vol. 59

Issue (3): 535-545 DOI: 10.3785/j.issn.1008-973X.2025.03.011

计算机技术

基于渐进特征融合及多尺度空洞注意力的遮挡鸟巢检测

尹向雷(

),屈少鹏,解永芳,苏妮

陕西理工大学电气工程学院，陕西汉中 723000

Occluded bird nest detection based on asymptotic feature fusion and multi-scale dilated attention

Xianglei YIN(

),Shaopeng QU,Yongfang XIE,Ni SU

College of Electrical Engineering, Shaanxi University of Technology, Hanzhong 723000, China

全文: PDF(5885 KB) HTML

摘要：

为了提高被遮挡鸟巢目标的检测性能与准确性，减少鸟类筑巢对电力系统稳定运行造成的威胁以及运维成本，提出基于改进YOLOv5的输电线路鸟巢检测方法. 该方法使用渐进特征金字塔网络优化原始特征金字塔网络结构，有效避免了非相邻层次之间较大的语意差距，增强了非相邻层次间的融合效果. 使用多尺度空洞注意力机制，使模型能够有效地提取不同尺度的语义信息，提高模型对遮挡鸟巢目标的检测性能. 采用轻量级MobileNetV3网络作为骨干网络，进一步降低模型复杂度. 消融实验与定性实验结果表明，改进后算法的召回率、精确率与平均精度均值相较于原始算法分别提升了2.0个百分点、0.7个百分点与1.7个百分点，权重大小与计算量分别减少了74.7个百分点与53.5个百分点. 对于遮挡鸟巢目标均表现出良好的性能，验证了改进方法的有效性.

关键词： 输电线路; 遮挡目标; YOLOv5; 注意力机制; 渐进特征金字塔网络

Abstract:

An improved YOLOv5 transmission line bird nest detection method was proposed, in order to improve the detection performance and accuracy of the occluded bird nest targets, as well as reduce the threat of bird nesting to the stable operation of the power system and the operation and maintenance cost. Firstly, the asymptotic feature pyramid network was used to optimize the original feature pyramid network structure, effectively avoiding the large semantic gap between non-adjacent layers, and enhancing the fusion effect between non-adjacent layers. Secondly, the multi-scale dilated attention mechanism was used to enable the model to effectively extract semantic information at different scales and improve the detection performance of the model for occluded bird nest targets. Finally, the lightweight MobileNetV3 network was adopted as the backbone network to further reduce the complexity of the model. Ablation experiments and qualitative experimental analysis demonstrated that, the recall, precision and mean average precision of the improved algorithm were respectively improved by 2.0 percentage point, 0.7 percentage point and 1.7 percentage point compared with the original algorithm, and the weight and the computational amount were reduced by 74.7 percentage point and 53.5 percentage point, respectively. The results showed good performance for the occluded bird nest targets, which verified the effectiveness of the improved method.

Key words: transmission line occlusion target YOLOv5 attention mechanism asymptotic feature pyramid network

收稿日期: 2024-03-07 出版日期: 2025-03-10

CLC:

TP 391.4

基金资助: 国家自然科学基金资助项目（62176146）；陕西省教育厅重点科学研究计划资助项目（20JS018）；陕西理工大学人才启动专项资助项目（SLGRCQD2114）.

作者简介: 尹向雷（1977—），男，副教授，博士，从事目标检测与跟踪以及机器视觉在电力系统中应用的研究. orcid.org/0000-0001-9962-7986. E-mail：thunder@snut.edu.cn

	服务
	把本文推荐给朋友
	加入引用管理器
	E-mail Alert
	作者相关文章
	尹向雷
	屈少鹏
	解永芳
	苏妮

引用本文:

尹向雷,屈少鹏,解永芳,苏妮. 基于渐进特征融合及多尺度空洞注意力的遮挡鸟巢检测[J]. 浙江大学学报(工学版), 2025, 59(3): 535-545.

Xianglei YIN,Shaopeng QU,Yongfang XIE,Ni SU. Occluded bird nest detection based on asymptotic feature fusion and multi-scale dilated attention. Journal of ZheJiang University (Engineering Science), 2025, 59(3): 535-545.

链接本文:

https://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2025.03.011 或 https://www.zjujournals.com/eng/CN/Y2025/V59/I3/535

图 1 YOLOv5s网络结构图

图 2 MMA-YOLOv5算法框架

表 1 轻量化网络对比实验结果

图 3 轻量化网络平均精度均值对比实验结果

图 4 深度可分离卷积

图 5 标准卷积

图 6 SE通道注意力机制

图 7 多尺度空洞注意力

图 8 注意力图可视化对比

图 9 AFPN渐进架构

图 10 改进前后算法损失函数曲线图

图 11 改进前后算法性能对比

图 12 平均精度均值对比

表 2 消融实验结果

表 3 不同算法对比实验结果

图 13 被遮挡的测试图像示意图

图 14 不同算法的检测结果

1	陈杰, 朱仕焜, 孙嫱, 等面向无人机前端轻量级应用的输电线路鸟巢智能检测[J]. 福州大学学报: 自然科学版, 2023, 51 (4): 539- 546 CHEN Jie, ZHU Shikun, SUN Qiang, et al Bird’s nest intelligent detection on transmission lines for unmanned aerial vehicle front-end lightweight application[J]. Journal of Fuzhou University: Natural Science Edition, 2023, 51 (4): 539- 546
2	WEN X, WU Q, WANG Y, et al High-risk region of bird streamer flashover in 110 kV composite insulators and design for bird-preventing shield[J]. International Journal of Electrical Power and Energy Systems, 2021, 131: 107010 doi: 10.1016/j.ijepes.2021.107010
3	裴少通, 张行远, 胡晨龙, 等基于ER-YOLO算法的跨环境输电线路缺陷识别方法[J]. 电工技术学报, 2024, 39 (9): 2825- 2840 PEI Shaotong, ZHANG Hangyuan, HU Chenlong, et al The defect detection method for cross-environment power transmission line based on the ER-YOLO algorithm[J]. Transactions of China Electrotechnical Society, 2024, 39 (9): 2825- 2840
4	何冠锋. 基于深度学习的输电线路鸟巢检测研究[D]. 东莞: 东莞理工学院, 2023. HE Guanfeng. Research on bird nest detection of transmission lines based on deep learning [D]. Dongguan: Dongguan University of Technology, 2023.
5	徐晶, 韩军, 童志刚, 等一种无人机图像的铁塔上鸟巢检测方法[J]. 计算机工程与应用, 2017, 53 (6): 231- 235 XU Jing, HAN Jun, TONG Zhigang, et al Method for detecting bird’s nest on tower based on UAV image[J]. Computer Engineering and Applications, 2017, 53 (6): 231- 235
6	LU J, XU X, LI X, et al Detection of bird’s nest in high power lines in the vicinity of remote campus based on combination features and cascade classifier[J]. IEEE Access, 2018, 6: 39063- 39071 doi: 10.1109/ACCESS.2018.2851588
7	李浩然, 高健, 吴田, 等基于改进Canny算子的绝缘子裂纹检测研究[J]. 智慧电力, 2021, 49 (2): 91- 98 LI Haoran, GAO Jian, WU Tian, et al Crack detection method of insulators based on improved canny operator[J]. Smart Power, 2021, 49 (2): 91- 98
8	LECUN Y, BENGIO Y, HINTON G Deep learning[J]. Nature, 2015, 521 (7553): 436- 444
9	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . Las Vegas: IEEE, 2016: 779–788.
10	LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multiBox detector [C]// European Conference on Computer Vision . [s.l.]: Springer, 2016: 21–37.
11	LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection [C]// Proceedings of the IEEE International Conference on Computer Vision . Venice: IEEE, 2017: 2999–3007.
12	GIRSHICK R. Fast R-CNN [C]// Proceedings of the IEEE International Conference on Computer Vision . Santiago: IEEE, 2015, 7: 1440–1448.
13	REN S, HE K, GIRSHICK R, et al Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39 (6): 1137- 1149 doi: 10.1109/TPAMI.2016.2577031
14	HE K, GKIOXARI G, DOLLÁR P, et al. Mask R-CNN [C]// Proceedings of the IEEE International Conference on Computer Vision . Venice: IEEE, 2017: 2980–2988.
15	魏业文, 李梅, 解园琳, 等基于改进Faster-RCNN的输电线路巡检图像检测[J]. 电力工程技术, 2022, 41 (2): 171- 178 WEI Yewen, LI Mei, XIE Yuanlin, et al Transmission line inspection image detection based on improved Faster-RCNN[J]. Electric Power Engineering Technology, 2022, 41 (2): 171- 178
16	王纪武, 罗海保, 鱼鹏飞, 等基于Faster R-CNN的多尺度高压塔鸟巢检测[J]. 北京交通大学学报, 2019, 43 (5): 37- 43 WANG Jiwu, LUO Haibao, YU Pengfei, et al Bird’s nest detection in multi-scale of high-voltage tower based on Faster R-CNN[J]. Journal of Beijing Jiaotong University, 2019, 43 (5): 37- 43
17	杨学存, 和沛栋, 陈丽媛, 等基于深度可分离卷积的轻量级YOLOv3输电线路鸟巢检测方法[J]. 智慧电力, 2021, 49 (12): 88- 95 YANG Xuecun, HE Peidong, CHEN Liyuan, et al Bird’s nest detection on lightweight YOLOv3 transmission line based on deep separable convolution[J]. Smart Power, 2021, 49 (12): 88- 95
18	赵霖, 王素珍, 邵明伟, 等基于改进YOLOv5的输电线路鸟巢缺陷检测方法[J]. 电子测量技术, 2023, 46 (3): 157- 165 ZHAO Lin, WANG Suzhen, SHAO Mingwe, et al Improved YOLOv5-based bird’s nest defect detection method for transmission lines[J]. Electronic Measurement Technology, 2023, 46 (3): 157- 165
19	王杨杨, 曹晖, 莫文昊基于深度学习的改进型YOLOv4输电线路鸟巢检测与识别[J]. 智慧电力, 2023, 51 (1): 101- 107 WANG Yangyang, CAO Hui, MO Wenhao Bird’s nest det1ection and identification on improved YOLOv4 transmission line based on deep learning[J]. Smart Power, 2023, 51 (1): 101- 107
20	ZHANG Z, HE G Recognition of bird nests on power transmission lines in aerial images based on improved YOLOv4[J]. Frontiers in Energy Research, 2022, 10: 870253 doi: 10.3389/fenrg.2022.870253
21	HAN G, LI T, LI Q, et al Improved algorithm for insulator and its defect detection based on YOLOX[J]. Sensors, 2022, 22 (16): 6186 doi: 10.3390/s22166186
22	肖粲俊, 潘睿志, 李超, 等基于改进YOLOv5s绝缘子缺陷检测技术研究[J]. 电子测量技术, 2022, 45 (24): 137- 144 XIAO Canjun, PAN Ruizhi, LI Chao, et al Research on defect detection technology based on improved YOLOv5s insulator[J]. Electronic Measurement Technology, 2022, 45 (24): 137- 144
23	张焕龙, 齐企业, 张杰, 等基于改进YOLOv5的输电线路鸟巢检测方法研究[J]. 电力系统保护与控制, 2023, 51 (2): 151- 159 ZHANG Huanlong, QI Qiye, ZHANG Jie, et al Bird nest detection method for transmission lines based on improved YOLOv5[J]. Power System Protection and Control, 2023, 51 (2): 151- 159
24	郝琨, 王阔, 王贝贝基于改进Mobilenet-YOLOv3的轻量级水下生物检测算法[J]. 浙江大学学报: 工学版, 2022, (8): 1622- 1632 HAO Kun, WANG Kuo, WANG Beibei Lightweight underwater biological detection algorithm based on improved Mobilenet-YOLOv3[J]. Journal of Zhejiang University: Engineering Science, 2022, (8): 1622- 1632
25	袁天乐, 袁巨龙, 朱勇建, 等基于改进YOLOv5的推力球轴承表面缺陷检测算法[J]. 浙江大学学报: 工学版, 2022, (12): 2349- 2357 YUAN Tianle, YUAN Julong, ZHU Yongjian, et al Surface defect detection algorithm of thrust ball bearing based on improved YOLOv5[J]. Journal of Zhejiang University: Engineering Science, 2022, (12): 2349- 2357
26	LI J, YAN D, LUAN K, et al Deep learning-based bird’s nest detection on transmission lines using UAV imagery[J]. Applied Sciences, 2020, 10 (18): 6147 doi: 10.3390/app10186147

[1]	刘登峰,郭文静,陈世海. 基于内容引导注意力的车道线检测网络[J]. 浙江大学学报(工学版), 2025, 59(3): 451-459.
[2]	姚明辉,王悦燕,吴启亮,牛燕,王聪. 基于小样本人体运动行为识别的孪生网络算法[J]. 浙江大学学报(工学版), 2025, 59(3): 504-511.
[3]	薛雅丽,贺怡铭,崔闪,欧阳权. 基于改进YOLOv5的SAR图像有向舰船目标检测算法[J]. 浙江大学学报(工学版), 2025, 59(2): 261-268.
[4]	杨冰,徐楚阳,姚金良,向学勤. 基于单目RGB图像的三维手部姿态估计方法[J]. 浙江大学学报(工学版), 2025, 59(1): 18-26.
[5]	冉庆东,郑力新. 基于改进YOLOv5的锂电池极片缺陷检测方法[J]. 浙江大学学报(工学版), 2024, 58(9): 1811-1821.
[6]	张凌嘉,周欣磊,许月萍,江衍铭. 基于YOLOv5和Mask-RCNN组合模型的社交媒体内涝灾害分析[J]. 浙江大学学报(工学版), 2024, 58(9): 1822-1831.
[7]	李灿林,王新玥,马利庄,邵志文,张文娇. 融合注意力机制和结构线提取的图像卡通化[J]. 浙江大学学报(工学版), 2024, 58(8): 1728-1737.
[8]	李忠良,陈麒,石琳,杨朝,邹先明. 时间感知组合的动态知识图谱补全[J]. 浙江大学学报(工学版), 2024, 58(8): 1738-1747.
[9]	吴书晗,王丹,陈远方,贾子钰,张越棋,许萌. 融合注意力的滤波器组双视图图卷积运动想象脑电分类[J]. 浙江大学学报(工学版), 2024, 58(7): 1326-1335.
[10]	马现伟,范朝辉,聂为之,李东,朱逸群. 对失效传感器具备鲁棒性的故障诊断方法[J]. 浙江大学学报(工学版), 2024, 58(7): 1488-1497.
[11]	杨军,张琛. 基于边界点估计与稀疏卷积神经网络的三维点云语义分割[J]. 浙江大学学报(工学版), 2024, 58(6): 1121-1132.
[12]	李运堂,李恒杰,张坤,王斌锐,关山越,陈源. 基于新型编码解码网络的复杂输电线识别[J]. 浙江大学学报(工学版), 2024, 58(6): 1133-1141.
[13]	魏翠婷,赵唯坚,孙博超,刘芸怡. 基于改进Mask R-CNN与双目视觉的智能配筋检测[J]. 浙江大学学报(工学版), 2024, 58(5): 1009-1019.
[14]	刘毅,陈一丹,高琳,洪姣. 基于多尺度特征融合的轻量化道路提取模型[J]. 浙江大学学报(工学版), 2024, 58(5): 951-959.
[15]	邢志伟,朱书杰,李彪. 基于改进图卷积神经网络的航空行李特征感知[J]. 浙江大学学报(工学版), 2024, 58(5): 941-950.

Viewed

Full text

Abstract

Cited

Shared

Discussed