摘 要:为解决数字图像中复杂多目标电力设备的分类识别与精确定位,提出了一种基于多特征融合与聚类分析的深度神经网络检测模型。该模型通过跳跃连接的信息通道快连边缘、角度及语义等多重特征,并通过阶段上采样融合不同尺度信息,以此构建出待检特征金字塔;然后对数据集进行聚类分析,利用金字塔中各尺度网络对原图像的映射比例计算出与目标最佳适配的区域生成框,最终得到适用于多尺度电力设备的检测模型。通过对比实验结果表明,本改进方法在检测时能够保持高精度并具有时效性,具备一定的工程实用价值。
关键词:深度学习;电力设备;多目标检测;特征融合;聚类分析;神经网络
DOI:10.19850/j.cnki.2096-4706.2023.03.016
中图分类号:TP391.4;TP18;TM761 文献标识码:A 文章编号:2096-4706(2023)03-0070-05
Identification and Localization of Multi-Scale Power Equipment Based on Fusion Characteristics and Cluster Analysis
MA Jingyi, YANG Jinlong
(Zhengzhou University of Science and Technology, Zhengzhou 450064, China)
Abstract: In order to solve the classification identification and precise positioning of complex multi-target power devices in digital images, deep neural network detection model based on multi-feature fusion and clustering analysis is proposed. In this model, the edge, angle and semantic features are connected quickly through the jump-connected information channel, and the different scale information is fused through the stage sampling to construct the Pyramid of features to be examined. Then, the data set is analyzed by clustering, and the region generating frame which best adapts to the target is calculated by using the mapping proportion of the original image to the scale network in the pyramid. Finally, the detection model for multi-scale power equipment is obtained. By comparing the experimental results, it shows that the improved method can maintain high accuracy and timeliness at the time of detection, and has certain engineering practical value.
Keywords: deep learning; power equipment; multi-object detection; feature fusion; cluster analysis; neural network
参考文献:
[1] 唐文虎,牛哲文,赵柏宁,等 . 数据驱动的人工智能技术在电力设备状态分析中的研究与应用 [J].高电压技术,2020,46(9):2985-2999.
[2] 周俊煌,黄廷城,谢小瑜,等 . 视频图像智能识别技术在输变电系统中的应用研究综述 [J]. 中国电力,2021,54(1):124-134+166.
[3] 林刚,王波,彭辉,等 . 基于改进 Faster-RCNN 的输电线巡检图像多目标检测及定位 [J]. 电力自动化设备,2019,39(5):213-218.
[4] 陈树勇,宋书芳,李兰欣,等 . 智能电网技术综述 [J]. 电网技术,2009,33(8):1-7.
[5] 章立 . 可见光图像弱小目标的检测与跟踪研究 [D]. 西安:西安科技大学,2018.
[6] 赵振兵,王乐 . 一种航拍绝缘子串图像自动定位方法 [J].仪器仪表学报,2014,35(3):558-565.
[7] 冯玲,黄新波,朱永灿 . 基于图像处理的输电线路覆冰厚度测量 [J]. 电力自动化设备,2011,31(10):76-80.
[8] LECUN Y,BENGIO Y,HINTON G. Deep Learning [J]. Nature,2015,521(28):436-444.
[9] KULKARNI A,CALLAN J. Selective Search [J].ACM Transactions on Information Systems (TOIS),2015,33(4):1-33.
[10] XIANG P,ZHOU H X,LI H. Hyperspectral anomaly Detection by Local Joint Subspace Process and Support Vector Machine [J]. International Journal of Remote Sensing,2020,41(10):3798-3819.
[11] GIRSHICK R. Fast R-CNN [C]//2015 IEEE International Conference on Computer Vision (ICCV).Santiago:IEEE,2015:1440-1448.
[12] REN S Q,HE K M,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.
[13] REDMON J,DIVVALA S,GIRSHICK R,et al. You only Look Once:Unified,Real-Time Object Detection [C]//2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Las Vegas:IEEE,2016:779-788.
[14] 赵振兵,李延旭,甄珍,等 . 结合 KL 散度和形状约束的Faster R-CNN 典型金具检测方法 [J]. 高电压技术,2020,46(9):3018-3026.
[15] 韩松臣,张比浩,李炜,等 . 基于改进 Faster-RCNN 的机场场面小目标物体检测算法 [J]. 南京航空航天大学学报,2019,51(6):735-741.
[16] HE K,ZHANG X Y,REN S Q,at al. Deep Residual Learning for Image Recognition [C]//2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Las Vegas: IEEE,2016:770-778.
[17] LIN T Y,DOLLÁR P,GIRSHICK R,et al. Feature Pyramid Networks for Object Detection [C]//2017 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Honolulu: IEEE,2017:936-944.
[18] 马秀麟,姚自明,邬彤,等 . 数据分析方法及应用——基于 SPSS 和 EXCEL 环境 [M]. 北京:人民邮电出版社,2015.
[19] NEUBECK A,GOOL L V. Efficient Non-Maximum Suppression [C]//18th International Conference on Pattern Recognition (ICPR'06).Hongkong:IEEE,2006:850-855.
[20] LONG J,SHELHAMER E,DARRELL T. Fully Convolutional Networks for Semantic Segmentation [C]//2015 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Boston:IEEE,2015:3431-3440
[21] KRIZHEVSKY A,SUTSKEVER I,HINTON G E. Imagenet Classification with Deep Convolutional Neural Networks [J]. Communications of the ACM,2017,60(6):84-90.
[22] 张建萍,刘希玉 . 基于聚类分析的 K-means 算法研究及应用 [J]. 计算机应用研究,2007(5):166-168.
[23] RUDER S. An Overview of Gradient Descent Optimization Algorithms [J/OL].arXiv:1609.04747 [cs.LG].[2022-09-03].https://arxiv. org/abs/1609.04747.
作者简介:马静怡(1995—),女,汉族,四川巴中人,助教,硕士,研究方向:目标检测;杨金龙(1990—),男,汉族,河南鹤壁人,助教,硕士,研究方向:精密仪器测量。