基于深度學(xué)習(xí)的絕緣子定向識(shí)別算法
doi: 10.11999/JEIT190350
-
山東理工大學(xué)建筑工程學(xué)院 淄博 255000
Insulator Orientation Detection Based on Deep Learning
-
Shandong University of Technology, Zibo 255000, China
-
摘要:
為了解決絕緣子目標(biāo)檢測(cè)中無(wú)法精確定位的問(wèn)題,該文基于深度學(xué)習(xí)提出一種絕緣子定向識(shí)別算法,通過(guò)在軸對(duì)齊檢測(cè)框中加入角度信息,可有效解決常規(guī)深度學(xué)習(xí)算法無(wú)法精確定位目標(biāo)的問(wèn)題。該算法首先將角度旋轉(zhuǎn)參數(shù)引入軸對(duì)齊矩形檢測(cè)框中構(gòu)成定向檢測(cè)框,然后將該參數(shù)偏移量作為第5參數(shù)加入到損失函數(shù)中進(jìn)行迭代回歸,同時(shí)為提高檢測(cè)精度在訓(xùn)練過(guò)程中使用Adam算法替代隨機(jī)梯度下降(SGD)算法進(jìn)行損失函數(shù)優(yōu)化,最終可獲得絕緣子定向檢測(cè)模型。實(shí)驗(yàn)分析表明,加入旋轉(zhuǎn)角度的定向檢測(cè)框可有效對(duì)絕緣子目標(biāo)進(jìn)行精確定位。
-
關(guān)鍵詞:
- 定向識(shí)別 /
- 絕緣子 /
- 深度學(xué)習(xí) /
- 角度旋轉(zhuǎn)
Abstract:In order to solve the problem of inaccurate location in insulator target detection, this paper proposes an insulator orientation recognition algorithm based on deep learning. By adding angle information to the axis alignment detection frame, it can effectively solve the problem that conventional deep learning algorithm can not accurately locate the target. First, the angular rotation parameters are introduced into the axially aligned rectangular detection frame to form a directional detection frame. Then the parameter offset is added as the fifth parameter to the loss function for iterative regression. At the same time, in order to improve the detection accuracy, Adam algorithm is used to replace Stochastic Gradient Descent (SGD) to optimize the loss function. Finally, the insulator directional detection model can be obtained. The experimental results show that the orientation detection frame with rotation angle can effectively locate the insulator target accurately.
-
Key words:
- Directional recognition /
- Insulator /
- Deep learning /
- Angle rotation
-
表 1 訓(xùn)練參數(shù)設(shè)定
參數(shù)名稱(chēng) 參數(shù)值 初始學(xué)習(xí)率 0.0001 學(xué)習(xí)率策略 Multistep 批處理大小 2 最大時(shí)期次數(shù) 100 每期迭代次數(shù) 1000 步長(zhǎng)值 60, 80, 100 下載: 導(dǎo)出CSV
表 2 方法AP對(duì)比
SSD模型(算法) 損失函數(shù)優(yōu)化方法 AP SSD300 SGD 0.561 SSD300 Adam 0.674 SSD512 SGD 0.736 SSD512 Adam 0.815 文獻(xiàn)[16]算法 – 0.761 下載: 導(dǎo)出CSV
-
BARTON J P and INFIELD D G. Energy storage and its use with intermittent renewable energy[J]. IEEE Transactions on Energy Conversion, 2004, 19(2): 441–448. doi: 10.1109/tec.2003.822305 WANG Jiafu, LIANG Xidong, and GAO Yanfeng. Failure analysis of decay-like fracture of composite insulator[J]. IEEE Transactions on Dielectrics and Electrical Insulation, 2014, 21(6): 2503–2511. doi: 10.1109/tdei.2014.004485 GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]. 2014 IEEE Conference on Computer Vision and Pattern Recogni-tion, Columbus, USA, 2014: 580–587. GIRSHICK R. Fast R-CNN[C]. 2015 IEEE International Conference on Computer Vision. Santiago, Chile, 2015: 1440–1448. doi: 10.1109/ICCV.2015.169. REN Shaoqing, HE Kaiming, GIRSHICK R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[C]. The 28th Interna-tional Conference on Neural Information Processing Systems, Montreal, Canada, 2015: 91–99. HE Kaiming, GKIOXARI G, DOLLáR P, et al. Mask R-CNN[C]. 2017 IEEE International Conference on Computer Vision, Venice, Italy, 2017: 2980–2988. doi: 10.1109/ICCV.2017.322. 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, Las Vegas, USA, 2016: 779–788. doi: 10.1109/CVPR.2016.91. REDMON J and FARHADI A. YOLO9000: Better, faster, stronger[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, USA, 2017: 6517–6525. REDMON J and FARHADI A. Yolov3: An incremental improvement[J]. arXiv:1804.02767, 2018. LIU Wei, ANGUELOV D, ERHAN D, et al. SSD: Single shot MultiBox detector[C]. The 14th European Conference on Computer Vision, Amsterdam, the Netherlands, 2016: 21–37. doi: 10.1007/978-3-319-46448-0_2. FU Chengyang, LIU Wei, RANGA A, et al. DSSD: Deconvolutional single shot detector[J]. arXiv:1701.06659, 2017. 陳慶, 閆斌, 葉潤(rùn), 等. 航拍絕緣子卷積神經(jīng)網(wǎng)絡(luò)檢測(cè)及自爆識(shí)別研究[J]. 電子測(cè)量與儀器學(xué)報(bào), 2017, 31(6): 942–953. doi: 10.13382/j.jemi.2017.06.018CHEN Qing, YAN Bin, YE Run, et al. Insulator detection and recognition of explosion fault based on convolutional neural networks[J]. Journal of Electronic Measurement and Instrumentation, 2017, 31(6): 942–953. doi: 10.13382/j.jemi.2017.06.018 LING Zenan, QIU R C, JIN Zhijian, et al. An accurate and real-time self-blast glass insulator location method based on faster R-CNN and U-net with aerial images[J]. arXiv:1801.05143, 2018. RONNEBERGER O, FISCHER P, and BROX T. U-net: Convolutional networks for biomedical image segmentation[C]. The 18th International Conference on Medical Image Computing and Computer-assisted Intervention, Munich, Germany, 2015: 234–241. ZHAO Zhenbing, FAN Xiaoqing, XU Guozhi, et al. Aggregating deep convolutional feature maps for insulator detection in infrared images[J]. IEEE Access, 2017, 5: 21831–21839. doi: 10.1109/ACCESS.2017.2757030 陳文賀, 李彩林, 袁斌, 等. 有效的絕緣子自爆缺陷定位方法[J]. 計(jì)算機(jī)工程與設(shè)計(jì), 2019, 40(8): 2346–2352. doi: 10.16208/j.issn1000-7024.2019.08.041CHEN Wenhe, LI Cailin, YUAN Bin, et al. Effective method to locate the self-explosion defect of insulators[J]. Computer Engineering and Design, 2019, 40(8): 2346–2352. doi: 10.16208/j.issn1000-7024.2019.08.041 黃新波, 劉新慧, 張燁, 等. 基于紅藍(lán)色差和改進(jìn)K-means算法的航拍絕緣子分類(lèi)識(shí)別方法[J]. 高電壓技術(shù), 2018, 44(5): 1528–1534. doi: 10.13336/j.1003-6520.hve.20180430018HUANG Xinbo, LIU Xinhui, ZHANG Ye, et al. Classi-fication recognition method of insulator in aerial image based on the red-blue difference and developed K-means algorithm[J]. High Voltage Engineering, 2018, 44(5): 1528–1534. doi: 10.13336/j.1003-6520.hve.20180430018 CHEN Chaoyue, GONG Weiguo, HU Yan, et al. Learning oriented region-based convolutional neural networks for building detection in satellite remote sensing images[C]. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Han-nover, Germany, 2017: 461–464. doi: 10.5194/isprs-archives-XLII-1-W1-461-2017. DENG Zhipeng, SUN Hao, ZHOU Shilin, et al. Toward fast and accurate vehicle detection in aerial images using coupled region-based convolutional neural networks[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2017, 10(8): 3652–3664. doi: 10.1109/JSTARS.2017.2694890 HE Tong, HUANG Weilin, QIAO Yu, et al. Accurate text localization in natural image with cascaded convolutional text network[J]. arXiv:1603.09423, 2016. YAO Cong, BAI Xiang, SANG Nong, et al. Scene text detection via holistic, multi-channel prediction[J]. arXiv:1606.09002, 2016. ZHANG Zheng, ZHANG Chengquan, SHEN Wei, et al. Multi-oriented text detection with fully convolutional networks[C]. 2016 IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, USA, 2016: 4159–4167. doi: 10.1109/CVPR.2016.451. LONG J, SHELHAMER E, DARRELL T. Fully convolu-tional networks for semantic segmentation[C]. 2015 IEEE Conference on Computer Vision and Pattern Recognition, Boston, USA, 2015: 3431–3440. doi: 10.1109/CVPR.2015.7298965. MA Jianqi, SHAO Weiyuan, Ye Hao, et al. Arbitrary-oriented scene text detection via rotation proposals[J]. IEEE Transactions on Multimedia, 2018, 20(11): 3111–3122. doi: 10.1109/TMM.2018.2818020 TANG Tianyu, ZHOU Shilin, DENG Zhipeng, et al. Arbitrary-oriented vehicle detection in aerial imagery with single convolutional neural networks[J]. Remote Sensing, 2017, 9(11): 1170–1186. doi: 10.3390/rs9111170 LIU Lei, PAN Zongxu, LEI Bin. Learning a rotation invariant detector with rotatable bounding box[J]. arXiv:1711.09405, 2017. ZHAO Zhenbing, LIU Ning, and WANG Le. Localization of multiple insulators by orientation angle detection and binary shape prior knowledge[J]. IEEE Transactions on Dielectrics and Electrical Insulation, 2015, 22(6): 3421–3428. doi: 10.1109/tdei.2015.004741 KINGMA D P and BA J. Adam: A method for stochastic optimization[J]. arXiv:1412.6980, 2014. DAI Jifeng, LI Yi, HE Kaiming, et al. R-FCN: Object detection via region-based fully convolutional net-works[C]. The 30th International Con-ference on Neural Information Processing systems, Bar-celona, Spain, 2016: 379–387. -