基于改进Faster-RCNN的输电线巡检图像多目标检测及定位
|
摘要
针对输电线巡检图像受光线、环境和拍摄角度等因素影响,图像中的电气设备呈现低分辨率和多形态化特征的问题,提出一种基于改进Faster-RCNN的巡检图像多目标检测及定位方法。该方法首先通过区域建议策略网络生成若干目标候选区域;然后基于实际巡检图像样本库,对卷积神经网络进行训练,以改善参数学习效果;最后利用正则化方法优化参数权重,提高检测速度,得到适应巡检图像多形态化特征的改进型Faster-RCNN模型。实际场景数据集测试结果表明,相比于数字图像处理、浅层机器学习、单阶法、双阶法、Mask-RCNN和Local Loss目标检测方法,所提改进型Faster-RCNN能够在不同分辨率和不同位置角度的巡检图像场景下保持较高的识别精度和速度,具有较高的工程实用价值。
Because the transmission line inspection images are influenced by light,environment,shooting angles and so on,the electrical equipment in the image presents low resolution and polymorphic characteristics. Aiming at this problem,a multi-target detection and location of transmission line inspection image based on improved Faster-RCNN(Faster-Region-Convolutional Neural Network) is proposed. Firstly,a number of target candidate regions are gene-rated by using region proposal network. Then,the CNN(Convolution Neural Network) is trained based on the actual inspection image sample database to improve the parameter learning effect. Finally,the regularization method is used to optimize the parameter weights and improve the detection speed,then the improved Faster-RCNN model suitable for polymorphic features of inspection images is established. The test results of actual scene datasets show that compared with the digital image processing,shallow machine learning,single-stage method,double-stage method,Mask-RCNN and Local Loss target detection method,the proposed improved Faster-RCNN maintains a higher recognition accuracy and speed in inspection images of different resolution and angle,and it has high engineering value.
Because the transmission line inspection images are influenced by light,environment,shooting angles and so on,the electrical equipment in the image presents low resolution and polymorphic characteristics. Aiming at this problem,a multi-target detection and location of transmission line inspection image based on improved Faster-RCNN(Faster-Region-Convolutional Neural Network) is proposed. Firstly,a number of target candidate regions are gene-rated by using region proposal network. Then,the CNN(Convolution Neural Network) is trained based on the actual inspection image sample database to improve the parameter learning effect. Finally,the regularization method is used to optimize the parameter weights and improve the detection speed,then the improved Faster-RCNN model suitable for polymorphic features of inspection images is established. The test results of actual scene datasets show that compared with the digital image processing,shallow machine learning,single-stage method,double-stage method,Mask-RCNN and Local Loss target detection method,the proposed improved Faster-RCNN maintains a higher recognition accuracy and speed in inspection images of different resolution and angle,and it has high engineering value.
引文
[1] 李鸣.基于无人机的电力输电线路巡检监测系统研究[D].北京:北京交通大学,2017.LI Ming.Study on monitor system of power transmission line based on unmanned aerial vehicle[D].Beijing:Beijing Jiaotong University,2017.
[2] 周封,任贵新.基于颜色空间变量的输电线图像分类及特征提取[J].电力系统保护与控制,2018,46(5):89-98.ZHOU Feng,REN Guixin.Image classification and feature extraction of transmission line based on color space variable[J].Power System Protection and Control,2018,46(5):89-98.
[3] 孙舒琬.支持向量机算法研究及其在目标检测上的应用[D].济南:山东大学,2017.SUN Shuwan.SVM algorithm study and its application in object detection[J].Jinan:Shandong University,2017.
[4] 杨帆,王干军,彭小圣,等.基于卷积神经网络的高压电缆局部放电模式识别[J].电力自动化设备,2018,38(5):123-128.YANG Fan,WANG Ganjun,PENG Xiaosheng,et al.Partial discharge pattern recognition of high-voltage cables based on convolutional neural network[J].Electric Power Automation Equipment,2018,38(5):123-128.
[5] THOMPSON A.The cascading Haar wavelet algorithm for computing the Walsh-Hadamard transform[J].IEEE Signal Processing Letters,2016,24(99):1020-1023.
[6] 钱生,陈宗海,林名强,等.基于条件随机场和图像分割的显著性检测[J].自动化学报,2015,41(4):711-724.QIAN Sheng,CHEN Zonghai,LIN Mingqiang,et al.Saliency detection based on conditional random field and image segmentation[J].Acta Automatica Sinica,2015,41(4):711-724.
[7] JUNG H G.Analysis of reduced-set construction using image reconstruction from a HOG feature vector[J].IET Computer Vision,2017,11(8):725-732.
[8] 张泽浩.基于动态贝叶斯网络的无人机巡检输电线路故障诊断研究[D].北京:华北电力大学,2016.ZHANG Zehao.Research on fault diagnosis of transmission line based on dynamic Bayesian network[D].Beijing:North China Electric Power University,2016.
[9] 伍洋.基于机器学习的航拍图像绝缘子识别方法研究[D].北京:华北电力大学,2016.WU Yang.Research on insulator recognition methods in aerial images based on machine learning[D].Beijing:North China Electric Power University,2016.
[10] LECUN Y,BENGIO Y,HINTON G.Deep learning[J].Nature,2015,521(7553):436-444.
[11] HUANG J,GUADARRAMA S,MURPHY K,et al.Speed /accuracy trade-offs for modern convolutional object detectors[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Las Vegas,NV,USA:IEEE,2016:3296-3297.
[12] ROSS G,JEFF D,TREVOR D,et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Procee-dings of the IEEE Conference on Computer Vision and Pattern Recog-nition.Boston,MA,USA:IEEE,2015:705-713.
[13] ROSS G.Fast R-CNN[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Honolulu,HI,USA:IEEE,2017:6517-6525.
[14] 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.
[15] 王万国,田兵,刘越,等.基于RCNN的无人机巡检图像电力小部件识别研究[J].地球信息科学学报,2017(2):256-263.WANG Wanguo,TIAN Bing,LIU Yue,et al.Study on the electrical devices detection in UAV images based on region based convolu-tional neural networks[J].Journal of Geo-Information Science,2017(2):256-263.
[16] HE K,GKIOXARI G,DOLLAR P,et al.Mask R-CNN[C]//International Conference on Computer Vision.Venice,Veneto,Italy:IEEE,2017:2980-2988.
[17] REDMON J,DIVVALA S,ROSS G,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,NV,USA:IEEE,2016:779-788.
[18] LIN T Y,GOYAL P,ROSS G,et al.Focal loss for dense object detection[C]//International Conference on Computer Vision.Venice,Veneto,Italy:IEEE,2017:2999-3007.
[19] HE K,ZHANG X,REN S,et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[C]//European Conference on Computer Vision.Zurich,Switzerland:IEEE,2014:346-361.
[20] FELZENSZWALB P F,HUTTENLOCHER D P.Efficient graph-based image segmentation[J].International Journal of Computer Vision,2004,59(2):167-181.
[2] 周封,任贵新.基于颜色空间变量的输电线图像分类及特征提取[J].电力系统保护与控制,2018,46(5):89-98.ZHOU Feng,REN Guixin.Image classification and feature extraction of transmission line based on color space variable[J].Power System Protection and Control,2018,46(5):89-98.
[3] 孙舒琬.支持向量机算法研究及其在目标检测上的应用[D].济南:山东大学,2017.SUN Shuwan.SVM algorithm study and its application in object detection[J].Jinan:Shandong University,2017.
[4] 杨帆,王干军,彭小圣,等.基于卷积神经网络的高压电缆局部放电模式识别[J].电力自动化设备,2018,38(5):123-128.YANG Fan,WANG Ganjun,PENG Xiaosheng,et al.Partial discharge pattern recognition of high-voltage cables based on convolutional neural network[J].Electric Power Automation Equipment,2018,38(5):123-128.
[5] THOMPSON A.The cascading Haar wavelet algorithm for computing the Walsh-Hadamard transform[J].IEEE Signal Processing Letters,2016,24(99):1020-1023.
[6] 钱生,陈宗海,林名强,等.基于条件随机场和图像分割的显著性检测[J].自动化学报,2015,41(4):711-724.QIAN Sheng,CHEN Zonghai,LIN Mingqiang,et al.Saliency detection based on conditional random field and image segmentation[J].Acta Automatica Sinica,2015,41(4):711-724.
[7] JUNG H G.Analysis of reduced-set construction using image reconstruction from a HOG feature vector[J].IET Computer Vision,2017,11(8):725-732.
[8] 张泽浩.基于动态贝叶斯网络的无人机巡检输电线路故障诊断研究[D].北京:华北电力大学,2016.ZHANG Zehao.Research on fault diagnosis of transmission line based on dynamic Bayesian network[D].Beijing:North China Electric Power University,2016.
[9] 伍洋.基于机器学习的航拍图像绝缘子识别方法研究[D].北京:华北电力大学,2016.WU Yang.Research on insulator recognition methods in aerial images based on machine learning[D].Beijing:North China Electric Power University,2016.
[10] LECUN Y,BENGIO Y,HINTON G.Deep learning[J].Nature,2015,521(7553):436-444.
[11] HUANG J,GUADARRAMA S,MURPHY K,et al.Speed /accuracy trade-offs for modern convolutional object detectors[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Las Vegas,NV,USA:IEEE,2016:3296-3297.
[12] ROSS G,JEFF D,TREVOR D,et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Procee-dings of the IEEE Conference on Computer Vision and Pattern Recog-nition.Boston,MA,USA:IEEE,2015:705-713.
[13] ROSS G.Fast R-CNN[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Honolulu,HI,USA:IEEE,2017:6517-6525.
[14] 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.
[15] 王万国,田兵,刘越,等.基于RCNN的无人机巡检图像电力小部件识别研究[J].地球信息科学学报,2017(2):256-263.WANG Wanguo,TIAN Bing,LIU Yue,et al.Study on the electrical devices detection in UAV images based on region based convolu-tional neural networks[J].Journal of Geo-Information Science,2017(2):256-263.
[16] HE K,GKIOXARI G,DOLLAR P,et al.Mask R-CNN[C]//International Conference on Computer Vision.Venice,Veneto,Italy:IEEE,2017:2980-2988.
[17] REDMON J,DIVVALA S,ROSS G,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,NV,USA:IEEE,2016:779-788.
[18] LIN T Y,GOYAL P,ROSS G,et al.Focal loss for dense object detection[C]//International Conference on Computer Vision.Venice,Veneto,Italy:IEEE,2017:2999-3007.
[19] HE K,ZHANG X,REN S,et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[C]//European Conference on Computer Vision.Zurich,Switzerland:IEEE,2014:346-361.
[20] FELZENSZWALB P F,HUTTENLOCHER D P.Efficient graph-based image segmentation[J].International Journal of Computer Vision,2004,59(2):167-181.