卷积神经网络及其在目标检测中的应用
|
摘要
针对目前武器装备在检测空中远距离弱小目标、假目标、遮挡等情况中智能化程度不高问题,分析了卷积神经网络的工作方式以及其应用在目标检测中的优势,讨论了基于卷积神经网络的目标检测算法在其它图像检测领域的应用情况及取得的最新成果,通过研究发现卷积神经网络利用其强大的特征学习能力使得检测过程更为高效化、智能化,将其应用到导弹武器系统中是未来提升防空作战效能的必然手段。
In view of the fact that the current weaponry is not intelligent in detecting long-range and weak targets,false targets,and occlusions in the air. The working mode of convolutional neural network and the advantages of its application in target detection are analyzed. The application and the latest achievements of the target detection algorithm based on convolutional neural network in other image detection fields are discussed. Convolutional neural networks use their powerful feature learning capabilities to make the detection process more efficient and intelligent is discovered through research. Its application to the missile weapon system is the inevitable means to improve the effectiveness of air defense operations in the future.
In view of the fact that the current weaponry is not intelligent in detecting long-range and weak targets,false targets,and occlusions in the air. The working mode of convolutional neural network and the advantages of its application in target detection are analyzed. The application and the latest achievements of the target detection algorithm based on convolutional neural network in other image detection fields are discussed. Convolutional neural networks use their powerful feature learning capabilities to make the detection process more efficient and intelligent is discovered through research. Its application to the missile weapon system is the inevitable means to improve the effectiveness of air defense operations in the future.
引文
[1] Duffy K R,Hubel D H. Receptive field properties of neu-rons in the primary visual cortex under photopic and sco-topic lighting conditions[J]. Vision Research,2007,47(19):2569-2574.
[2] Fukushima K,Miyake S,Ito T. Neocognitron:A neuralnetwork model for a mechanism of visual pattern recogni-tion[J]. IEEE Transactions on Systems,Man,and Cy-bernetics,1983,(5):826-834.
[3] Le Cun Y,Boser B,Denker J S,et al. Backpropagationapplied to handwritten zip code recognition[J]. Neural-Computation,1989,1(4):541-551.
[4] Hinton G E,Osindero S,Teh Y W. A fast learning algo-rithmfor deep belief nets[J]. Neural Computation,2006,18(7):1527-1554.
[5]卢宏涛,张秦川.深度卷积神经网络在计算机视觉中的应用研究综述[J].数据采集与处理,2016,31(01):1-17.
[6] Ren S Q,He K M,Girshick R,et al. Faster R-CNN:Towards real-time object detection with region proposalnetworks[C]. Proceedings of Advances in Neural Infor-mation Processing Systems,2015:91-99.
[7] Li Y,He K,Sun J. R-FCN:Object detection via region-based fully convolutional networks[C]. Advances in Neu-ral Information Processing Systems,2016:379-387.
[8] Liu W,Anguelov D,Erhan D,et al. SSD:Single shot-multibox detector[C]. European conference on computervision. Springer,Cham,2016:21-37.
[9] Redmon J,Divvala S K,Girshick R B,et al. You onlylook once:Unified,real-time object detection[C]. In:CVPR(2016).
[10]郑昌艳,梅卫,王刚.基于深度卷积神经网络的蛇形机动航迹图像识别[J].火力与指挥控制,2016,41(5):66-70.
[11]冯小雨,梅卫,胡大帅.基于改进Faster R-CNN的空中目标检测[J].光学学报,2018(6):250-258.
[12] Krizhevsky A,Sutskever I,Hinton G E. Imagenet clas-sification with deep convolutional neural networks[C].Advances in Neural Information Processing Systems,2012:1097-1105.
[13] Simonyan K,Zisserman A. Very deep convolutional net-works for large-scale image recognition[C]. Internation-al Conference on Learning Representations(ICLR),2015.
[14] He K,Zhang X,Ren S,et al. Deep residual learning forimage recognition[J]. Computer Vision and PatternRecognition,2016.
[15] Szegedy C,Liu W,Jia Y,et al. Going deeper with con-volutions[C]. Proceedings of the IEEE Conference onComputer Vision and Pattern Recognition,2015:1-9.
[16] Chao Y,Vijayanarasimhan S,et al. Rethinking the fas-ter R-CNN architecture for temporal action localization.2018-1-6. http://arxiv. org/abs/1804. 07667,2018.
[17] He K,Zhang X,Ren S,et al. Spatial pyramid poolingin deep convolutional networks for visual recognition[C]. Proceedings of European Conference on ComputerVision,2014:346-361.
[18] He K,Zhang X,Ren S,et al. Delving deep into rectifi-ers:Surpassing human-level performance on imagenetclassification[EB/OL]. http://arxiv. org/abs/1502.03167,2015.
[19] Girshick,Ross,et al. Rich feature hierarchies for accu-rate object detection and semantic segmentation[C].Proceedings of the IEEE conference on computer visionand pattern recognition,2014.
[20] Girshick,Ross. Fast r-cnn[C]. Proceedings of theIEEE International Conference on Computer Vision.2015.
[21] Uijlings J R R,Sande K E A,Gevers A W M. Selectivesearch for object recognition[J]. Smeulders. Interna-tional Journal of Computer Vision,2013(2).
[22] Zitnick C L,Dollár P. Edge boxes:locating object pro-posals from edges[C]. Proceedings of the 13th EuropeanConference on Computer Vision-ECCV 2014. Switzer-land:Springer International Publishing, 2014:391-405.
[23] Bell S,Zitnick C L,Bala K,et al. Inside-outside net:Detecting objects in context with skip pooling and recur-rent neural networks[C]. In:CVPR.(2016).
[24] Kong,Tao,et al. Hyper Net:towards accurate regionproposal generation and joint object detection[C]. ar Xivpreprint ar Xiv:1604. 00600(2016).
[25] Felzenszwalb P,Girshick R,McAllester D,et al. Ob-ject detection with discriminatively trained part basedmodels[C]. TPAMI,2010.
[26] Redmon J,Farhadi A. Yolo9000:Better,faster,stron-ger[C]. In Computer Vision and Pattern Recognition(CVPR),2017.
[27] Redmon J,Farhadi A. Yolov3:An incremental im-provement[C]. ar Xiv preprint ar Xiv,2018.
[28] Ioffe S,Szegedy C. Batch normalization:Acceleratingdeep network training by reducing internal covariateshift[J]. ar Xiv preprint ar Xiv:1502. 03167,2015.
[2] Fukushima K,Miyake S,Ito T. Neocognitron:A neuralnetwork model for a mechanism of visual pattern recogni-tion[J]. IEEE Transactions on Systems,Man,and Cy-bernetics,1983,(5):826-834.
[3] Le Cun Y,Boser B,Denker J S,et al. Backpropagationapplied to handwritten zip code recognition[J]. Neural-Computation,1989,1(4):541-551.
[4] Hinton G E,Osindero S,Teh Y W. A fast learning algo-rithmfor deep belief nets[J]. Neural Computation,2006,18(7):1527-1554.
[5]卢宏涛,张秦川.深度卷积神经网络在计算机视觉中的应用研究综述[J].数据采集与处理,2016,31(01):1-17.
[6] Ren S Q,He K M,Girshick R,et al. Faster R-CNN:Towards real-time object detection with region proposalnetworks[C]. Proceedings of Advances in Neural Infor-mation Processing Systems,2015:91-99.
[7] Li Y,He K,Sun J. R-FCN:Object detection via region-based fully convolutional networks[C]. Advances in Neu-ral Information Processing Systems,2016:379-387.
[8] Liu W,Anguelov D,Erhan D,et al. SSD:Single shot-multibox detector[C]. European conference on computervision. Springer,Cham,2016:21-37.
[9] Redmon J,Divvala S K,Girshick R B,et al. You onlylook once:Unified,real-time object detection[C]. In:CVPR(2016).
[10]郑昌艳,梅卫,王刚.基于深度卷积神经网络的蛇形机动航迹图像识别[J].火力与指挥控制,2016,41(5):66-70.
[11]冯小雨,梅卫,胡大帅.基于改进Faster R-CNN的空中目标检测[J].光学学报,2018(6):250-258.
[12] Krizhevsky A,Sutskever I,Hinton G E. Imagenet clas-sification with deep convolutional neural networks[C].Advances in Neural Information Processing Systems,2012:1097-1105.
[13] Simonyan K,Zisserman A. Very deep convolutional net-works for large-scale image recognition[C]. Internation-al Conference on Learning Representations(ICLR),2015.
[14] He K,Zhang X,Ren S,et al. Deep residual learning forimage recognition[J]. Computer Vision and PatternRecognition,2016.
[15] Szegedy C,Liu W,Jia Y,et al. Going deeper with con-volutions[C]. Proceedings of the IEEE Conference onComputer Vision and Pattern Recognition,2015:1-9.
[16] Chao Y,Vijayanarasimhan S,et al. Rethinking the fas-ter R-CNN architecture for temporal action localization.2018-1-6. http://arxiv. org/abs/1804. 07667,2018.
[17] He K,Zhang X,Ren S,et al. Spatial pyramid poolingin deep convolutional networks for visual recognition[C]. Proceedings of European Conference on ComputerVision,2014:346-361.
[18] He K,Zhang X,Ren S,et al. Delving deep into rectifi-ers:Surpassing human-level performance on imagenetclassification[EB/OL]. http://arxiv. org/abs/1502.03167,2015.
[19] Girshick,Ross,et al. Rich feature hierarchies for accu-rate object detection and semantic segmentation[C].Proceedings of the IEEE conference on computer visionand pattern recognition,2014.
[20] Girshick,Ross. Fast r-cnn[C]. Proceedings of theIEEE International Conference on Computer Vision.2015.
[21] Uijlings J R R,Sande K E A,Gevers A W M. Selectivesearch for object recognition[J]. Smeulders. Interna-tional Journal of Computer Vision,2013(2).
[22] Zitnick C L,Dollár P. Edge boxes:locating object pro-posals from edges[C]. Proceedings of the 13th EuropeanConference on Computer Vision-ECCV 2014. Switzer-land:Springer International Publishing, 2014:391-405.
[23] Bell S,Zitnick C L,Bala K,et al. Inside-outside net:Detecting objects in context with skip pooling and recur-rent neural networks[C]. In:CVPR.(2016).
[24] Kong,Tao,et al. Hyper Net:towards accurate regionproposal generation and joint object detection[C]. ar Xivpreprint ar Xiv:1604. 00600(2016).
[25] Felzenszwalb P,Girshick R,McAllester D,et al. Ob-ject detection with discriminatively trained part basedmodels[C]. TPAMI,2010.
[26] Redmon J,Farhadi A. Yolo9000:Better,faster,stron-ger[C]. In Computer Vision and Pattern Recognition(CVPR),2017.
[27] Redmon J,Farhadi A. Yolov3:An incremental im-provement[C]. ar Xiv preprint ar Xiv,2018.
[28] Ioffe S,Szegedy C. Batch normalization:Acceleratingdeep network training by reducing internal covariateshift[J]. ar Xiv preprint ar Xiv:1502. 03167,2015.