基于深度神经网络的迷彩目标发现仿真学习方法
|
摘要
针对自动化迷彩目标发现学习中有效样本严重不足的问题,借鉴AlphaGo的技术思想,提出了一种基于样本模拟的深度神经网络仿真训练方法。建立了迷彩场景仿真合成模型,通过设计图像空间的复合算法、场景图像深度特征提取策略、目标融合度测量策略,以及图聚类采样算法,批量化地生成了可用于深度神经网络训练和学习的具有代表性的迷彩场景仿真样本;设计了基于深度残差神经网络的迷彩目标发现模型,并引入了多尺度网络训练方法。模拟样本和真实场景图像的实验结果表明,所提方法可有效应用于迷彩目标的自动化识别与评估。
Aiming at the problem of serious lack of effective samples in the automatic discovery of camouflage targets,a simulation training method is proposed based on the sample simulation of a deep neural network and the technical idea of AlphaGo.A simulation synthesis model of camouflage scenes is established.The compound algorithm in the image space,the deep feature extraction strategy of scene images,the measurement strategy of target fusion degree,and the sampling algorithm for graph clustering are designed,respectively.Thus the representative samples for camouflage scene simulation are batch generated,which can be used for the deep neural network training and learning.Moreover,a discovery model of camouflage targets is designed based on a deep residual neural network,in which a multi-scale network training strategy is considered.The experimental results on the simulated samples and real scene images show that the proposed method can be effectively used for the automatic discovery and evaluation of camouflage targets.
Aiming at the problem of serious lack of effective samples in the automatic discovery of camouflage targets,a simulation training method is proposed based on the sample simulation of a deep neural network and the technical idea of AlphaGo.A simulation synthesis model of camouflage scenes is established.The compound algorithm in the image space,the deep feature extraction strategy of scene images,the measurement strategy of target fusion degree,and the sampling algorithm for graph clustering are designed,respectively.Thus the representative samples for camouflage scene simulation are batch generated,which can be used for the deep neural network training and learning.Moreover,a discovery model of camouflage targets is designed based on a deep residual neural network,in which a multi-scale network training strategy is considered.The experimental results on the simulated samples and real scene images show that the proposed method can be effectively used for the automatic discovery and evaluation of camouflage targets.
引文
[1]Zhang H R,Li Y B,Xing R K,et al.Evaluation of air defense missile infrared camouflage capability based on set pair analysis[J].Laser&Optoelectronics Progress,2018,55(7):070402.张海瑞,李彦彬,邢瑞康,等.基于集对分析的防空导弹装备红外伪装能力评估[J].激光与光电子学进展,2018,55(7):070402.
[2]Guo T,Hua W S,Liu X,et al.Comprehensive evaluation of optical camouflage effect based on hyperspectra[J].Laser&Optoelectronics Progress,2016,53(10):101002.郭彤,华文深,刘恂,等.一种基于高光谱的光学伪装效果综合评价方法[J].激光与光电子学进展,2016,53(10):101002.
[3]Bai X Q,Liao N F,Huang H,et al.Evaluation of ship camouflage effect on sea based on color difference and spectral characteristics[J].Laser&Optoelectronics Progress,2018,55(9):093301.白雪琼,廖宁放,黄浩,等.基于色差和光谱特性的海面船只隐身效果评估[J].激光与光电子学进展,2018,55(9):093301.
[4]Cai W,Wu F C,Yang Z Y,et al.Research on magneto-optic modulation technology and application[J].Laser&Optoelectronics Progress,2015,52(6):060003.蔡伟,伍樊成,杨志勇,等.磁光调制技术与应用研究[J].激光与光电子学进展,2015,52(6):060003.
[5]Wang T,Niu M S,Bu M M,et al.Polarizationdifference imaging system with adjustable optical path and its characteristics[J].Acta Optica Sinica,2017,37(7):0711001.王田,牛明生,步苗苗,等.可调光程的差分偏振成像系统及其特性研究[J].光学学报,2017,37(7):0711001.
[6]Tao F,Song M X,Hong J,et al.Polarization calibration method for simultaneous imaging polarimeter based on off-axis three-mirror[J].Acta Optica Sinica,2018,38(9):0912005.陶菲,宋茂新,洪津,等.基于离轴三反的同时全偏振成像仪的偏振定标方法[J].光学学报,2018,38(9):0912005.
[7]Wang X L,Wang F,Liu X,et al.Hyperspectral polarization characteristics of typical camouflage target under desert background[J].Laser&Optoelectronics Progress,2018,55(5):051101.王小龙,王峰,刘晓,等.荒漠背景下典型伪装目标的高光谱偏振特性[J].激光与光电子学进展,2018,55(5):051101.
[8]Cui B S,Xue S Q,Ji Y J,et al.Camouflage effectiveness evaluation based on image feature[J].Infrared and Laser Engineering,2010,39(6):1178-1183.崔宝生,薛士强,姬艳军,等.图像特征的伪装效果评估技术[J].红外与激光工程,2010,39(6):1178-1183.
[9]Xu W D,LüX L,Chen B,et al.A model based on texture analysis for the performance evaluation of camouflage screen equipment[J].Acta Armamentarii,2002,23(3):329-331.许卫东,吕绪良,陈兵,等.一种基于纹理分析的伪装器材效果评价模型[J].兵工学报,2002,23(3):329-331.
[10]Lin W,Chen Y H,Wang J Y,et al.Camouflage assessment method based on image features and psychological perception quantity[J].Acta Armamentarii,2013,34(4):412-417.林伟,陈玉华,王吉远,等.基于图像特征与心理感知量的伪装效果评价方法[J].兵工学报,2013,34(4):412-417.
[11]Wang P Y,Zhao D H,Li M F.Optical camouflage effect assessment based on digital image inpainting technology[J].Laser&Optoelectronics Progress,2018,55(3):031011.王鹏烨,赵德辉,李明锋.基于图像修复技术的目标可见光伪装效果评价[J].激光与光电子学进展,2018,55(3):031011.
[12]Zhang R,Li W P,Mo T.Review of deep learning[J].Information and Control,2018,47(4):385-397,410.张荣,李伟平,莫同.深度学习研究综述[J].信息与控制,2018,47(4):385-397,410.
[13]Krizhevsky A,Sutskever I,Hinton G E.ImageNet classification with deep convolutional neural networks[J].Communications of the ACM,2017,60(6):84-90.
[14]He K M,Zhang X Y,Ren S Q,et al.Deep residual learning for image recognition[C]∥Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,2016:770-778.
[15]Huang G,Liu Z,Maaten L V D,et al.Densely connected convolutional networks[C]∥Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,2017:2261-2269.
[16]Xin Y,Yang J,Xie Z Q.A semantic overlapping community detecting algorithm in social networks based on random walk[J].Journal of Computer Research and Development,2015,52(2):499-511.辛宇,杨静,谢志强.基于随机游走的语义重叠社区发现算法[J].计算机研究与发展,2015,52(2):499-511.
[17]Wang Y,Tang J,Rao Q F,et al.High efficient K-means algorithm for determining optimal number of clusters[J].Journal of Computer Applications,2014,34(5):1331-1335.王勇,唐靖,饶勤菲,等.高效率的K-means最佳聚类数确定算法[J].计算机应用,2014,34(5):1331-1335.
[18]An Z,Xu X P,Yang J H,et al.Design of augmented reality head-up display system based on image semantic segmentation[J].Acta Optica Sinica,2018,38(7):0710004.安喆,徐熙平,杨进华,等.结合图像语义分割的增强现实型平视显示系统设计与研究[J].光学学报,2018,38(7):0710004.
[19]Li S M,Lei G Q,Fan R.Depth map superresolution based on two-channel convolutional neural network[J].Acta Optica Sinica,2018,38(10):1010002.李素梅,雷国庆,范如.基于双通道卷积神经网络的深度图超分辨研究[J].光学学报,2018,38(10):1010002.
[20]Sun H Q,Pang Y W.An neural network framework of self-learning uncertainty[J].Acta Optica Sinica,2018,38(6):0620002.孙汉卿,庞彦伟.一种自学习不确定度的神经网络架构[J].光学学报,2018,38(6):0620002.
[21]Long J,Shelhamer E,Darrell T.Fully convolutional networks for semantic segmentation[C]∥Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,2015:3431-3440.
[22]Isola P,Zhu J Y,Zhou T H,et al.Image-to-image translation with conditional adversarial networks[C]∥Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,2017:5967-5976.
[23]Goodfellow I J,Pouget-Abadie J,Mirza M,et al.Generative adversarial nets[C]∥Advances in Neural Information Processing Systems,2014:2672-2680.
[2]Guo T,Hua W S,Liu X,et al.Comprehensive evaluation of optical camouflage effect based on hyperspectra[J].Laser&Optoelectronics Progress,2016,53(10):101002.郭彤,华文深,刘恂,等.一种基于高光谱的光学伪装效果综合评价方法[J].激光与光电子学进展,2016,53(10):101002.
[3]Bai X Q,Liao N F,Huang H,et al.Evaluation of ship camouflage effect on sea based on color difference and spectral characteristics[J].Laser&Optoelectronics Progress,2018,55(9):093301.白雪琼,廖宁放,黄浩,等.基于色差和光谱特性的海面船只隐身效果评估[J].激光与光电子学进展,2018,55(9):093301.
[4]Cai W,Wu F C,Yang Z Y,et al.Research on magneto-optic modulation technology and application[J].Laser&Optoelectronics Progress,2015,52(6):060003.蔡伟,伍樊成,杨志勇,等.磁光调制技术与应用研究[J].激光与光电子学进展,2015,52(6):060003.
[5]Wang T,Niu M S,Bu M M,et al.Polarizationdifference imaging system with adjustable optical path and its characteristics[J].Acta Optica Sinica,2017,37(7):0711001.王田,牛明生,步苗苗,等.可调光程的差分偏振成像系统及其特性研究[J].光学学报,2017,37(7):0711001.
[6]Tao F,Song M X,Hong J,et al.Polarization calibration method for simultaneous imaging polarimeter based on off-axis three-mirror[J].Acta Optica Sinica,2018,38(9):0912005.陶菲,宋茂新,洪津,等.基于离轴三反的同时全偏振成像仪的偏振定标方法[J].光学学报,2018,38(9):0912005.
[7]Wang X L,Wang F,Liu X,et al.Hyperspectral polarization characteristics of typical camouflage target under desert background[J].Laser&Optoelectronics Progress,2018,55(5):051101.王小龙,王峰,刘晓,等.荒漠背景下典型伪装目标的高光谱偏振特性[J].激光与光电子学进展,2018,55(5):051101.
[8]Cui B S,Xue S Q,Ji Y J,et al.Camouflage effectiveness evaluation based on image feature[J].Infrared and Laser Engineering,2010,39(6):1178-1183.崔宝生,薛士强,姬艳军,等.图像特征的伪装效果评估技术[J].红外与激光工程,2010,39(6):1178-1183.
[9]Xu W D,LüX L,Chen B,et al.A model based on texture analysis for the performance evaluation of camouflage screen equipment[J].Acta Armamentarii,2002,23(3):329-331.许卫东,吕绪良,陈兵,等.一种基于纹理分析的伪装器材效果评价模型[J].兵工学报,2002,23(3):329-331.
[10]Lin W,Chen Y H,Wang J Y,et al.Camouflage assessment method based on image features and psychological perception quantity[J].Acta Armamentarii,2013,34(4):412-417.林伟,陈玉华,王吉远,等.基于图像特征与心理感知量的伪装效果评价方法[J].兵工学报,2013,34(4):412-417.
[11]Wang P Y,Zhao D H,Li M F.Optical camouflage effect assessment based on digital image inpainting technology[J].Laser&Optoelectronics Progress,2018,55(3):031011.王鹏烨,赵德辉,李明锋.基于图像修复技术的目标可见光伪装效果评价[J].激光与光电子学进展,2018,55(3):031011.
[12]Zhang R,Li W P,Mo T.Review of deep learning[J].Information and Control,2018,47(4):385-397,410.张荣,李伟平,莫同.深度学习研究综述[J].信息与控制,2018,47(4):385-397,410.
[13]Krizhevsky A,Sutskever I,Hinton G E.ImageNet classification with deep convolutional neural networks[J].Communications of the ACM,2017,60(6):84-90.
[14]He K M,Zhang X Y,Ren S Q,et al.Deep residual learning for image recognition[C]∥Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,2016:770-778.
[15]Huang G,Liu Z,Maaten L V D,et al.Densely connected convolutional networks[C]∥Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,2017:2261-2269.
[16]Xin Y,Yang J,Xie Z Q.A semantic overlapping community detecting algorithm in social networks based on random walk[J].Journal of Computer Research and Development,2015,52(2):499-511.辛宇,杨静,谢志强.基于随机游走的语义重叠社区发现算法[J].计算机研究与发展,2015,52(2):499-511.
[17]Wang Y,Tang J,Rao Q F,et al.High efficient K-means algorithm for determining optimal number of clusters[J].Journal of Computer Applications,2014,34(5):1331-1335.王勇,唐靖,饶勤菲,等.高效率的K-means最佳聚类数确定算法[J].计算机应用,2014,34(5):1331-1335.
[18]An Z,Xu X P,Yang J H,et al.Design of augmented reality head-up display system based on image semantic segmentation[J].Acta Optica Sinica,2018,38(7):0710004.安喆,徐熙平,杨进华,等.结合图像语义分割的增强现实型平视显示系统设计与研究[J].光学学报,2018,38(7):0710004.
[19]Li S M,Lei G Q,Fan R.Depth map superresolution based on two-channel convolutional neural network[J].Acta Optica Sinica,2018,38(10):1010002.李素梅,雷国庆,范如.基于双通道卷积神经网络的深度图超分辨研究[J].光学学报,2018,38(10):1010002.
[20]Sun H Q,Pang Y W.An neural network framework of self-learning uncertainty[J].Acta Optica Sinica,2018,38(6):0620002.孙汉卿,庞彦伟.一种自学习不确定度的神经网络架构[J].光学学报,2018,38(6):0620002.
[21]Long J,Shelhamer E,Darrell T.Fully convolutional networks for semantic segmentation[C]∥Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,2015:3431-3440.
[22]Isola P,Zhu J Y,Zhou T H,et al.Image-to-image translation with conditional adversarial networks[C]∥Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,2017:5967-5976.
[23]Goodfellow I J,Pouget-Abadie J,Mirza M,et al.Generative adversarial nets[C]∥Advances in Neural Information Processing Systems,2014:2672-2680.