基于CNN的湍流图像退化强度分类研究
|
摘要
湍流图像的复原一直是退化图像领域的研究热点,但依据湍流干扰强度对图像进行分类研究相对较少.不同场景的高空航拍图像进行大气湍流处理.调整湍流退化强度值,生成2000张对应的湍流干扰图像,再对这些图像进行预处理后送入卷积神经网络中进行湍流退化强度分类,最后通过优化搭建的卷积神经网络模型的激活函数以及对学习率的调整进一步提升分类准确率.实验表明,卷积神经网络对不同干扰强度的湍流退化图像分类准确率达到80%左右,结果表明该方法对大气湍流退化图像的复原具有一定指导意义.
Turbulence image restoration is a hot topic in the field of meteorology, but there are few studies on the classification of turbulent disturbance intensity images. Aerial images of different scenarios are processed for atmospheric turbulence. 2000 corresponding turbulent interference images are generated by adjusting the intensity of turbulence degradation value, and then they are sent into the convolutional neural network after the image preprocessing in turbulent degradation intensity classification, finally by optimizing the structures, convolutional neural network model of the activation function and the adjustment of vector for further improving classification accuracy. Experiments show that the classification accuracy of turbulent degradation images which have different interference intensities by convolutional neural network is about 80%, which demonstrates that this method is of guiding significance for the restoration of atmospheric turbulent degradation images.
Turbulence image restoration is a hot topic in the field of meteorology, but there are few studies on the classification of turbulent disturbance intensity images. Aerial images of different scenarios are processed for atmospheric turbulence. 2000 corresponding turbulent interference images are generated by adjusting the intensity of turbulence degradation value, and then they are sent into the convolutional neural network after the image preprocessing in turbulent degradation intensity classification, finally by optimizing the structures, convolutional neural network model of the activation function and the adjustment of vector for further improving classification accuracy. Experiments show that the classification accuracy of turbulent degradation images which have different interference intensities by convolutional neural network is about 80%, which demonstrates that this method is of guiding significance for the restoration of atmospheric turbulent degradation images.
引文
1黄真理.湍流的分形特征.力学进展,2000,30(4):581-596.
2闫传忠.基于分形的大气湍流随机相位屏数值模拟[硕士学位论文].青岛:中国海洋大学,2009.
3沙婵娟.基于分形的自然场景模拟[硕士学位论文].贵阳:贵州师范大学,2009.
4石秀英.基于估计点扩散函数的湍流图像复原算法研究[硕士学位论文].长春:长春理工大学,2012.
5王珺楠,邱欢,张丽娟,等.基于小波域Curvelet变换的湍流图像去噪算法.液晶与显示,2017,32(11):905-913.
6洪汉玉,张天序.基于小波分解的湍流退化图像的快速复原算法.红外与毫米波学报,2003,22(6):451-456.[doi:10.3321/j.issn:1001-9014.2003.06.012]
7黎明,杨杰.基于支持向量机的湍流退化图像加速复原算法.红外与毫米波学报,2009,28(6):472-475.[doi:10.3321/j.issn:1001-9014.2009.06.016]
8 Hubel DH,Wiesel TN.Receptive fields,binocular interaction and functional architecture in the cat’s visual cortex.The Journal of Physiology,1962,160(1):106-154.[doi:10.1113/jphysiol.1962.sp006837]
9 LeCun Y,Bottou L,Bengio Y,et al.Gradient-based learning applied to document recognition.Proceedings of the IEEE,1998,86(11):2278-2324.[doi:10.1109/5.726791]
10 Srivastava N,Hinton G,Krizhevsky A,et al.Dropout:Asimple way to prevent neural networks from overfitting.The Journal of Machine Learning Research,2014,15(1):1929-1958.
11 Huang GB,Lee H,Learned-Miller E.Learning hierarchical representations for face verification with convolutional deep belief networks.Proceedings of 2012 IEEE Conference on Computer Vision and Pattern Recognition.Providence,RI,USA.2012.2518-2525.
12王海霞.基于卷积神经网络的接地网腐蚀程度分类方法研究[硕士学位论文].西安:西安科技大学,2017.
13包青平.基于深度学习的服装图像分类与检索[硕士学位论文].杭州:浙江大学,2017.
14祝军.基于深度学习模型的图像分类研究[硕士学位论文].宁波:宁波大学,2015.
15石祥滨,房雪键,张德园,等.基于深度学习混合模型迁移学习的图像分类.系统仿真学报,2016,28(1):167-173,182.
16张弛.基于卷积神经网络的鞋印图像分类算法研究[硕士学位论文].大连:大连海事大学,2016.
17吉春旭.基于卷积神经网络的岩石图像分类研究与应用[硕士学位论文].西安:西安石油大学,2017.
18龚震霆,陈光喜,曹建收.卷积神经网络在脑脊液图像分类上的应用.计算机工程与设计,2017,38(4):1056-1061.
19顾郑平,朱敏.基于深度学习的鱼类分类算法研究.计算机应用与软件,2018,35(1):200-205.
2闫传忠.基于分形的大气湍流随机相位屏数值模拟[硕士学位论文].青岛:中国海洋大学,2009.
3沙婵娟.基于分形的自然场景模拟[硕士学位论文].贵阳:贵州师范大学,2009.
4石秀英.基于估计点扩散函数的湍流图像复原算法研究[硕士学位论文].长春:长春理工大学,2012.
5王珺楠,邱欢,张丽娟,等.基于小波域Curvelet变换的湍流图像去噪算法.液晶与显示,2017,32(11):905-913.
6洪汉玉,张天序.基于小波分解的湍流退化图像的快速复原算法.红外与毫米波学报,2003,22(6):451-456.[doi:10.3321/j.issn:1001-9014.2003.06.012]
7黎明,杨杰.基于支持向量机的湍流退化图像加速复原算法.红外与毫米波学报,2009,28(6):472-475.[doi:10.3321/j.issn:1001-9014.2009.06.016]
8 Hubel DH,Wiesel TN.Receptive fields,binocular interaction and functional architecture in the cat’s visual cortex.The Journal of Physiology,1962,160(1):106-154.[doi:10.1113/jphysiol.1962.sp006837]
9 LeCun Y,Bottou L,Bengio Y,et al.Gradient-based learning applied to document recognition.Proceedings of the IEEE,1998,86(11):2278-2324.[doi:10.1109/5.726791]
10 Srivastava N,Hinton G,Krizhevsky A,et al.Dropout:Asimple way to prevent neural networks from overfitting.The Journal of Machine Learning Research,2014,15(1):1929-1958.
11 Huang GB,Lee H,Learned-Miller E.Learning hierarchical representations for face verification with convolutional deep belief networks.Proceedings of 2012 IEEE Conference on Computer Vision and Pattern Recognition.Providence,RI,USA.2012.2518-2525.
12王海霞.基于卷积神经网络的接地网腐蚀程度分类方法研究[硕士学位论文].西安:西安科技大学,2017.
13包青平.基于深度学习的服装图像分类与检索[硕士学位论文].杭州:浙江大学,2017.
14祝军.基于深度学习模型的图像分类研究[硕士学位论文].宁波:宁波大学,2015.
15石祥滨,房雪键,张德园,等.基于深度学习混合模型迁移学习的图像分类.系统仿真学报,2016,28(1):167-173,182.
16张弛.基于卷积神经网络的鞋印图像分类算法研究[硕士学位论文].大连:大连海事大学,2016.
17吉春旭.基于卷积神经网络的岩石图像分类研究与应用[硕士学位论文].西安:西安石油大学,2017.
18龚震霆,陈光喜,曹建收.卷积神经网络在脑脊液图像分类上的应用.计算机工程与设计,2017,38(4):1056-1061.
19顾郑平,朱敏.基于深度学习的鱼类分类算法研究.计算机应用与软件,2018,35(1):200-205.