基于深度置信网络(DBN)的赤潮高光谱遥感提取研究
|
摘要
赤潮是严重的海洋灾害,有效监测赤潮对于保护海洋生态环境具有重要意义。高光谱遥感具有光谱分辨率高、图谱合一等优势,适合于海洋赤潮监测。深度学习是机器学习领域的前沿,为高光谱遥感分类提供了新的思路。深度置信网络(Deep Belief Network,DBN)兼具监督分类与非监督分类的特点,通过构建DBN模型,将DBN应用于赤潮灾害遥感监测中,应用渤海机载高光谱遥感数据开展赤潮分类,以达到提取高光谱图像中赤潮水体范围的目的。通过设置对照实验,对比经典的SVM监督分类方法与ISODATA非监督分类方法,发现DBN模型在相同实验条件下具有更高的分类精度,赤潮遥感提取精度提高了3%～11%。
Red tide is a kind of serious marine disaster. Effective monitoring on red tides is of great significance for the protection of marine ecological environment. Hyperspectral remote sensing has the advantages of high spectral resolution and combines image with spectrum, which is suitable for marine red tide monitoring. Deep learning is the frontier of machine learning, which provides a new idea for hyperspectral remote sensing classification. Deep Belief Network(DBN) has the characteristics of both supervised classification and unsupervised classification. By constructing DBN model, DBN is applied to remote sensing monitoring on red tide disasters, and the Airborne Hyperspectral Remote Sensing Data of the Bohai Sea are used to classify red tides, in order to extract the range of red tide water in hyperspectral images. Compared with the classical SVM supervised classification method and ISODATA unsupervised classification method, the DBN model has higher classification accuracy under the same experimental conditions, and the accuracy of red tide remote sensing extraction is improved by 3%-11%.
Red tide is a kind of serious marine disaster. Effective monitoring on red tides is of great significance for the protection of marine ecological environment. Hyperspectral remote sensing has the advantages of high spectral resolution and combines image with spectrum, which is suitable for marine red tide monitoring. Deep learning is the frontier of machine learning, which provides a new idea for hyperspectral remote sensing classification. Deep Belief Network(DBN) has the characteristics of both supervised classification and unsupervised classification. By constructing DBN model, DBN is applied to remote sensing monitoring on red tide disasters, and the Airborne Hyperspectral Remote Sensing Data of the Bohai Sea are used to classify red tides, in order to extract the range of red tide water in hyperspectral images. Compared with the classical SVM supervised classification method and ISODATA unsupervised classification method, the DBN model has higher classification accuracy under the same experimental conditions, and the accuracy of red tide remote sensing extraction is improved by 3%-11%.
引文
[1]马毅.赤潮航空高光谱遥感赤潮探测技术研究[D].青岛:中国科学院海洋研究所,2003.Ma Yi. Research on red tide detection technology by aerial hyperspectral remote sensing[D]. Qingdao:Institute of Oceanography,Chinese Academy of Sciences, 2003.
[2]朱艳,刘晓莉,杨哲海.高光谱数据的降维及Tabu搜索算法的应用[J].测绘科学技术学报,2007,24(1):22-25.ZhuYan, Liu Xiao-li, Yang Zhe-hai. Dimension reduction of hyperspectral data and application of Tabu search algorithms[J]. Journal of Surveying and Mapping Science and Technology,2007,24(1):22-25.
[3]赵杰,周可法.联合光谱角匹配与随机森林的蚀变信息提取[J].遥感信息,2017,32(6):51-55.ZhaoJie, Zhou Ke-fa. Combined spectral angle matching and alteration information extraction of random forest[J]. Remote Sensing Information, 2017, 32(6):51-55.
[4]王立国,杜心平. K均值聚类和孪生支持向量机相结合的高光谱图像半监督分类[J].应用科技,2017,44(3):12-18.Wang Li-guo, Du Xin-ping. Semi-supervised classification of hyperspectral images based on K-means clustering and twin support vector machine[J]. Applied Technology,2017,44(3):12-18.
[5] Foody G M, Mathur A. A relative evaluation of multiclass image classification by support vector machines[J]. IEEE Transactions on Geoscience&Remote Sensing, 2004, 42(6):1335-1343.
[6] Hinton G E, Osindero S, Teh Y. A fast learning algorithm for deep belief nets[J]. Neural Computation,2006,18(7):1527-1554.
[7] YUD. Deep learning and its applications to signal and in-formation processing[J]. IEEE Signal Processing Magazine, 2011, 28(1):145-154.
[8] Bengio Y, Courville A, Vincent P. Representation learning:a review and new perspectives[J]. IEEE Transactions on Pattern Analysis&Machine Intelligence, 2013, 35(8):1798-1828.
[9]冉鹏,王灵,李昕.改进Softmax分类器的深度卷积神经网络及其在人脸识别中的应用[J].上海大学学报,2018,24(3):75-90.Ran Peng, Wang Ling, Li Xin. Deep convolution neural network with improved softmax classifier and its application in face recognition[J]. Journal of Shanghai University,2018,24(3):75-90.
[10] Mohamed A R, Dahl G E, Hinton G E. Acoustic modeling using deep belief networks[J]. IEEE Transactions on Audio Speech&Language Processing, 2011, 20(1):14-22.
[11]孙巧巧.基于深度学习的高光谱图像分类及参数设计研究[D].青岛:青岛科技大学, 2017.Sun Qiao-qiao. Research on hyperspectral image classification and parameter design based on deep learning[D].Qingdao:Qingdao University of Science and Technology, 2017.
[12]孙利平.基于深度学习的高光谱遥感图像分类算法研究[D].西安:西安电子科技大学,2018.Sun Li-ping. Research on hyperspectral remote sensing image classification algorithms based on deep learning[D]. Xi'an:Xi'an University of Electronic Science and Technology, 2018.
[13] Larochelle H, Bengio Y, Louradour J, et al.Exploring strategies for training deep neural networks[J]. Journal of Machine Learning Research, 2009,10(1):1-40.
[2]朱艳,刘晓莉,杨哲海.高光谱数据的降维及Tabu搜索算法的应用[J].测绘科学技术学报,2007,24(1):22-25.ZhuYan, Liu Xiao-li, Yang Zhe-hai. Dimension reduction of hyperspectral data and application of Tabu search algorithms[J]. Journal of Surveying and Mapping Science and Technology,2007,24(1):22-25.
[3]赵杰,周可法.联合光谱角匹配与随机森林的蚀变信息提取[J].遥感信息,2017,32(6):51-55.ZhaoJie, Zhou Ke-fa. Combined spectral angle matching and alteration information extraction of random forest[J]. Remote Sensing Information, 2017, 32(6):51-55.
[4]王立国,杜心平. K均值聚类和孪生支持向量机相结合的高光谱图像半监督分类[J].应用科技,2017,44(3):12-18.Wang Li-guo, Du Xin-ping. Semi-supervised classification of hyperspectral images based on K-means clustering and twin support vector machine[J]. Applied Technology,2017,44(3):12-18.
[5] Foody G M, Mathur A. A relative evaluation of multiclass image classification by support vector machines[J]. IEEE Transactions on Geoscience&Remote Sensing, 2004, 42(6):1335-1343.
[6] Hinton G E, Osindero S, Teh Y. A fast learning algorithm for deep belief nets[J]. Neural Computation,2006,18(7):1527-1554.
[7] YUD. Deep learning and its applications to signal and in-formation processing[J]. IEEE Signal Processing Magazine, 2011, 28(1):145-154.
[8] Bengio Y, Courville A, Vincent P. Representation learning:a review and new perspectives[J]. IEEE Transactions on Pattern Analysis&Machine Intelligence, 2013, 35(8):1798-1828.
[9]冉鹏,王灵,李昕.改进Softmax分类器的深度卷积神经网络及其在人脸识别中的应用[J].上海大学学报,2018,24(3):75-90.Ran Peng, Wang Ling, Li Xin. Deep convolution neural network with improved softmax classifier and its application in face recognition[J]. Journal of Shanghai University,2018,24(3):75-90.
[10] Mohamed A R, Dahl G E, Hinton G E. Acoustic modeling using deep belief networks[J]. IEEE Transactions on Audio Speech&Language Processing, 2011, 20(1):14-22.
[11]孙巧巧.基于深度学习的高光谱图像分类及参数设计研究[D].青岛:青岛科技大学, 2017.Sun Qiao-qiao. Research on hyperspectral image classification and parameter design based on deep learning[D].Qingdao:Qingdao University of Science and Technology, 2017.
[12]孙利平.基于深度学习的高光谱遥感图像分类算法研究[D].西安:西安电子科技大学,2018.Sun Li-ping. Research on hyperspectral remote sensing image classification algorithms based on deep learning[D]. Xi'an:Xi'an University of Electronic Science and Technology, 2018.
[13] Larochelle H, Bengio Y, Louradour J, et al.Exploring strategies for training deep neural networks[J]. Journal of Machine Learning Research, 2009,10(1):1-40.