1. 毕业设计（论文）主要目标：
高光谱图像光谱维度高、数据冗余大且存在非线性特性，直接进行分类易导致Hughes现象，而卷积神经网络能够自动地从图像中提取空间特征和光谱信息，本文的设计目标是找到一种对高光谱遥感数据进行准确分类的方法。
2. 毕业设计（论文）主要内容：
1.利用python从原始高光谱遥感数据中获取大量13x13大小的像元.每个像元的大小为13x13x200，每一个像元对应一个groudtruth，以groudtruth作为标签。
2.再将读取到的数据集打乱，做成数据集。其中10%作为训练集，剩余90%作为测试集。
3.使用深度残差网络实现高光谱遥感数据的准确分类，设置好网络参数。将训练数据和测试数据分别送入深度残差网络进行训练与测试。
3. 主要参考文献
| [1] Bioucas-Dias J M, Plaza A, Camps-Valls G, Scheunders P.Hyperspectral remote sensing data analysis and future challenges. IEEEGeoscience and Remote Sensing Magazine, 2013, 1(2): 636 [2] Bioucas-Dias J M, Plaza A, Camps-Valls G, Scheunders P.Hyperspectral remote sensing data analysis and future challenges. IEEEGeoscience and Remote Sensing Magazine, 2013, 1(2): 636 [3] Fauvel M, Tarabalka Y, Benediktsson J A, Chanussot J. Advances inspectral-spatial classication of hyperspectral images. Proceedings of theIEEE, 2013, 101(3): 652675 [4] Landgrebe D. Hyperspectral image data analysis. IEEE SignalProcessing Magazine, 2002, 19(1): 1728 [5] Nasrabadi N M. Hyperspectral target detection: an overview ofcurrent and future challenges. IEEE Signal Processing Magazine, 2014, 31(1):3444 [6] Plaza A, Martinez P, Perez R, Plaza J. A new approach to mixedpixel classication of hyperspectral imagery based on extended morphologicalproles. Pattern Recognition, 2004, 37(6): 10971116 [7] Samat A, Li J, Liu S C, Du P J, Miao Z L, Luo J Q. Improved hyperspectralimage classication by active learning using pre-designed mixed pixels.Pattern Recognition, 2015, 51(C): 4358 [8] Imani M, Ghassemian H. Band clustering-based feature extraction forclassication of hyperspectral images using limited training samples. IEEEGeoscience and Remote Sensing Letters, 2014, 11(8): 13251329 [9] Li F, Xu L, Siva P, Wong A. Hyperspectral image classication withlimited labeled training samples using enhanced ensemble learning andconditional random elds. IEEE Journal of Selected Topics in Applied EarthObservations and Remote Sensing, 2015, 8(6): 112 [10] Hughes G. On the mean accuracy of statistical pattern recognizers.IEEE Transactions on Information Theory, 1968, 14(1): 5563 [11] Prasad S, Bruce L M. Limitations of principal components analysisfor hyperspectral target recognition. IEEE Geoscience and Remote SensingLetters, 2008, 5(4): 625629 [12] Li W, Prasad S, Fowler J E, Bruce L M. Locality-preservingdimensionality reduction and classication for hyperspectral image analysis.IEEE Transactions on Geoscience and Remote Sensing, 2012, 50(4): 11851198 [13] Li J, Bioucas J M, Plaza A. Hyperspectral image segmentation usinga new Bayesian approach with active learning. IEEE Transactions on Geoscienceand Remote Sensing, 2011, 49(10): 39473960 [14] Liao W, Pizurica A, Scheunders P, Philips W. Semisupervised localdiscriminant analysis for feature extraction in hyperspectral images. IEEETransactions on Geoscience and Remote Sensing, 2013, 51(1): 184198 [15] Li W, Prasad S, Fowler J E, Bruce L M. Locality-preservingdiscriminant analysis in kernel-induced feature spaces for hyperspectralimage classication. IEEE Geoscience and Remote Sensing Letters, 2011, 8(5):894898 [16] Hinton G E, Osindero S, Teh Y W. A fast learning algorithm fordeep belief nets. Neural computation, 2006, 18(7): 15271554 [17] Krizhevsky A, Sutskever I, Hinton G. Imagenet classication withdeep convolutional neural networks. In: Proceedings of Advances in NeuralInformation Processing Systems. Nevada, USA: Curran Associates, 2012.10971105 [18] Hinton G E, Salakhutdinov R R. Reducing the dimensionality of datawith neural networks. Science, 2006, 313(5786): 504507 |
