基于深度卷積神經(jīng)網(wǎng)絡(luò)和多核學(xué)習(xí)的遙感圖像分類方法
doi: 10.11999/JEIT180628
-
1.
河海大學(xué)計(jì)算機(jī)與信息學(xué)院 ??南京 ??211100
-
2.
南京師范大學(xué)物理科學(xué)與技術(shù)學(xué)院 ??南京 ??210000
Remote Sensing Image Classification Method Based on Deep Convolution Neural Network and Multi-kernel Learning
-
1.
College of Computer and Information, Hohai University, Nanjing 211100, China
-
2.
School of Physics and Technology, Nanjing Normal University, Nanjing 210000, China
-
摘要:
為解決傳統(tǒng)遙感圖像分類方法特征提取過(guò)程復(fù)雜、特征表現(xiàn)力不強(qiáng)等問(wèn)題，該文提出一種基于深度卷積神經(jīng)網(wǎng)絡(luò)和多核學(xué)習(xí)的高分辨率遙感圖像分類方法。首先基于深度卷積神經(jīng)網(wǎng)絡(luò)對(duì)遙感圖像數(shù)據(jù)集進(jìn)行訓(xùn)練，學(xué)習(xí)得到兩個(gè)全連接層的輸出將作為遙感圖像的兩種高層特征；然后采用多核學(xué)習(xí)理論訓(xùn)練適合這兩種高層特征的核函數(shù)，并將它們映射到高維空間，實(shí)現(xiàn)兩種高層特征在高維空間的自適應(yīng)融合；最后在多核融合特征的基礎(chǔ)上，設(shè)計(jì)一種基于多核學(xué)習(xí)-支持向量機(jī)的遙感圖像分類器，對(duì)遙感圖像進(jìn)行精確分類。實(shí)驗(yàn)結(jié)果表明，與目前已有的基于深度學(xué)習(xí)的遙感圖像分類方法相比，該算法在分類準(zhǔn)確率、誤分類率和Kappa系數(shù)等性能指標(biāo)上均有所提升，在實(shí)驗(yàn)測(cè)試集上3個(gè)指標(biāo)分別達(dá)到了96.43%, 3.57%和96.25%，取得了令人滿意的結(jié)果。
-
關(guān)鍵詞:
- 高分辨率遙感圖像 /
- 分類 /
- 卷積神經(jīng)網(wǎng)絡(luò) /
- 多核學(xué)習(xí)
Abstract:To solve the problems of complex feature extraction process and low characteristic expressiveness of traditional remote sensing image classification methods, a high resolution remote sensing image classification method based on deep convolution neural network and multi-kernel learning is proposed. Firstly, the deep convolution neural network is constructed to train the remote sensing image data set to learn the outputs of two fully connected layers, which are taken as two high-level features of remote sensing images. Then, the multi-kernel learning is used to train the kernel functions for these two high-level features, so that they can be mapped to the high dimensional space, where these two features are fused adaptively. Finally, with the combined features, a remote sensing image classifier based on Multi-Kernel Learning-Support Vector Machine (MKL-SVM) is designed for remote sensing image classification. Experimental results show that compared with the existing deep learning based remote sensing classification methods, the proposed algorithm achieves improved results in terms of classification accuracy, error, and Kappa coefficient. On the experimental test set, the above three indicators reach 96.43%, 3.57%, and 96.25% respectively, and satisfactory results are obtained.
-
表 1 3種算法各分類性能指標(biāo)值
指標(biāo) 算法1 算法2 本文算法 Ac 0.8808 0.9337 0.9643 Er 0.1192 0.0663 0.0357 Kappa系數(shù) 0.8748 0.9304 0.9625 
下載: 導(dǎo)出CSV
-
WANG Xin, SHEN Siqiu, NING Chen, et al. Multi-class remote sensing objects recognition based on discriminative sparse representation[J]. Applied Optics, 2016, 55(6): 1381–1394. doi: 10.1364/ao.55.001381 WANG Xin, XIONG Xingnan, NING Chen, et al. Integration of heterogeneous features for remote sensing scene classification[J]. Journal of Applied Remote Sensing, 2018, 12(1): 15–23. doi: 10.1117/1.jrs.12.015023 NING Chen, LIU Wenbo, ZHANG Gong, et al. Enhanced synthetic aperture radar automatic target recognition method based on novel features[J]. Applied Optics, 2016, 55(31): 8893–8904. doi: 10.1364/ao.55.008893 馮子勇. 基于深度學(xué)習(xí)的圖像特征學(xué)習(xí)和分類方法的研究及應(yīng)用[D]. [博士論文], 華南理工大學(xué), 2016.FENG Ziyong. Research and application of image feature learning and classification methods based on deep learning[D]. [Ph.D. dissertation], South China University of Technology, 2016. 林妙真. 基于深度學(xué)習(xí)的人臉識(shí)別研究[D]. [碩士論文], 大連理工大學(xué), 2013.LIN Miaozhen. Research on face recognition based on deep learning[D]. [Master dissertation], Dalian University of Technology, 2013. 奚雪峰, 周國(guó)棟. 面向自然語(yǔ)言處理的深度學(xué)習(xí)研究[J]. 自動(dòng)化學(xué)報(bào), 2016, 42(10): 1445–1465. doi: 10.16383/j.aas.2016.c150682XI Xuefeng and ZHOU Guodong. A survey on learning for natural language processing[J]. Acta Automatica Sinica, 2016, 42(10): 1445–1465. doi: 10.16383/j.aas.2016.c150682 張建華. 基于深度學(xué)習(xí)的語(yǔ)音識(shí)別應(yīng)用研究[D]. [碩士論文], 北京郵電大學(xué), 2015.ZHANG Jianhua. Research into speech recognition application based on deep learning[D]. [Master dissertation], Beijing University of Posts and Telecommunications, 2015. 劉大偉, 韓玲, 韓曉勇. 基于深度學(xué)習(xí)的高分辨率遙感影像分類研究[J]. 光學(xué)學(xué)報(bào), 2016, 36(4): 306–314. doi: 10.3788/AOS201636.0428001LIU Dawei, HAN Ling, and HAN Xiaoyong. High spatial resolution remote sensing image classification based on deep learning[J]. Acta Optica Sinica, 2016, 36(4): 306–314. doi: 10.3788/AOS201636.0428001 曲景影, 孫顯, 高鑫. 基于CNN模型的高分辨率遙感圖像目標(biāo)識(shí)別[J]. 國(guó)外電子測(cè)量技術(shù), 2016, 35(8): 45–50. doi: 10.19652/j.cnki.femt.2016.08.011QU Xianying, SUN Xian, and GAO Xin. Remote sensing image target recognition based on CNN[J]. Foreign Electronic Measurement Technology, 2016, 35(8): 45–50. doi: 10.19652/j.cnki.femt.2016.08.011 周敏, 史振威, 丁火平. 遙感圖像飛機(jī)目標(biāo)分類的卷積神經(jīng)網(wǎng)絡(luò)方法[J]. 中國(guó)圖象圖形學(xué)報(bào), 2017, 22(5): 702–708. doi: 10.11834/jig.160595ZHOU Min, SHI Zhenwei, and DING Huoping. Aircraft classification in remote-sensing images using convolutional neural networks[J]. Journal of Image and Graphics, 2017, 22(5): 702–708. doi: 10.11834/jig.160595 黃潔, 姜志國(guó), 張浩鵬, 等. 基于卷積神經(jīng)網(wǎng)絡(luò)的遙感圖像艦船目標(biāo)檢測(cè)[J]. 北京航空航天大學(xué)學(xué)報(bào), 2017, 43(9): 1841–1848. doi: 10.13700/j.bh.1001-5965.2016.0755HUANG Jie, JIANG Zhiguo, ZHANG Haopeng, et al. Ship object detection in remote sensing images using convolutional neural networks[J]. Journal of Beijing University of Aeronautics and Astronsutics, 2017, 43(9): 1841–1848. doi: 10.13700/j.bh.1001-5965.2016.0755 LECUN Y, BOTTOU L, BENGIO Y, et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11): 2278–2324. doi: 10.1109/5.726791 GAO Ligang, CHEN Paiyu, and YU Shimeng. Demonstration of convolution kernel operation on resistive crosspoint array[J]. IEEE Electron Device Letters, 2016, 7(7): 870–873. doi: 10.1109/LED.2016.2573140 GIRSHICK R, IANDOLA F, DARRELL T, et al. Deformable part models are convolutional neural networks[C]. Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition. Boston, USA, 2015: 437–446. doi: 10.1109/CVPR.2015.7298641. HINTON G E. How neural networks learn from experience[J]. Scientific American, 1992, 267(3): 145–151. doi: 10.1038/scientificamerican0992-144 BOUVRIE J. Notes on convolutional neural networks[R]. Massachusetts: Center for biological and computational learning, 2006: 38–44. RUMELHAR T D E, HINTON G E, and WILLIAMS R J. Learning representations by back-propagating errors[J]. Nature, 1986, 323(6088): 533–536. doi: 10.1038/323533a0 KRIZHEVSKY A, SUTSKEVER I, and HINTON G E. ImageNet classification with deep convolutional neural networks[C]. Proceedings of Advances in Neural Information Processing Systems 25, Lake Tahoe, USA, 2012: 1097–1105. doi: 10.1145/3065386. NING Chen, LIU Wenbo, and WANG Xin. Infrared object recognition based on monogenic features and multiple kernel learning[C]. The 2018 IEEE International Conference on Image, Vision and Computing. Chongqing, China, 2018: 204–208. -
計(jì)量
- 文章訪問(wèn)數(shù): 4261
- HTML全文瀏覽量: 1475
- PDF下載量: 366
- 被引次數(shù): 0
下載:
