Remote sensing image classification based on DenseNet feature hashing
-
摘要: 针对遥感场景的精准分类,提出了密集网络特征哈希的场景分类算法。基于密集网络输出的高层语义特征,经全连接层过渡降维后,激活函数生成归一化的特征向量作为分类层的判断输入,形成端到端的分类网络。训练后的网络作为特征提取器,将测试数据激活层特征映射生成二值哈希码,最后采用支持向量机分类。所提出的算法分别在UC Merced,WHU和NWPU-RESISC45公开数据集进行了验证,分别与传统局部特征描述子、迁移学习、深度特征编码3个层次的多种算法进行了对比,实验结果表明,相比于传统中低层语义特征,分类准确度得到大幅度提高; 相比于深度学习网络的迁移,密集特征映射表达精细,聚集影像核心类别判断要素,更符合遥感影像的特征分布; 相比于深度特征编码算法,特征结构简单,分类精度高,迁移和拓展性强,可以满足不同遥感场景分类要求。Abstract: To achieve accurate remote sensing scene classification, this study proposed a classification algorithm based on DenseNet feature hashing. First, dimension reduction was conducted for high-level semantic features output by a DenseNet through a fully connected layer. Then, normalized feature vectors were generated as the input of the classification layer using an activation function, and an end-to-end classification network was formed. Using the trained network as a feature extractor, the features of the activation layer of test data were mapped into binary hash codes. Finally, the remote sensing scene classification was conducted using support vector machine. The new algorithm was validated on public data sets UC Merced, WHU, and NWPU-RESISC45, and its classification effect was compared with that of multiple algorithms at three levels, namely the conventional local feature descriptor, transfer learning, and depth feature coding. The experimental results are as follows. The new algorithm had significantly higher classification accuracy than conventional algorithms based on mid- and low-level semantic features. Compared with the algorithm based on transfer learning, the proposed algorithm has fine-scale DenseNet feature mapping and accumulates elements used to determine core categories of images and, thus, is more suitable for the feature distribution of remote sensing images. Compared with the depth feature coding algorithm, the new algorithm has a simple feature structure, high classification accuracy, and strong transferability and extensibility and, thus, can meet the classification requirements of different remote sensing scenarios.
-
Key words:
- transfer learning /
- feature coding /
- DenseNet /
- hash code
-
-
[1] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[C]// Proceedings of the International Conference on Neural Information Processing Systems. Lake Tahoe: NIPS, 2012:1097-1105.
[2] He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016:770-778.
[3] Huang G, Liu Z, Van Der Maaten L, et al. Densely connected convolutional networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Honolulu:IEEE, 2017:4700-4708.
[4] Wang G, Fan B, Xiang S, et al. Aggregating rich hierarchical features for scene classification in remote sensing imagery[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2017, 10(9):4104-4115.
[5] Razavian A S, Sullivan J, Carlsson S, et al. Visual instance retrieval with deep convolutional networks[J]. ITE Transactions on Media Technology and Applications, 2016, 4(3):251-258.
[6] Zheng L, Yang Y, Tian Q. SIFT meets CNN:A decade survey of instance retrieval[J]. IEEE Transactions on Pattern Analysis Machine Intelligence, 2018, 40(5):1224-1244.
[7] Hu F, Xia G S, Hu J, et al. Transferring deep convolutional neural networks for the scene classification of high-resolution remote sensing imagery[J]. Remote Sensing, 2015, 7(11):14680-14707.
[8] Cheng G, Li Z, Yao X, et al. Remote sensing image scene classification using bag of convolutional features[J]. IEEE Geoscience and Remote Sensing Letters, 2017, 14(10):1735-1739.
[9] 王鑫, 李可, 宁晨, 等. 基于深度卷积神经网络和多核学习的遥感图像分类方法[J]. 电子与信息学报, 2019, 41(5):1098-1105.
[10] Wang X, Li K, Ning C, et al. Remote sensing image classification method based on deep convolution neural network and multi-kernel learning[J]. Journal of Electronics & Information Technology, 2019, 41(5):1098-1105.
[11] He N, Fang L, Li S, et al. Remote sensing scene classification using multilayer stacked covariance pooling[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(12):6899-6910.
[12] Cheng G, Yang C, Yao X, et al. When deep learning meets metric learning:Remote sensing image scene classification via learning discriminative CNNs[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(5):2811-2821.
[13] 刘异, 庄姊琪, 闫利, 等. 联合Fisher核编码和卷积神经网络的影像场景分类[J]. 遥感信息, 2018, 33(4):8-15.
[14] Liu Y, Zhuang Z Q, Yan L, et al. Combined Fisher kernel coding framework with convolutional neural network for remote sensing scene classification[J]. Remote Sensing Information, 2018, 33(4):8-15.
[15] 余东行, 张保明, 赵传, 等. 联合卷积神经网络与集成学习的遥感影像场景分类[J]. 遥感学报, 2020, 24(6):717-727.
[16] Yu D H, Zhang B M, Zhao C, et al. Scene classification of remote sensing image using ensemble convolutional neural network[J]. Journal of Remote Sensing, 2020, 24(6):717-727.
[17] 李亚飞, 董红斌. 基于卷积神经网络的遥感图像分类研究[J]. 智能系统学报, 2018, 13(4):550-556.
[18] Li Y F, Dong H B. Classification of remote-sensing image based on convolutional neural network[J]. CAAI Transactions on Intelligent Systems, 2018, 13(4):550-556.
[19] 张哲益, 曹卫华, 朱蕊, 等. 基于脉冲卷积神经网络稀疏表征的高分辨率遥感图像场景分类方法[J]. 控制与决策, 2022, 37(9):2305-2313.
[20] Zhang Z Y, Cao W H, Zhu R, et al. Sparse representation with spike convolutional neural networks for scene classification of remote sensing images of high resolution[J]. Control and Decision, 2022, 37(9):2305-2313.
[21] Selvaraju R R, Cogswell M, Das A, et al. Grad-cam:Visual explanations from deep networks via gradient-based localization[C]// Proceedings of the IEEE International Conference on Computer Vision. Venice:IEEE, 2017:618-626.
[22] Lin K, Yang H F, Hsiao J H, et al. Deep learning of binary hash codes for fast image retrieval[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops. Boston:IEEE, 2015:27-35.
[23] Lazebnik S, Schmid C, Ponce J. Beyond bags of features:Spatial pyramid matching for recognizing natural scene categories[C]// 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06). New York: IEEE, 2006:2169-2178.
[24] Arandjelovic R, Zisserman A. All about VLAD[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Portland:IEEE, 2013:1578-1585.
[25] 李国祥, 马文斌, 王继军. 稠密特征编码的遥感场景分类算法[J]. 小型微型计算机系统, 2021, 42(4):766-772.
[26] Li G X, Ma W B, Wang J J. Remote sensing image classification based on dense feature coding[J]. Journal of Chinese Computer Systems, 2021, 42(4):766-772.
[27] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J/OL]. arxiv, 2014[2022-04-24]. https://arxiv.org/pdf/1409.1556.pdf.
[28] Szegedy C, Liu W, Jia Y, et al. Going deeper with convolutions[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Boston:IEEE, 2015:1-9.
[29] Arandjelovic R, Gronat P, Torii A, et al. NetVLAD:CNN architecture for weakly supervised place recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016:5297-5307.
[30] Lin T, Roychowdhury A, Maji S. Bilinear convolutional neural networks for fine-grained visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(6):1309-1322.
[31] Sánchez J, Perronnin F, Mensink T, et al. Image classification with the fisher vector:Theory and practice[J]. International Journal of Computer Vision, 2013, 105(3):222-245.
-
计量
- 文章访问数: 776
- PDF下载数: 69
- 施引文献: 0
下载: