Application of label clustering loss in the classification of remote sensing images
-
摘要: 遥感影像的场景信息在影像解译和各领域的实际生产生活中具有重要的应用价值,针对遥感影像类内差异性大、类间差异性小的特点,该文在中心损失函数的基础上进一步研究,提出了一种新的标签聚类损失函数。首先使用类标签中心初始化方法对类别中心进行参数初始化,其次使用正弦衰减学习率使模型在预热阶段保持类别中心的稳定性,然后使用欧氏距离与余弦距离来进行类内特征的聚集以及类间中心的远离。并且使用VGG16和ResNet50两个网络模型在NWPU-RESISC45数据集上进行验证,准确率分别提高了2.3%和5.7%。通过试验表明,该方法能够有效地实现特征的聚集与类别中心的远离,提升网络模型的准确率,在遥感影像分类任务中具有一定的发展前景。Abstract: Scene information of remote sensing images has important application value in image interpretation and actual production and life in various fields. In view of the characteristics of remote sensing images with large intra-class differences and small inter-class differences, this paper further studies the center loss function and proposes a new label clustering loss function. Firstly, the class center is initialized by using the class label center initialization method. Secondly, the sinusoidal attenuation learning rate is used to keep the stability of the class center in the preheating stage. Then, Euclidean distance and cosine distance are used to gather the intra-class features and keep them away from the class center. Furthermore, two network models, VGG16 and ResNet50, are used to verify on NWPU-RESISC45 data set, and the accuracy is improved by 2.3% and 5.7% respectively. Experiments show that the method proposed in this paper can effectively cluster the features and separate class centers, and improve the accuracy of the network model, which has a certain development prospect in the classification of remote sensing images.
-
-
[1] 李冠东, 张春菊, 王铭恺, 等. 卷积神经网络迁移的高分影像场景分类学习[J]. 测绘科学, 2019, 44(4):116-123,174.[1] Li G D, Zhang C J, Wang M K, et al. High resolution remote sensing image classification learning based on convolutional neural network migration[J]. Surveying and Mapping Science, 2019, 44(4):116-123,174.[2] 何小飞, 邹峥嵘, 陶超, 等. 联合显著性和多层卷积神经网络的高分影像场景分类[J]. 测绘学报, 2016, 45(9):1073-1080.[2] He X F, Zou Z R, Tao C, et al. High resolution image scene classification based on saliency and multilayer convolutional neural network[J]. Acta Surveying and Mapping, 2016, 45(9):1073-1080.[3] Liu Y, Zhong Y, Qin Q. Scene classificationbased on multiscale convolutional neural network[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(12):7109-7121. [4] Zan C, Wang S, X Hou, et al. Recurrent transformer network for remote sensing scene categorisation[C]. // Proceedings of the 2018 British Machine Vision Conference.Durham:BMVA, 2018:266.[5] Wang Q, Liu S, Chanussot J, et al. Scene classification with recurrent attention of VHR remote sensing images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(2):1155-1167. [6] Sun H, Li S, Zheng X, et al. Remote sensing scene classification by gated bidirectional network[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019(99):1-15.[7] Li Y, Liu M, Zhang S. Classification of optical remote sensing images based on convolutional neural network[C]// 2019 6th International Conference on Control,Decision and Information Technologies (CoDIT).IEEE, 2019:801-806.[8] 杨斌, 李成华, 江小平, 等. 一种用于提升深度学习分类模型准确率的正则化损失函数[J]. 中南民族大学学报(自然科学版), 2020, 39(1):74-78.[8] Yang B, Li C H, Jiang X P, et al. A regularized loss function for improving the accuracy of deep learning classification model[J]. Journal of Central South University for Nationalities,Natural Science Edition, 2020, 39(1):74-78.[9] Liu Y, Huang C. Scene classification via triplet networks[J]. IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing, 2018, 11(1):220-237.[10] Khatun A, Denman S, Sridharan S, et al. Joint identification-verification for person re-identification:A four stream deep learning approach with improved quartet loss function[J]. Computer Vision and Image Understanding, 2020,197-198:102989.[11] Shi W, Gong Y, Tao X, et al. Fine-grained image classification using modified DCNNs trained by cascaded softmax and generalized large-margin losses[J]. IEEE Transactions on Neural Networks and Learning Systems, 2018, 30(3):683-694. [12] Reddy A C, Saraschandrika A, Reddy A V. Study of the Clustering algorithms for hyper spectral remote sensing images[J]. Journal of Hyperspectral Remote Sensing, 2020, 10(2):117-121. [13] Gong C, Han J, Lu X. Remote sensing image scene classification:Benchmark and state of the art[J]. Proceedings of the IEEE, 2017, 105(10):1865-1883. [14] 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,(99):1-12.[15] 杨宏炳, 迟勇欣, 王金光. 基于剪枝网络的知识蒸馏对遥感卫星图像分类方法[J/OL]. 计算机应用研究, 2021, 38(8):2469-2473.[15] Yang H B, Chi Y X, Wang J G. Remote sensing satellite image classification method based on knowledge distillation of pruning net work[J/OL]. Computer Application Research, 2021, 38(8):2469-2473.[16] 马欣悦, 王梨名, 祁昆仑, 等. 基于多尺度循环注意力网络的遥感影像场景分类方法[J]. 地球科学, 2021, 46(10):3740-3752.[16] Ma X Y, Wang L M, Qi K L, et al. Remotesensing image scene classification method based on multi-scale cyclic attention network[J]. Earth Science, 2021, 46(10):3740-3752.[17] Zhou Z, Zheng Y, Ye H, et al. Satellite image scene classification via ConvNet with context aggregation[J]. PCM, 2018(2):329-339.[18] 业巧林, 许平, 张冬. 基于深度学习特征和支持向量机的遥感图像分类[J]. 林业工程学报, 2019, 4(2):125-131.[18] Ye Q L, Xu X P, Zhang D. Remote sensing image classification based on deep learning feature and support vector machine[J]. Journal of Forestry Engineering, 2019, 4(2):125-131.[19] Wang J, Liu W, Ma L, et al. IORN:An effective remote sensing image scene classification framework[J]. IEEE Geoscience and Remote Sensing Letters, 2018, 15(11):1695-1699.
-
计量
- 文章访问数: 962
- PDF下载数: 29
- 施引文献: 0
下载: