Deep Induction Network for Small Samples Classification of Hyperspectral Images
Deep Induction Network for Small Samples Classification of Hyperspectral Images
复制标题
用于高光谱图像小样本分类的深度归纳网络
DOI：
10.1109/jstars.2020.3002787
复制
发表时间：
2020-01-01
影响因子：
5.5
通讯作者：
Wei, Xiangpo
中科院分区：
文献类型：
--
作者：
Gao, Kuiliang;Guo, Wenyue;Wei, Xiangpo
近年来，深度学习模型在高光谱图像（HSI）分类中取得了巨大的成功。然而，由于深度学习模型的参数空间大与HSI中标记样本不足之间的矛盾，大多数深度学习模型在小样本条件下无法获得满意的结果。针对这一问题，本文设计了一种基于归纳网络的深度模型，以提高HSI在小样本条件下的分类性能。具体而言，采用了典型的元训练策略，使模型具有更强的泛化能力，从而可以用很少的标记样本（例如每个类5个样本）准确地区分新的类。此外，为了处理HSI中同一类样本的不同特征所引起的干扰，利用动态路由算法引入了类智能诱导模块，该模块可以将样本智能表示诱导为类智能级别表示。所获得的类层次表示具有更好的可分离性，使所设计的模型能够生成更准确和鲁棒的分类结果。在三个公共HSI上进行了大量实验以验证所提出方法的有效性。结果表明，在小样本条件下，我们的方法优于现有的深度学习方法。
Recently, the deep learning models have achieved great success in hyperspectral images (HSI) classification. However, most of the deep learning models fail to obtain satisfactory results under the condition of small samples due to the contradiction between the large parameter space of the deep learning models and the insufficient labeled samples in HSI. To address the problem, a deep model based on the induction network is designed in this article to improve the classification performance of HSI under the condition of small samples. Specifically, the typical meta-training strategy is adopted, enabling the model to acquire stronger generalization ability, so as to accurately distinguish the new classes with only a few labeled samples (e.g., five samples per class). Moreover, in order to deal with the disturbance caused by the various characteristics of the samples in the same class in HSI, the class-wise induction module is introduced utilizing the dynamic routing algorithm, which can induce the sample-wise representations to the class-wise level representations. The obtained class-wise level representations possess better separability, allowing the designed model to generate more accurate and robust classification results. Extensive experiments are carried out on three public HSI to verify the effectiveness of the proposed method. The results demonstrate that our method outperforms existing deep learning methods under the condition of small samples.