Recently, much research has shown that deep learning methods are superior in scene classification for remote sensing images (HSIs). However, the lack of labelled samples and computational resources present serious obstacles for HSI scene classification. To address these problems, we consider the pragmatic deep learning problem with only a few or no labelled samples. A general semi-supervised scene classification method based on clustering and transfer learning named GSCCTL is proposed to transfer outstanding supervised learning models to semi-supervised scene classification. Specifically, the GSCCTL model iterates between (1) scene clustering and (2) transfer learning to improve scene classification performance for the HSI data set. Since the clustering results may contain noise, we propose a new method to calculate the distance between the samples and the clustering center and add a reliable sample selection operation to the data after clustering. Initially, the model’s scene classification performance is weak, and transfer learning is carried out on a few samples close to the clustering center. In subsequent iterations, the model is constantly optimized, more and more samples are selected as reliable samples for training, and the model’s scene classification performance is strengthened. Through iterative clustering and transfer learning, the model is progressively optimized and gradually converged until the end condition is reached. Extensive experiments on the UCMerced, AID, and NWPU-RESISC45 remote sensing data sets demonstrate that the GSCCTL model based on clustering and transfer learning improved scene classification accuracy with only a few labelled samples and is suitable for semi-supervised scene classification.

With the rapid development of remote sensing technology, change detection (CD) methods based on remote sensing images have been widely used in land resource planning, disaster monitoring, and urban expansion, among other fields. The purpose of CD is to accurately identify changes on the Earth’s surface. However, most CD methods focus on changes between the pixels of multitemporal remote sensing image pairs while ignoring the coupled relationships between them. This often leads to uncertainty about edge pixels with regard to changing objects and misclassification of small objects. To solve these problems, we propose a CD method for remote sensing images that uses a coupled dictionary and deep learning. The proposed method realizes the spatial-temporal modeling and correlation of multitemporal remote sensing images through a coupled dictionary learning module and ensures the transferability of reconstruction coefficients between multisource image blocks. In addition, we constructed a differential feature discriminant network to calculate the dissimilarity probability for the change area. A new loss function that considers true/false discrimination loss, classification loss, and cross-entropy loss is proposed. The most discriminating features can be extracted and used for CD. The performance of the proposed method was verified on two well-known CD datasets. Extensive experimental results show that the proposed method is superior to other methods in terms of precision, recall, F₁-score, , and .

The convolutional neural network (CNN) has promoted hyperspectral images (HSIs) classification performance. However, the size of the convolutional kernel is fixed, whereas the size of objects in HSIs varies greatly; training a CNN requires a large number of samples with label, but manually tagging each pixel of HSIs is time-consuming and labor-intensive. To address above problems, a semi-supervised robust self-ensemble network (SRSN) is proposed in this letter. The SRSN contains a basic network and an ensemble network. The two networks can learn from each other to realize self-ensemble learning. Specifically, the deformable convolution, which is originally applied to the spatial dimension, is extended to the spectral dimension, thereby effectively solves the problem of CNN’s fixed convolutional kernel. Concurrently, to enhance the performance of the semi-supervised classifier, a consistency filter is proposed to screen unlabeled samples with high confidence. Experiments were carried out on the international common test datasets. The experimental results fully prove that the SRSN model proposed in this letter is superior to other methods and achieves 97.28%, 82.88%, and 89.13% OA of PaviaCenter, Houston2013, and WHU-Hi-HongHu datasets.