Change detection (CD), aims to detect the changing area of the same scene at different times, which is an important application of remote sensing images. As the key data source of CD, hyperspectral image (HSI) is widely used in CD technology because of its rich spectral-spatial information. However, how to mine the multi-level spatial information of dual-temporal hyperspectral images (HSIs) and focus on the features of the pixels to be classified individually remains a problem in the spatial attention mechanism (SAM). To make full use of the spectral-spatial information of HSIs, in this paper we propose a CNN framework with compact band weighting and multi-scale spatial attention (CBW-MSSANet) for HSI pixel-level CD. The main contributions of this article are as follows: 1) a new method of pseudo-label training sample selection based on k-means (KM) centroid distance is designed; 2) apply the compact band weighting (CBW) module to HSI CD to take full advantage of the spectral information of HSIs; 3) a multi-scale spatial attention (MSSA) module is developed for pixel-level CD, which can mine multi-level spatial information and pay more attention to the features of the pixels to be classified, and combine the spatial information of adjacent pixels to make it more conducive to pixel-level CD. Experimental results on four real HSI datasets demonstrated that the performance of MSSA surpasses the classical single-scale SAM, and CBW-MSSANet is superior to some representative CD methods.
Accurately recognizing the semantic categories of LR (low-resolution) aerial photos is an indispensable technique in remote sensing. In practice, however, this task is nontrivial due to: 1) the difficulty to encode human visual perception in the recognition process, 2) the intolerable human resources to label sufficient training LR aerial photos, and 3) the challenge to select high quality features for categorization. To handle these problems, a novel cross-resolution perceptual knowledge propagation (CPKP) is proposed, focusing on leveraging the visual perceptual experiences deeply learned from HR (high-resolution) aerial photos to enhance categorizing LR ones. Specifically, by mimicking human vision system, a novel low-rank model is proposed to decompose each LR aerial photo into multiple visually/semantically salient foreground regions coupled with the background non-salient ones. This model can 1) produce the gaze shifting path (GSP) simulating human gaze shifting sequence, and 2) calculate the deep feature for each GSP. Afterward, a kernel-induced feature selection (FS) algorithm is formulated to obtain a succinct set of deep GSP features discriminative across LR and HR aerial photos. Based on these, the labels from LR and HR aerial photos are collaboratively utilized to train a linear classifier for categorizing LR ones. Noticeably, our CPKP framework can effectively optimize the linear classifier training. This is because labels of HR aerial photos can be acquired more conveniently than LR ones practically. Comprehensive visualization results and comparative study have validated the superiority of our method.
For the classification of hyperspectral imagery (HSI), the convolutional neural network (CNN) can learn the discriminative spatial-spectral information of the image better than the traditional classification methods. However, when CNN uses the local receptive field to extract the features of HSI, it may cause the feature expression of the same pixel on the feature map to be inconsistent, and eventually cause noise in the classification results. To overcome this, we introduce the attention mechanism in the CNN model to improve the feature expressiveness. A spectral-spatial attention aggregation network (SSAAN) for HSI classification is designed, and there are two attention branches in our method. The spectral attention module with the squeeze-and-excitation (SESAM) automatically obtains the importance of each feature channel of HSI, and then enhances the useful band features and suppresses the less-useful band features according to this importance. In the spatial attention module with selective kernel (SKSAM), first, different convolution kernels of 2D-CNN are used to extract the shallow-middle-deep layer features from the principal components after dimension reduction, and the pixel spatial information from the three paths is combined and aggregated. Then, the feature maps of kernels of different sizes are aggregated according to the selection weights. Finally, the feature vectors obtained from the two branches of the spatial attention module and the spectral attention module are connected to further improve feature representation, and the classification result is obtained by the softmax function. Experimental results through three real HSI data sets show that our proposed method SSAAN achieves better performance compared to the state-of-the-art methods.
Superpixel segmentation has been demonstrated to be a powerful tool in hyperspectral image (HSI) classification. Each superpixel region can be regarded as a homogeneous region, which is composed of a series of spatial neighboring pixels. However, a superpixel region may contain the pixels from different classes. To further explore the optimal representations of superpixels, a new framework based on two k selection rules is proposed to find the most representative training and test samples. The proposed method consists of the following four steps: first, a superpixel segmentation algorithm is performed on the HSI to cluster the pixels with similar spectral features into the same superpixel. Then, a domain transform recursive filtering is used to extract the spectral-spatial features of the HSI. Next, the k nearest neighbor (KNN) method is utilized to select k 1 representative training samples and k 2 test pixels for each superpixel, which can effectively overcome the within-class variations and between-class interference, respectively. Finally, the class label of superpixels can be determined by measuring the averaged distances among the selected training and test samples. Experiments conducted on four real hyperspectral datasets show that the proposed method provides competitive classification performances with respect to several recently proposed spectral-spatial classification methods.
This letter presents the hyperspectral imagery (HSI) noisy label detection using a spectral angle and the local outlier factor (SALOF) algorithm. The noisy label is caused by a mislabeled training pixel, and thus, noisy training samples mixed with correct and incorrect labels are formed in the supervised classification. The LOF algorithm is first used in the noisy label detection of the HSI to improve the supervised classification accuracy. The proposed method SALOF mainly includes the following steps. First, k nearest neighbors of different training samples of each class are calculated based on the spectral angle mapper. Second, the reachability distance and local reachability density of all training samples are obtained. Third, the LOF is determined among different classes of training samples. Then, a segmentation threshold of the LOF is established to achieve an abnormal probability of these training samples. Finally, the support vector machines are applied to measure the detection efficiency of the proposed method. The experiments performed on the Kennedy Space Center data set demonstrate that the proposed method can effectively detect noisy labels.
In cross-domain hyperspectral image (HSI) classification, the labeled samples of the target domain are very limited, and it is a worthy attention to obtain sufficient class information from the source domain to categorize the target domain classes (both the same and new unseen classes). This article investigates this problem by employing few-shot learning (FSL) in a meta-learning paradigm. However, most existing cross-domain FSL methods extract statistical features based on convolutional neural networks (CNNs), which typically only consider the local spatial information among features, while ignoring the global information. To make up for these shortcomings, this article proposes novel convolutional transformer-based few-shot learning (CTFSL). Specifically, FSL is first performed in the classes of source and target domains simultaneously to build the consistent scenario. Then, a domain aligner is set up to map the source and target domains to the same dimensions. In addition, a convolutional transformer (CT) network is utilized to extract local-global features. Finally, a domain discriminator is executed subsequently that can not only reduce domain shift but also distinguish from which domain a feature originates. Experiments on three widely used hyperspectral image datasets indicate that the proposed CTFSL method is superior to the state-of-the-art cross-domain FSL methods and several typical HSI classification methods in terms of classification accuracy.
In this paper, a new spectral-spatial classification method based on texture pattern separation (TPS) is proposed for hyperspectral image (HSI) classification that consists of the following steps. First, the principal component analysis (PCA) method is used to reduce the dimensions of the original HSI. Then, the processed image is partitioned into several three-dimensional subcube images, each of which can be decomposed into a texture layer and background layer using joint convolutional analysis and synthesis sparse representation (JCAS) to reflect meaningful information and disturbing information, respectively. Next, texture layer images are fed into different kernel pixelwise classifiers for classification. Finally, majority voting is utilized to obtain the final classification result. Through comparisons with other well-known classification methods, the proposed TPS method shows outstanding performance, even with a quite limited number of training samples.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)