The Collaborative Representation-based Detector (CRD) performs anomaly detection for hyperspectral data using a linear representation of local neighbors for background estimation, which may not fully capture the informational content and spectral variability in complex hyperspectral images with heterogenous background. To deal with this aspect, the Collaborative Representation-based Attention Network (CRAN) is introduced in this letter, providing a nonlinear representation of data samples for background estimation. Both local neighbors and global samples are used in parallel, and their outputs are fused through a cross-attention mechanism. Experimental results show a good performance of CRAN in comparison with several state-of-the-art anomaly detectors.

Abstract Polarimetric synthetic aperture radar (POLSAR) systems provide high resolution images containing polarimetric information. So, they have high capability in land cover classification. In this work, a binary coding‐based polarimetric‐morphological (BCPM) feature extraction is proposed for POLSAR image classification. At first, a set of polarimetric features is proposed. Then, a new morphological framework is introduced for contextual feature extraction from the POLSAR cube. The coherence matrix is composed from diagonal and non‐diagonal elements with different information. These elements are analysed separately in the proposed method. Moreover, the amplitude and phase components of the non‐diagonal elements are individually analysed using morphological filters by reconstruction. Finally, a binary coding‐based polarimetric‐spatial feature reduction, which uses the first order statistics, is proposed for feature transformation. The experiments on three real POLSAR images and a synthetic dataset show the superior performance of BCPM compared to several classification methods.

Incorporation of spatial information besides rich spectral information of hyperspectral image significantly enhances data classification accuracy. A morphology‐based feature extraction and classification framework is proposed here, which includes the local neighbourhood information in a spatial window for extension of training set. The proposed method is morphology‐based structure‐preserving projection (MSPP) and tries to preserve the data structure in spectral–spatial feature space. Moreover, MSPP increases the class discrimination ability by defining a similarity matrix constructed by extended spectral–spatial training samples. The experimental results show the superiority of MSPP compared to some state‐of‐the‐art classification methods from the classification accuracy point of view.

A hyperspectral anomaly detector, which uses the benefits of a proposed background feature extraction method and the kernel trick, is introduced in this study. The proposed method is differential image based detector (DID). The DID method uses the differential image to estimate the variations of background in the feature extraction phase. The proposed feature extraction method slows the variations of background. So, it suppresses the background signal and highlights the anomalous signal. The separation between anomalous targets and background clutter is increased in the projected feature space. Before feeding the transformed data into the Reed–Xiaoli (RX) detector, the kernel learning method is applied. The kernel technique transforms the projected data with a linear non‐Gaussian model to a potentially high‐dimensional feature space with the non‐linear Gaussian domain. Experiments are conducted on two real hyperspectral images. The experimental results indicate that the proposed DID method outperforms RX and some state‐of‐the‐art anomaly detection approaches.

Morphological profiles (MPs) are efficiently exploited for modelling the geometrical features of structures in a scene. They increase the discriminability between different classes. The degree of processing of images depends on the geometrical structure and shape of the used structure element (SE) in the transformation. Since the geometric structures of an image are not the same in the whole image, the use of a fixed shape for SE may not be so efficient. Thus, it is proposed to extract an edge patch image‐based morphological profile (EPIMP), which considers SEs with different shapes for different areas of image. The used SE in each patch of image is corresponding to the shape (i.e. edge image) of that patch. The proposed method is experimented on both multispectral and hyperspectral images and the obtained results show that the proposed method is much more efficient than the conventional MPs. Moreover, the experiments show the superiority of EPIMP compared with some state‐of‐the‐art spectral‐spatial classification methods such as generalised composite kernel, multiple feature learning, weighted joint collaborative representation and multiple‐structure‐element non‐linear multiple kernel learning.

An improved version of Reed-Xiaoli (RX) detector is proposed in this letter, which uses the benefits of median-mean line (MML) metric. The background data may be contaminated by anomalies. The anomalous outliers contributed in the estimate of background statistics decrease the differences between anomalous targets and background clutter. So, the performance of an RX detector is degraded. To deal with the negative effects of anomalous outliers, and to rectify the position of background data, the MML metric is used for providing more reliable background samples. Therefore, more stable background statistics (mean and covariance matrix) are estimated. The experimental results show the better performance of the proposed MML-RX method compared with some state-of-the-art anomaly detection methods with reasonable computation time.

Representation-based classifiers such as nearest regularized subspace (NRS) have been recently developed for hyperspectral image classification. The joint collaborative representation (JCR) and the weighted JCR (WJCR) methods added spatial information to the pixel-wise NRS classifier. While JCR adopts the same weights for extraction of spatial features from the surrounding pixels, WJCR uses the similarity between the central pixel and its surroundings to assign different weights to neighbor pixels. Two improved versions of WJCR are introduced in this paper. The first method, WJCR based on median-mean line, is proposed to cope with the negative effect of outlying neighbors. The second method, WJCR based on angular separation (AS), uses the benefits of the AS measurement to decrease the contribution of redundant information due to the highly correlated neighbors. The experimental results on some real hyperspectral data sets show the good efficiency of the proposed methods compared to other state-of-the-art NRS-based classifiers.

Feature extraction using weighted training (FEWT) samples is proposed in this letter. Different spectral bands (features) play different roles in identification of land-cover classes. In the FEWT, the relative importance of each feature of a training sample in predicting the class label of that sample is obtained and considered as a weight for that feature. Then, the weighted training samples can be used in each arbitrary feature extraction method. In this letter, we use the weighted training samples in supervised locality preserving projection. The experimental results on three popular hyperspectral images show that FEWT has better performance and more speed than some state-of-the-art supervised feature extraction methods using limited number of available training samples.

Feature extraction plays a central role in classification of hyperspectral data. We propose a clustering-based feature extraction (CBFE) method in this letter. The proposed method is supervised and only needs to calculate the first-order statistics. Thus, CBFE has better performance than some popular supervised feature extraction methods such as linear discriminant analysis, generalized discriminant analysis, and nonparametric weighted feature extraction in small sample size situation. In addition, CBFE works better than unsupervised approaches such as principal component analysis in classification applications. CBFE considers a vector associated with each band that is composed by the mean values of all classes in that band. Then, a clustering method such as k-means is run to group the similar bands in one cluster. The selected number of clusters is equal to the number of extracted features. Experiments carried out on two different hyperspectral data sets demonstrate that the CBFE has better performance in comparison with some conventional feature extraction methods.

Hyperspectral images provide a large volume of spectral bands. Feature extraction (FE) is an important preprocessing step for classification of high-dimensional data. Supervised FE methods such as linear discriminant analysis, generalized discriminant analysis, and nonparametric weighted FE use the criteria of class separability. Theses methods maximize the between-class scatter matrix and minimize the within-class scatter matrix. We propose a supervised FE method in this letter, which uses no statistical moments. Thus, it works well using limited training samples. The proposed FE method consists of two important phases. In the first phase, an attraction point for each class is found. In the second phase, by using an appropriate transformation, the samples of each class move toward the attraction point of their class. The experimental results on two real hyperspectral images demonstrate that FE using attraction points has better performance in comparison with some other supervised FE methods in a small sample size situation.