Hyperspectral anomaly detection: a performance comparison of existing techniques

Whether or not a hyperspectral anomaly detector is effective is determined by two crucial issues, anomaly detectability and background suppressibility (BS), both of which are very closely related to two factors, the datasets used for a selected hyperspectral anomaly detector and detection measures used for its performance evaluation. This paper explores how anomaly detectability and BS play key roles in hyperspectral anomaly detection (HAD). To address these two issues, we investigate three key elements attributed to HAD. One is a selected hyperspectral anomaly detector, and another is the datasets used for experiments. The third one is the detection measures used to evaluate the effectiveness of a hyperspectral anomaly detector. As for hyperspectral anomaly detectors, twelve commonly used anomaly detectors were evaluated and compared. To address the appropriate use of datasets for HAD, seven popular and widely used datasets were studied for HAD. As for the third issue, the traditional area under a receiver operating characteristic (ROC) curve of detection probability—PD versus false alarm probability, PF, (AUC(D,F))—was extended to 3D ROC analysis where a 3D ROC curve was developed to generate three 2D ROC curves from which eight detection measures could be derived to evaluate HAD in all round aspects, including anomaly detectability, BS and joint anomaly detectability and BS. Qualitative analysis showed that many works reported in the literature which claimed that their developed hyperspectral anomaly detectors performed better than other anomaly detectors are actually not true because they overlooked these two issues. Specifically, a comprehensive study via extensive experiments demonstrated that these 3D ROC curve-derived detection measures can be further used to address the various characterizations of different data scenes and also to provide explanations as to why certain data scenes are not suitable for HAD.

Hyperspectral imagery systems have the ability to collect 3D digital images with rich spatial and spectral information. Anomaly detection is one of the interesting applications over last two decades in hyperspectral imagery. In this paper, we propose Collaborative Representation-Based with Outlier Removal Anomaly Detector (CRBORAD) method for HSI Anomaly Detection. We use both spectral and spatial information for detecting anomalies instead of using only spectral information that was introduced in our previous work. The proposed detector can adaptively estimate the background by its adjacent pixels within a sliding dual window. Before estimating background pixels, we remove outlier pixels that are significantly different from majority of pixels. It leads us to precise background approximation and better accuracy for detecting anomalies in subsequent stages. The residual image is constituted by subtracting the predicted background from the original HSI, and anomalies can be determined in the residual image, finally. Kernel extension of the proposed approach is also presented. We implemented the proposed algorithms on San Diego airport hyperspectral data. CRBORAD results are illustrated using receiver-operating-characteristic (ROC) curves, Area Under Curve (AUC) values and intuitive images. Comparing the results of the current study with four popular and previous methods shows that CRBORAD provides us an accurate method for detecting anomalies.

In this paper, collaborative representation is proposed for anomaly detection in hyperspectral imagery. The algorithm is directly based on the concept that each pixel in background can be approximately represented by its spatial neighborhoods, while anomalies cannot. The representation is assumed to be the linear combination of neighboring pixels, and the collaboration of representation is reinforced by l <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> -norm minimization of the representation weight vector. To adjust the contribution of each neighboring pixel, a distance-weighted regularization matrix is included in the optimization problem, which has a simple and closed-form solution. By imposing the sum-to-one constraint to the weight vector, the stability of the solution can be enhanced. The major advantage of the proposed algorithm is the capability of adaptively modeling the background even when anomalous pixels are involved. A kernel extension of the proposed approach is also studied. Experimental results indicate that our proposed detector may outperform the traditional detection methods such as the classic Reed-Xiaoli (RX) algorithm, the kernel RX algorithm, and the state-of-the-art robust principal component analysis based and sparse-representation-based anomaly detectors, with low computational cost.

Anomaly detection is one of the most important applications of hyperspectral imagery. In this paper, we propose a hyperspectral image anomaly detection method using collaborative representation detector with PCA remove outlier (PCAroCRD). The algorithm has two versions: global and local. In the collaborative representation detection (CRD) algorithm, if the test pixel is an anomalous pixel and several samples from background are also anomalous and similar to test pixel, then the judge error is likely to happen. In the proposed algorithm, PCA is adopted to extract main pixels information and remove abnormal pixels from background so as to obtain higher detection accuracy before estimating the samples. Experimental results indicate that the proposed algorithm outperforms the original collaborative representation detector (CRD), kernel version of CRD (KCRD), advanced CRD (CRDBORAD), and the traditional detection methods such as Global Reed-Xiaoli (RX) algorithm and the Local RX.

Recently, collaborative representation detector (CRD) has been popularly used for hyperspectral anomaly detection. For the original CRD, the least squares solution becomes more unstable when more classes, i.e., samples for anomaly detection are involved, and the detection error is likely to happen if the test pixel is an anomalous pixel and several samples from background are similar anomalous. In this paper, we propose a hyperspectral anomaly detection method that uses CRD with principal component analysis (PCA) for removing outlier (PCAroCRD). According to the different background modeling methods, global and local versions are proposed. In the proposed algorithm, the spatial-domain PCA is adopted to extract main pixel information of global/local background that will be used as samples for collaborative representation, and simultaneously the information of abnormal pixels in global/local background can be removed. Fewer useful samples can also keep the detection result stable. Experimental results indicate that the PCAroCRD outperforms the original CRD, kernel version of CRD, advanced CRD (CRDBORAD), morphology-based CRD, Global Reed–Xiaoli (RX) algorithm, and the Local RX.

In this paper we present a new methodology for automated target detection and identification in hyperspectral imagery. The standard paradigm for target detection in hyperspectral imagery is to run a detection algorithm, typically statistical in nature, and visually inspect each high-scoring pixel to decide whether it is a true detection or a false alarm. Detection filters have constant false alarm rates (CFARs) approaching 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-5</sup> , but these can still result in a large number of false alarms given multiple images and a large number of target materials. Here we introduce a new methodology for target detection and identification in hyperspectral imagery in conjunction with a data-driven target detection algorithm that shows promise for hard targets. The result is a greatly reduced false alarm rate and a practical methodology for aiding an analyst in quantitatively evaluating detected pixels.

Collaborative representation-based detector (CRD), as the most representative anomaly detection method, has been widely applied in the field of hyperspectral anomaly detection (HAD). However, the sliding dual window of the original CRD introduces high computational complexity. Moreover, most HAD models only consider a single spectral or spatial feature of the hyperspectral image (HSI), which is unhelpful for improving detection accuracy. To solve these problems, in terms of speed and accuracy, we propose a novel anomaly detection approach, named Random Collective Representation-based Detector with Multiple Feature (RCRDMF). This method includes the following steps. This method first extract the different features include spectral feature, Gabor feature, extended multiattribute profile (EMAP) feature, and extended morphological profile (EMP) feature matrix from the HSI image, which enables us to improve the accuracy of HAD by combining the multiple spectral and spatial features. The ensemble and random collaborative representation detector (ERCRD) method is then applied, which can improve the anomaly detection speed. Finally, an adaptive weight approach is proposed to calculate the weight for each feature. Experimental results on six hyperspectral datasets demonstrate that the proposed approach has the superiority over accuracy and speed.

Anomaly detection in hyperspectral imaging is crucial for remote sensing, driving the development of numerous algorithms. However, systematic studies reveal a dichotomy where algorithms generally excel at either detecting anomalies in specific datasets or generalizing across heterogeneous datasets (i.e., lack adaptability). A key source of this dichotomy may center on the singular and like biases frequently employed by existing algorithms. Current research lacks experimentation into how integrating insights from diverse biases might counteract problems in singularly biased approaches. Addressing this gap, we propose stacking-based ensemble learning for hyperspectral anomaly detection (SELHAD). SELHAD introduces the integration of hyperspectral anomaly detection algorithms with diverse biases (e.g., Gaussian, density, partition) into a singular ensemble learning model and learns the factor to which each bias should contribute so anomaly detection performance is optimized. Additionally, it introduces bootstrapping strategies into hyperspectral anomaly detection algorithms to further increase robustness. We focused on five representative algorithms embodying common biases in hyperspectral anomaly detection and demonstrated how they result in the previously highlighted dichotomy. Subsequently, we demonstrated how SELHAD learns the interplay between these biases, enabling their collaborative utilization. In doing so, SELHAD transcends the limitations inherent in individual biases, thereby alleviating the dichotomy and advancing toward more adaptable solutions.

Although hyperspectral anomaly detection is commonly conducted in the visible, near-infrared, and shortwave infrared spectral regions, there has been less research on hyperspectral anomaly detection in the longwave infrared (LWIR) hyperspectral region. The radiance of thermal infrared hyperspectral imagery is determined by the temperature and emissivity. To avoid the detection uncertainty caused by the single factor of temperature, emissivity can be introduced to detect anomalies. However, in the emissivity domain, the spectral contrast and signal-to-noise ratio (SNR) are low, which makes it difficult to separate the anomalies from the background. In this paper, an anomaly detection method combining emissivity and a segmented low-rank prior (EaSLRP) is proposed for use with thermal infrared hyperspectral imagery. The EaSLRP method is divided into three parts—1) temperature/emissivity retrieval, 2) extraction of the thermal infrared hyperspectral background information, and 3) Mahalanobis distance detection. A homogeneous region generation method is also proposed to solve the problem of the complex global background leading to inaccurate background estimation. The GoDec method is used for matrix decomposition and background information extraction and to remove some of the noise. The proposed Mahalanobis distance detector then uses the background component and original image for anomaly detection, while highlighting the spectral difference between the anomalies and background. This method can also suppress the influence of noise, to some extent. The experimental results obtained with airborne Fourier transform thermal infrared spectrometer hyperspectral images demonstrate that the EaSLRP method is effective when compared with the Reed–Xiaoli detector (RXD), the segmented RX detector (SegRX), the low-rank and sparse representation-based detector (LRASR), the low-rank and sparse matrix decomposition (LRaSMD)-based Mahalanobis distance method (LSMAD), and the locally enhanced low-rank prior method (LELRP-AD).

Lately, generative adversarial networks (GAN)-based methods have drawn extensive attention and achieved a promising performance in the field of hyperspectral anomaly detection (HAD) owing to GAN’s powerful data generation capability. However, without considering the background spatial features, most of these methods can not obtain a GAN with a strong background generation ability. Besides, they fail to address the hyperspectral image (HSI) redundant information disturbance problem in the anomaly detection part. To solve these issues, the unsupervised generative adversarial network with background spatial feature enhancement and irredundant pooling (BEGAIP) is proposed for HAD. To make better use of features, spatial and spectral features union extraction idea is also applied to the proposed model. To be specific, in spatial branch, a new background spatial feature enhancement way is proposed to get a data set containing relatively pure background information to train GAN and reconstruct a more vivid background image. In a spectral branch, irredundant pooling (IP) is invented to remove redundant information, which can also enhance the background spectral feature. Finally, the features obtained from the spectral and spatial branch are combined for HAD. The experimental results conducted on several HSI data sets display that the model proposed acquire a better performance than other relevant algorithms.

The processing chain leading to specific material detection in hyperspectral imagery implies the use of atmospherically corrected images of emissivity or reflectance before comparing image signatures to a database of materials' signatures. This is a sensible approach for the reflective hyperspectral bands l and when the pixels are completely filled with a uniform material in the LWIR bands (8 to 12 microns). In the LWIR, the atmospheric correction process is different of what is used in the reflective bands and involves the use of a temperature and emissivity separation process (TES). If the pixel is not filled with a uniform material and the measured radiance is produced from the mix of materials having different emissivity and temperatures, the output of the TES will not be linear in temperature and in emissivity and will be contaminated by the non-linear mix of the temperature and emissivity of the materials leading to a potential for confusion during the detection process. In this paper, we propose a detection approach using the ground leaving radiance that is used directly to perform detection using emissivity signatures contained in a database. The detection results using this process are compared with the detection results using the output of a TES algorithm. The study is performed in simulation without noise and with the exact knowledge of the downwelling irradiance. The results show that a detection algorithm using the ground leaving radiance performs better than its counterpart using the emissivity when the difference in temperature increases.

An Iterative spectral-spatial hyperspectral anomaly detection (ISSHAD) was recently developed to augment the data cube to be processed to construct a new data cube via a feedback process in an iterative manner. As a result of such a feedback process hyperspectral anomaly detection (ISSHAD) significantly improve HAD performance. As an opposite approach to ISSHAD, this paper develops a new approach to HAD, called cascaded spectral-spatial hyperspectral anomaly detection (CSSHAD), which performs HAD by a cascade of anomaly detection modules (ADMs) in a feedforward network in which each cascaded anomaly detection module (ADM) implements the same anomaly detector to produce an anomaly detection map (ADMap). Then a spatial filter is designed to capture spatial information from the ADMap to generate a real-valued spatial filtered ADMap (SF-Map) which will be used as the input to the next following ADM. Finally, the depth of CSSHAD is determined by an automatic stopping rule. To further improve anomaly detectability and background suppressibility, a recent developed concept, effective anomaly space (EAS) is incorporated into CSSHAD to derive EAS-CSSHAD. Unlike ISSHAD which is a feedback HAD network, CSSHAD is a feedforward HAD network that operates a cascade of ADMs by feeding forward SFMaps in a similar manner that ISSHAD uses an iterative feedback process. Accordingly, CSSHAD and ISSHAD can be considered to dual HAD methods as companion methods. Experimental results also demonstrate that CSSHAD performs comparably to ISSHAD with less computational complexity due to its use of feedforward structure rather than feedback process.

In this paper we examine the efficacy of using the closest distance to center algorithm in conjunction with ellipsoidal multivariate trimming (MVT) to find outliers in a hyperspectral image. MVT is applied here as a global anomaly detector on images that are pre-processed into clusters using a technique called X-means. Under the assumption that there are no more than 5% outliers in any given cluster set, we develop a method, based upon principal component analysis pre-processing, to create a flexible threshold for determining the percentage of data to retain with MVT. Using a retention percentage that more adequately reflects the actual number of outlier-free observations allows one to form estimates of the mean and covariance matrix that more effectively decrease the effects of swamping and masking as compared to using a set percentile for retention. These ideas are tested against real and synthetically generated hyperspectral imagery.

Anomaly detection is a fundamental task in hyperspectral image (HSI) processing. However, most existing methods rely on pixel feature vectors and overlook the relational structure information between pixels, limiting the detection performance. In this article, we propose a novel approach to hyperspectral anomaly detection that characterizes the HSI data using a vertex- and edge-weighted graph with the pixels as vertices. The constructed graph encodes rich structural information in an affinity matrix. A crucial innovation of our method is the ability to obtain internal relations between pixels at multiple topological scales by processing different powers of the affinity matrix. This power processing is viewed as a graph evolution, which enables anomaly detection using vertex extraction formulated as a quadratic programming problem on graphs of varying topological scales. We also design a hierarchical guided filtering architecture to fuse multiscale detection results derived from graph evolution, which significantly reduces the false alarm rate. Our approach effectively characterizes the topological properties of HSIs, leveraging the structural information between pixels to improve anomaly detection accuracy. Experimental results on four real HSIs demonstrate the superior detection performance of our proposed approach compared to some state-of-the-art hyperspectral anomaly detection methods.

In this paper, we introduce a new approach to address the challenge of generalization in hyperspectral anomaly detection (AD). Our method eliminates the need for adjusting parameters or retraining on new test scenes as required by most existing methods. Employing an image-level training paradigm, we achieve a general anomaly enhancement network for hyperspectral AD that only needs to be trained once. Trained on a set of anomaly-free hyperspectral images with random masks, our network can learn the spatial context characteristics between anomalies and background in an unsupervised way. Additionally, a plug-and-play model selection module is proposed to search for a spatial-spectral transform domain that is more suitable for AD task than the original data. To establish a unified benchmark to comprehensive evaluate our method and existing methods, we develop a large-scale hyperspectral AD dataset (HAD100) that includes 100 real test scenes with diverse anomaly targets. In comparison experiments, we combine our network with a parameter-free detector, and achieve the optimal balance between detection accuracy and inference speed among state-of-the-art AD methods. Experimental results also show that our method still achieves competitive performance when the training and test set are captured by different sensor devices. Our code is available at https://github.com/ZhaoxuLi123/AETNet.