Background Subspace Research Articles

Target detection in hyperspectral imaging combining replacement and additive models

In hyperspectral imaging the replacement model where a target, if present, partly replaces the disturbance is often advocated. In this paper, we consider a somehow more realistic model where only the low-rank background is substituted for the target while a residual noise, which belongs to the orthogonal complement, is unaffected by the presence/absence of the target. A two-step generalized likelkihood ratio test is formulated for such a model. Furthermore we show that the log likelihood can be well approximated by a weighted combination of the log likelihoods of the FTMF and the AMF, and that the dimension of the background subspace is the tuning parameter which enables to balance between these two well-known detectors. A comparison with standard techniques on real hyperspectral data reveals a good performance of the new detectors.

Data-augmented matched subspace detector for hyperspectral subpixel target detection

The performance of subspace-based methods such as matched subspace detector (MSD) and MSD with interaction effects (MSDinter) heavily depends on the background subspace and the target subspace. Nonetheless, constructing a representative target subspace is challenging due to the limited availability of target spectra in a collected hyperspectral image. In this paper, we propose two new hyperspectral target detection methods termed data-augmented MSD (DAMSD) and data-augmented MSDinter (DAMSDI) that can effectively solve the scarcity problem of target spectra and from which a representative target-background mixed subspace can be learned. We first synthesise target-background mixed spectra based on classical hyperspectral mixing models and then learn a target-background mixed subspace via principal component analysis. Compared with MSD and MSDinter, the learned mixed subspace is more representative as spectral variability of target spectra is explained to the largest extent and it leads to an improvement in computational speed and numerical stability. We demonstrate the efficacy of DAMSD and DAMSDI for subpixel target detection on two public hyperspectral image datasets.

A Hyperspectral Subspace Target Detection Method Based on AMUSE

Target detection is an important research topic in hyperspectral image processing, and has been the focus of research of many scholars. Compared with many target detection methods, the subspace target detection method is superior. This paper proposes a hyperspectral subspace target detection method LBSE-AMUSE based on the Local Background Subspace Estimation (LBSE) method and the AMUSE algorithm (Algorithm for Multiple Signals Extraction). HyMap airborne hyperspectral remote sensing data is used as the data source, and the performance between the LBSE and LBSE-AMUSE methods is compared by using quantitative indicators such as false alarm rate, detection rate and ROC curve. The experimental results show that the LBSE-AMUSE method has a simple structure, short running time and low false alarm rate compared with the LBSE method, which can suppress the background highlighting target to some extent.

Attribute profile based target detection using collaborative and sparse representation

Abstract Two hyperspectral target detection methods are introduced in this paper. The proposed methods use the spatial information contained in attribute profiles (APs) in addition to the original spectral information. The first detector is AP based collaborative representation (AP-CR) and the second one is AP based sparse representation (AP-SR). Since the thinning operators extract the details of image, the spatial features extracted by them are used to compose the target subspace. In contrast, since the thickening operators conduct the image details to be similar to the surrounding background, they are used for extraction of spatial features composing the background subspace. The proposed AP-CR and AP-SR methods, by generating two appropriate spectral–spatial subspaces, individually considered for target and background dictionaries, show a superior performance in several popular hyperspectral data from the detection probability and the running time point of views.

IBRS: An Iterative Background Reconstruction and Suppression Framework for Hyperspectral Target Detection

With the development of hyperspectral imaging technology and detection algorithm research, hyperspectral target detection has become indispensable in the field of target recognition and reconnaissance. But limited by the complex distribution of land covers and the difficulties to distinguish spectral characteristics between the target and its surroundings, existing target detection algorithms may still be insufficient on identifying target spectrum and suppressing background spectra. In this paper, we propose a novel iterative background reconstruction and suppression framework for hyperspectral target detection. By means of iteratively integrating the dictionary training algorithm with sparse-coding method to estimate the background training dictionary and its corresponding sparse representation, hyperspectral imagery (HSI) dataset is purified into background subspace matrix. Then, we add the constructed background suppression regularization term on the constrained energy minimization model to form the framework as a linearly background constrained optimization filter model. The optimal solution is then solved by a convex optimization interior-point method and improved by an iteratively reweighting method to achieve the goal of separating the target and the background. The proposed algorithm is applied to two synthetic hyperspectral datasets and two real-HSI datasets; the experimental results show that our algorithm performs a superior detection result compared to other classical hyperspectral target detection algorithms. It also provides a fresheye idea to integrate a variety of theoretical algorithms to explore the way of constructing and solving target detection model.

An online background subtraction algorithm deployed on a NAO humanoid robot based monitoring system

In this paper, we design a fast background subtraction algorithm and deploy this algorithm on a monitoring system based on NAO humanoid robot. The proposed algorithm detects a contiguous foreground via a contiguously weighted linear regression (CWLR) model. It consists of a background model and a foreground model. The background model is a regression based low rank model. It seeks a low rank background subspace and represents the background as the linear combination of the basis spanning the subspace. The foreground model promotes the contiguity in the foreground detection. It encourages the foreground to be detected as whole regions rather than separated pixels. We formulate the background and foreground model into a contiguously weighted linear regression problem. This problem can be solved efficiently via an alternating optimization approach which includes continuous and discrete variables. Given an image sequence, we use the first few frames to incrementally initialize the background subspace, and we determine the background and foreground in the following frames in an online scheme using the proposed CWLR model, with the background subspace continuously updated using the detected background information. The proposed algorithm is implemented by Python on a NAO humanoid robot based monitoring system. This system consists of a control station and a Nao robot. The Nao robot acts as a mobile probe. It captures image sequence and sends it to the control station. The control station serves as a control terminal. It sends commands to control the behavior of Nao robot, and it processes the image data sent by Nao. This system can be used for living environment monitoring and form the basis for many vision-based applications like fall detection and scene understanding. The experimental comparisons with most recent algorithms on both benchmark dataset and NAO captures demonstrate the high effectiveness of the proposed algorithm.

Hyperspectral anomaly detection based on constrained eigenvalue–eigenvector model

Anomaly detection in a large area using hyperspectral imaging is an important application in real-time remote sensing. Anomaly detectors based on subspace models are suitable for such an anomaly and usually assume the main background subspace and its dimensions are known. These detectors can detect the anomaly for a range of values of the dimension of the subspace. The objective of this paper is to develop an anomaly detector that extends this range of values by assuming main background subspace with an unknown user-specified dimension and by imposing covariance of error to be a diagonal matrix. A pixel from the image is modeled as the sum of a linear combination of the unknown main background subspace and an unknown error. By having more unknown quantities, there are more degrees of freedom to find a better way to fit data to the model. By having a diagonal matrix for the covariance of the error, the error components become uncorrelated. The coefficients of the linear combination are unknown, but are solved by using a maximum likelihood estimation. Experimental results using real hyperspectral images show that the anomaly detector can detect the anomaly for a significantly larger range of values for the dimension of the subspace than conventional anomaly detectors.

A Sparse Representation-Based Binary Hypothesis Model for Target Detection in Hyperspectral Images

In this paper, a new sparse representation-based binary hypothesis (SRBBH) model for hyperspectral target detection is proposed. The proposed approach relies on the binary hypothesis model of an unknown sample induced by sparse representation. The sample can be sparsely represented by the training samples from the background-only dictionary under the null hypothesis and the training samples from the target and background dictionary under the alternative hypothesis. The sparse vectors in the model can be recovered by a greedy algorithm, and the same sparsity levels are employed for both hypotheses. Thus, the recovery process leads to a competition between the background-only subspace and the target and background subspace, which are directly represented by the different hypotheses. The detection decision can be made by comparing the reconstruction residuals under the different hypotheses. Extensive experiments were carried out on hyperspectral images, which reveal that the SRBBH model shows an outstanding detection performance.

Target detection based on a dynamic subspace

For hyperspectral target detection, it is usually the case that only part of the targets pixels can be used as target signatures, so can we use them to construct the most proper background subspace for detecting all the probable targets? In this paper, a dynamic subspace detection (DSD) method which establishes a multiple detection framework is proposed. In each detection procedure, blocks of pixels are calculated by the random selection and the succeeding detection performance distribution analysis. Manifold analysis is further used to eliminate the probable anomalous pixels and purify the subspace datasets, and the remaining pixels construct the subspace for each detection procedure. The final detection results are then enhanced by the fusion of target occurrence frequencies in all the detection procedures. Experiments with both synthetic and real hyperspectral images (HSI) evaluate the validation of our proposed DSD method by using several different state-of-the-art methods as the basic detectors. With several other single detectors and multiple detection methods as comparable methods, improved receiver operating characteristic curves and better separability between targets and backgrounds by the DSD methods are illustrated. The DSD methods also perform well with the covariance-based detectors, showing their efficiency in selecting covariance information for detection.

A Subspace-Based Change Detection Method for Hyperspectral Images

Remote sensing change detection has played an important role in many applications. Most traditional change detection methods deal with single-band or multispectral remote sensing images. Hyperspectral remote sensing images offer more detailed information on spectral changes so as to present promising change detection performance. The challenge is how to take advantage of the spectral information at such a high dimension. In this paper, we propose a subspace-based change detection (SCD) method for hyperspectral images. Instead of dealing with band-wise changes, the proposed method measures spectral changes. SCD regards the observed pixel in the image of Time 2 as target and constructs the background subspace using the corresponding pixel in the image of Time 1, and additional information. In this paper, two types of additional information, i.e., spatial information in the neighborhood of the corresponding pixel in Time 1, and the spectral information of undesired land-cover types, are used to construct the background subspace for special applications. The subspace distance is calculated to determine whether the target is anomalous with respect to the background subspace. The anomalous pixels are considered as changes. Here, orthogonal subspace projection is employed to calculate the subspace distance, which makes full use of the advantage of the abundant spectral information in hyperspectral imagery, and is also easy to apply. The experimental results using Hyperion data and HJ-1A HSI data indicate that SCD gives more accurate detection results, with a lower false alarm rate, compared with other state-of-the-art methods. SCD with additional information also gives satisfactory results in the experiments, reducing the false alarms caused by misregistration and suppressing the change of undesired land-cover types.

Background modeling by subspace learning on spatio-temporal patches

This paper presents a novel background model for video surveillance—Spatio-Temporal Patch based Background Modeling (STPBM). We use spatio-temporal patches, called bricks, to characterize both the appearance and motion information. Our method is based on the observation that all the background bricks at a given location under all possible lighting conditions lie in a low dimensional background subspace, while bricks with moving foreground are widely distributed outside. An efficient online subspace learning method is presented to capture the subspace, which is able to model the illumination changes more robustly than traditional pixel-wise or block-wise methods. Experimental results demonstrate that the proposed method is insensitive to drastic illumination changes yet capable of detecting dim foreground objects under low contrast. Moreover, it outperforms the state-of-the-art in various challenging scenes with illumination changes.

Net analyte signal with floating reference theory in non-invasive blood glucose sensing by near-infrared spectroscopy

Based on the floating reference theory, a new method for extracting the net analyte signal (NAS) is proposed. The noise background subspace is spanned by spectra at the floating radial reference point, and then, the spectra at the measurement point are projected on the subspace. Thereafter, the glucose concentrations in intralipid solutions are investigated through Monte Carlo simulation and experiments, and the partial least squares (PLS) models with and without NAS analysis are built. The root mean square errors of calibration and prediction reach to 28.87% and 27.33%, respectively. The results confirm the existence of information induced by glucose concentration variations as well as the validity of the floating reference theory.

An Automatic Approach to Adaptive Local Background Estimation and Suppression in Hyperspectral Target Detection

This paper deals with subspace-based target detection in hyperspectral images. Specifically, it focuses on a general detection scheme where, first, background is suppressed through orthogonal-subspace projection and then target detection is accomplished. An adequate estimation of the background subspace is essential to a successful outcome. The background subspace has been typically estimated globally. However, global approaches may be ineffective for small-target-detection applications since they tend to overestimate the background interference affecting a given target. This may result in a low target residual energy after background suppression that is detrimental to detection performance. In this paper, we propose a novel and fully automatic algorithm for local background-subspace estimation (LBSE). Local background has typically a lower inherent complexity than that of global background. By estimating the background subspace over a local neighborhood of the test pixel, the resulting background-subspace dimension is expected to be low, thus resulting in a higher target residual energy after suppression which benefits the detection performance. Specifically, the proposed LBSE acts on a per-pixel basis, thus adaptively tailoring the estimated basis to the local complexity of background. Both simulated and real hyperspectral data are employed to investigate the detection-performance improvements offered by LBSE with respect to both global and local methodologies previously presented.

SIGNAL PROCESSING ALGORITHMS FOR STARING SINGLE PIXEL HYPERSPECTRAL SENSORS

Remote sensing of chemical warfare agents (CWA) with stand-off hyperspectral sensors has a wide range of civilian and military applications. These sensors exploit the spectral changes in the ambient photon flux produced thermal emission or absorption after passage through a region containing the CWA cloud. In this work we focus on (a) staring single-pixel sensors that sample their field of view at regular intervals of time to produce a time series of spectra and (b) scanning single or multiple pixel sensors that sample their FOV as they scan. The main objective of signal processing algorithms is to determine if and when a CWA enters the FOV of the sensor. We shall first develop and evaluate algorithms for staring sensors following two different approaches. First, we will assume that no threat information is available and we design an adaptive anomaly detection algorithm to detect a statistically-significant change in the observed spectrum. The algorithm processes the observed spectra sequentially-in-time, estimates adaptively the background, and checks whether the next spectrum differs significantly from the background based on the Mahalanobis distance or the distance from the background subspace. In the second approach, we will assume that we know the spectral signature of the CWA and develop sequential-in-time adaptive matched filter detectors. In both cases, we assume that the sensor starts its operation before the release of the CWA; otherwise, staring at a nearby CWA-free area is required for background estimation. Experimental evaluation and comparison of the proposed algorithms is accomplished using data from a long-wave infrared (LWIR) Fourier transform spectrometer.

Object localization in the presence of a strong heterogeneous background in fluorescent tomography

We propose a method for object localization in fluorescent tomography (FT) in the presence of a highly heterogeneous background. Existing approaches typically assume a homogeneous background distribution; thus, they are incapable of accurately accounting for the more general case of an unconstrained, possibly heterogeneous, background. The proposed method iteratively solves the inverse problem over a solution space partitioned into a background subspace and an object subspace to simultaneously estimate the background and localize the target fluorescent objects. Simulation results of this algorithm applied to continuous-wave FT demonstrate effective localization of target objects in the presence of highly heterogeneous background distributions.

A REVIEW OF INTEGRABLE DEFORMATIONS IN AdS/CFT

Marginal β deformations of [Formula: see text] super-Yang–Mills theory are known to correspond to a certain class of deformations of the S5 background subspace of type IIB string theory in AdS5×S5. An analogous set of deformations of the AdS5 subspace is reviewed here. String energy spectra computed in the near-pp-wave limit of these backgrounds match predictions encoded by discrete, asymptotic Bethe equations, suggesting that the twisted string theory is classically integrable in this regime. These Bethe equations can be derived algorithmically by relying on the existence of Lax representations, and on the Riemann–Hilbert interpretation of the thermodynamic Bethe ansatz. This letter is a review of a seminar given at the Institute for Advanced Study, based on research completed in collaboration with McLoughlin.

Invariant subpixel material detection in hyperspectral imagery

We present an algorithm for subpixel material detection in hyperspectral data that is invariant to the illumination and atmospheric conditions. The algorithm does not require atmospheric correction. The target material spectral reflectance is the only required prior information. A target material subspace model is constructed from the reflectance using a physical model and a background subspace model is estimated directly from the image. These two subspace models are used to compute maximum-likelihood estimates (MLEs) for the target material component and the background component at each image pixel. These estimates form the basis of a generalized likelihood ratio test for subpixel material detection. We present experimental results, using Hyperspectral Digital Imagery Collection Experiment (HYDICE) imagery, that demonstrate the utility of the algorithm for subpixel material detection under varying illumination and atmospheric conditions.