Adaptive Coherence/cosine Estimator Research Articles

Ensemble-Based Cascaded Constrained Energy Minimization for Hyperspectral Target Detection

Ensemble learning is an important group of machine learning techniques that aim to enhance the nonlinearity and generalization ability of a learning system by aggregating multiple learners. We found that ensemble techniques show great potential for improving the performance of traditional hyperspectral target detection algorithms, while at present, there are few previous works have been done on this topic. To this end, we propose an Ensemble based Constrained Energy Minimization (E-CEM) detector for hyperspectral image target detection. Classical hyperspectral image target detection algorithms like Constrained Energy Minimization (CEM), matched filter (MF) and adaptive coherence/cosine estimator (ACE) are usually designed based on constrained least square regression methods or hypothesis testing methods with Gaussian distribution assumption. However, remote sensing hyperspectral data captured in a real-world environment usually shows strong nonlinearity and non-Gaussianity, which will lead to performance degradation of these classical detection algorithms. Although some hierarchical detection models are able to learn strong nonlinear discrimination of spectral data, due to the spectrum changes, these models usually suffer from the instability in detection tasks. The proposed E-CEM is designed based on the classical CEM detection algorithm. To improve both of the detection nonlinearity and generalization ability, the strategies of “cascaded detection”, “random averaging” and “multi-scale scanning” are specifically designed. Experiments on one synthetic hyperspectral image and two real hyperspectral images demonstrate the effectiveness of our method. E-CEM outperforms the traditional CEM detector and other state-of-the-art detection algorithms. Our code will be made publicly available.

SparseCEM and SparseACE for Hyperspectral Image Target Detection

Due to the limitation of the spatial resolution of hyperspectral sensors, in real hyperspectral remote sensing images, targets of interest usually only occupy a few pixels (or even subpixels). Under such circumstances, we hope that the output of the detection algorithm is sparse. However, the existing detection algorithms seldom restrict this sparsity. Among the developed detection algorithms, the constrained energy minimization (CEM) and the adaptive coherence/cosine estimator (ACE) are two famous and widely used algorithms. In this letter, based on the CEM and the ACE, we propose the novel sparse CEM (SparseCEM) and sparse ACE (SparseACE) using the $\ell_{1}$ -norm regularization term to restrict the output to be sparse. Furthermore, we convert our detection models to second-order cone program problems, which can be efficiently solved by using the interior point method. The experiments on two real hyperspectral images demonstrate the effectiveness of the proposed algorithms.

Target detection using difference measured function based matched filter for hyperspectral imagery

Automatic target detection is a research focus in hyperspectral image processing field. Algorithms such as the matched filter (MF) and the adaptive coherence/cosine estimator (ACE) directly use the prior knowledge of target spectral signatures to detect targets. In this paper, we propose a difference measured function based matched filter (DFMF), which could include the famous algorithm MF as a special case. The DFMF uses a new measured function to build an objective function, and utilizes the gradient descent method to find an optimal projection vector. After finding the optimal projection vector, the interesting targets can be detected in the projection space. The experimental results demonstrate the proposed algorithm could detect interesting targets effectively and performs better than some other experimental algorithms.

Recursive spectral similarity measure-based band selection for anomaly detection inhyperspectral imagery

Band selection has been widely used in hyperspectral image processing for dimensionreduction. In this paper, a recursive spectral similarity measure-based band selection(RSSM-BBS) method is presented. Unlike most of the existing image-based band selectiontechniques, it is for two hyperspectral signatures with its main focus on their spectralseparability. Furthermore, it is unsupervised and based on the recursive calculation of thespectral similarity measure with an additional band. In order to demonstrate the utility ofthe proposed method, an anomaly detection algorithm is developed, which firstextracts the anomalous target spectrum from the image using the automatictarget detection and classification algorithm (ATDCA), followed by the maximumspectral screening (MSS) to obtain a good estimate of the background, and thenimplements RSSM-BBS to select bands that participate in the subsequent adaptivecosine/coherence estimator (ACE) target detection. As shown in the experimentalresult on the AVIRIS dataset, the detection performance of the ACE has beenimproved greatly with the bands selected by RSSM-BBS over that using full bands.

Adaptive multidimensional Wiener filtering for target detector improvement

In this paper, we consider the problem of hyperspectral image denoising. Current denoising is based on multichannel restoration filters assuming the separability of the signal covariance, which describes the between-channel and within-channel relationships. We propose a new algorithm for a spectral band restoration scheme, the adaptive multidimensional Wiener filter, based on a local signal model, without assuming spectral and spatial separability. The proposed filter can be applied as a preprocessing step for detection in hyperspectral imagery. We highlight the target detection improvement when the developed method is used before existing methods the well-known hyperspectral imagery detectors as: AMF (Adaptive Matched Filter), ACE (Adaptive coherence/cosine Estimator) and RX (Reed and Xiaoli algotithm). We demonstrate that integrating a multidimensional restoration leads to significant improvement of the detection probability. The performance of our method is exemplified using real-world HYDICE images.

Fast converging adaptive matched filter and adaptive cosine/coherence estimator

This paper deals with the problem of detecting a signal known up to a scaling factor in the presence of Gaussian disturbance with unknown covariance matrix. We focus on scenarios where the number of secondary data which share the same covariance of the data under test is very limited. Thus, we propose a modified version of the adaptive matched filter (AMF) and of the adaptive cosine/coherence estimator (ACE) which can be applied even if the sample covariance matrix is rank deficient. Even though these detectors do not ensure the constant false alarm rate (CFAR) property with respect to the covariance matrix of the disturbance, a sensitivity analysis has shown that the threshold setting is very robust with respect to possible discrepancies between the design and the operating conditions. Finally, numerical results highlight that the proposed receivers outperform the conventional AMF and ACE in low sample situations.