Collaborative Representation Model Research Articles

Background covariance discriminative dictionary learning for hyperspectral target detection

Hyperspectral target detection (HTD) aims to identifying targets within a hyperspectral image (HSI) based on provided target spectra. In the current HTD field, representation-based detectors have attracted much attention. However, there are two prominent challenges that are particularly noteworthy. First, the background class encompasses diverse land covers, making its accurate representation challenging. Second, the detection ability can be significantly influenced by the abnormalities and noise in HSI. To tackle these concerns, we propose a novel background covariance discriminative dictionary learning (BCDDL) model for HTD. To enhance the background representation ability and overcome the sparse noise, we combine the dictionary learning with spectral covariance descriptors and undertake background reconstruction in regional scale. Specifically, the input HSI is pre-processed into superpixels, the spectral covariance of each superpixel is used to provide a compact and flexible description of local regional statistical properties. Further, a novel spatial clustering-based dictionary learning method is proposed to learn the background discriminative covariances dictionary. The collaborative representation model within symmetric positive definite (SPD) manifold is utilized to reconstruct background region and get the residual. By merging the background residual with pixel-wise target reconstruction residual, we derive final detection output. Comprehensive experiments on two public hyperspectral datasets and two novel GaoFen-5 datasets demonstrate the superiority of our BCDDL approach over 10 state-of-the-art methods, especially in terms of suppressing background.

Complemented subspace-based weighted collaborative representation model for imbalanced learning

Collaborative representation-based classifiers (CRCs) have demonstrated remarkable classification performance in various pattern recognition fields. However, their success heavily relies on a balanced class distribution. More radically, significantly skewed class distributions may severely affect the effectiveness of the CRCs. To address this limitation, in this study, a complemented subspace-based regularization technique is integrated into the CRC framework for tackling imbalanced classification tasks. To enhance recognition accuracy for minority classes, a class weight learning algorithm is presented, in which the weight of each class is adaptively obtained according to the prior distribution information of the original data. Furthermore, the proposed model provides an efficient closed-form solution, ensuring computational efficiency comparable to CRCs. Extensive experiments on diverse imbalanced datasets are carried out to substantiate the superiority of our method over numerous advanced imbalanced learning algorithms.

Multi-band remote sensing image fusion based on collaborative representation

Remote sensing image fusion extracts the spatial information of the panchromatic (PAN) image to sharpen the geometric structure of a multi-spectral (MS) image. Traditional algorithms that solve the fusion image problem by applying various transformations often result in some losses of spatial and spectral details. To improve the quality of the fusion result, we develop a novel fusion method based on collaborative representation for multi-band remote sensing images. In the developed collaborative representation model, a spectral preservation coefficient based on the spectral contribution and spectral-spatial dependency is designed to retain the spectral in the low-resolution MS (LRMS) image. An intensity modulation coefficient based on the spatial difference between the PAN and MS images that is spectral dependent is designed to adaptively recover and modulate the spatial of the MS image. Through the proposed collaborative representation model, the LRMS, low-resolution PAN, and PAN images, and the designed coefficients collaboratively represent the fusion image. The results of the experiment on various satellite datasets show that the method we develop is effective and robust in enhancing pansharpening.

Self-paced collaborative representation with manifold weighting for hyperspectral anomaly detection

ABSTRACT In the acquisition process of hyperspectral images (HSIs), each band may be contaminated with different degrees of mixing noise. For hyperspectral anomaly detection (HAD) tasks, bands with higher noise contamination levels provide more interference information, thus affecting the detection results. In order to reduce the negative effect of noise in HSIs and improve the robustness of the detector, we propose a self-paced collaborative representation with manifold weighting hyperspectral anomaly detector (SPCRMW). Each band is given an order to join the collaborative representation model according to the degree of noise contamination. Moreover, a novel manifold learning reconstruction-based regularization matrix is proposed to reduce the effect of potential anomalous pixels mixed in the background on collaborative representations. It can automatically assign weights to the background pixels by manifold learning reconstruction error. The results compared with six state-of-the-art HAD methods on three real hyperspectral datasets are presented and illustrate the superiority of the proposed SPCRMW method.

Nonnegative-Constrained Joint Collaborative Representation With Union Dictionary for Hyperspectral Anomaly Detection

Recently, many collaborative representation-based (CR) algorithms have been proposed for hyperspectral anomaly detection. CR-based detectors approximate the image by a linear combination of background dictionaries and the coefficient matrix, and derive the detection map by utilizing recovery residuals. However, these CR-based detectors are often established on the premise of precise background features and strong image representation, which are very difficult to obtain. In addition, pursuing the coefficient matrix reinforced by the general $l_2$-min is very time consuming. To address these issues, a nonnegative-constrained joint collaborative representation model is proposed in this paper for the hyperspectral anomaly detection task. To extract reliable samples, a union dictionary consisting of background and anomaly sub-dictionaries is designed, where the background sub-dictionary is obtained at the superpixel level and the anomaly sub-dictionary is extracted by the pre-detection process. And the coefficient matrix is jointly optimized by the Frobenius norm regularization with a nonnegative constraint and a sum-to-one constraint. After the optimization process, the abnormal information is finally derived by calculating the residuals that exclude the assumed background information. To conduct comparable experiments, the proposed nonnegative-constrained joint collaborative representation (NJCR) model and its kernel version (KNJCR) are tested in four HSI data sets and achieve superior results compared with other state-of-the-art detectors.

Hyperspectral Anomaly Detection With Relaxed Collaborative Representation

Anomaly detection has become an important remote sensing application due to the abundant spectral and spatial information contained in hyperspectral images. Recently, hyperspectral anomaly detection methods based on collaborative representation model have attracted significant attention. Nevertheless, these methods have to face two main challenges: (1) all features (spectral signatures) are constrained to share the same representation coefficient, which ignores the differences among features; (2) existing dictionaries for pixel-by-pixel detection model are usually not reliable. To address these issues, this paper proposes a new relaxed collaborative representation detector for hyperspectral anomaly detection by using a novel non-global dictionary. The proposed detector conducts collaborative representation on each feature dimension of the pixel under test, and simultaneously constrains the coding vectors of different features to be similar. To the best of our knowledge, this is the first time that a detection model is built from each feature dimension. To adjust the contributions of each feature, an adaptive feature weight constrained version of the method is also proposed. The non-global dictionary is constructed by combining the k-nearest neighbor method and an existing global dictionary, which is more reliable and practical than the widely used dual windows dictionary. In addition, this paper also designs a band selection strategy for the proposed method. Experiments on five real datasets indicate that the proposed method suppresses background well and outperforms other classical and state-of-the-art methods.

Spatial‐spectral feature extraction of hyperspectral images using tensor‐based collaborative graph analysis

Although the collaborative graph-based discriminant analysis (CGDA) method has shown promising performance for the feature extraction of the hyperspectral image (HSI), both the intrinsic local subspace structures and spatial structural information are ignored in CGDA. To address these problems, a novel spatial-spectral feature extraction method, i.e. tensor-based collaborative graph analysis, is proposed in this letter. Specifically, the spectral similarity is utilized to calculate the Tikhonov matrix, which can constrain the testing samples to be represented by similar training samples in the collaborative representation model. To fully exploit the 3D spatial-spectral structural information, the collaborative representation model is extended to tensor space by using the third-order tensor representation of HSI, in which samples are constructed by small local patches around the central pixels. Depending on the above techniques, the quality of the collaborative graph constructed by the coefficient matrix can be significantly improved to obtain the discriminative low-dimensional features. Experiments on three HSI data sets demonstrate the superiority of the proposed method.

A neighborhood prior constrained collaborative representation for classification

With the development of machine learning and computer vision, classification technology is becoming increasingly important. Due to the advantage in efficiency and effectiveness, collaborative representation-based classifiers (CRC) have been applied to many practical cognitive fields. In this paper, we propose a new neighborhood prior constrained collaborative representation model for pattern classification. Compared with the naive CRC models which approximate the test sample with all the training data globally, our proposed methods emphasize the guidance of the neighborhood priors in the coding process. Two different kinds of neighbor priors and the models’ weighted extensions are explored from the view of sample representation ability and relationships between the samples. Consequently, the contributions of different samples can be distinguished adaptively and the obtained representations can be more discriminative for the recognition. Experimental results on several popular databases can verify the effectiveness of our proposed methods in comparison with other state-of-the-art classifiers.

Hyperspectral Leaf Image-Based Cucumber Disease Recognition Using the Extended Collaborative Representation Model.

Collaborative representation (CR)-based classification has been successfully applied to plant disease recognition in cases with sufficient training samples of each disease. However, collecting enough training samples is usually time consuming and labor-intensive. Moreover, influenced by the non-ideal measurement environment, samples may be corrupted by variables introduced by bad illumination and occlusions of adjacent leaves. Consequently, an extended collaborative representation (ECR)-based classification model is presented in this paper. Then, it is applied to cucumber leaf disease recognition, which constructs a pure spectral library consisting of several representative samples for each disease and designs a universal variation spectral library that deals with linear variables superimposed on samples. Thus, each query sample is encoded as a linear combination of atoms from these two spectral libraries and disease identity is determined by the disease of minimal reconstruction residuals. Experiments are conducted on spectral curves extracted from normal leaves and the disease lesions of leaves infected with cucumber anthracnose and brown spot. The diagnostic accuracy is higher than 94.7% and the average online diagnosis time is short, about 1 to 1.3 ms. The results indicate that the ECR-based classification model is feasible in the fast and accurate diagnosis of cucumber leaf diseases.

Face Recognition Algorithm Based on Correlation Coefficient and Ensemble-Augmented Sparsity

The representation-based classification method has become a research hotspot in recent years. Representation-based classifiers assign class labels directly to test samples based on a structured dictionary. The structured dictionary is composed of training samples. We call the samples in the dictionary as atoms. To further improve the expression ability of different training atoms in the classifier to the test samples, we propose an ensemble-enhanced sparse classification algorithm based on the correlation coefficient. The model proceeds from two aspects: the representation coefficient level and the similarity between different samples. First, a sparse dictionary is combined with the multiplicative property of training samples to build and solve the sparse and collaborative representation algorithm. The fusion representation coefficients are obtained by weighted sparse representation coefficients and collaborative representation coefficients. The test samples are reconstructed by fusing the representation coefficients, and the minimization recovery residuals for each class of samples are calculated. Second, the correlation coefficient value between the test sample and the training sample is calculated, and the maximum correlation quotient between the test sample and the training sample is obtained. Finally, the maximum correlation quotient and the minimization recovery residual are weighted, and the decision classification is carried out in the decision function to achieve the final face recognition. Experiments on AR, Extended Yale B, Georgia Tech and other general face databases show the effectiveness of the algorithm. The main contribution of this model is that it is more robust than a single sparse or collaborative representation model and can improve recognition accuracy.

A New Local Knowledge-Based Collaborative Representation for Image Recognition

Recently, collaborative representation based classifiers (CRC) have shown outstanding performances in recognition tasks. The key to success of most CRC algorithms states that the testing samples can be coded well by a suitable dictionary globally, while the local knowledge between samples has not been fully considered. We observe that the representations of similar samples have a high degree of similarity. In order to take advantage of this important similarity information, this paper proposes a new local knowledge-based collaborative representation model for image classification. Specifically, certain adjacent training samples of the testing image should be determined firstly, and then the representations of these neighborhoods can be applied to guide the coefficients of the testing samples to be more discriminative. Further, we derive a robust version of the proposed method to treat the face recognition with occlusions or corruptions. Extensive experiments are carried out to show the superiority of the proposed method over other state-of-the-art classifiers on various image recognition tasks.

Collaborative representation-based discriminant neighborhood projections for face recognition

Manifold learning as an efficient dimensionality reduction method has been extensively used. However, manifold learning suffers from the problem of manual selection of parameters, which seriously affects the algorithm performance. Recently, applications of collaborative representation have received concern in some fields such as image processing and pattern recognition research. Based on manifold learning and collaborative representation, this paper develops a new algorithm for feature extraction, which is called collaborative representation-based discriminant neighborhood projections (CRDNP). In CRDNP, we first construct intra-class and inter-class neighborhood graphs of the input data as well as a weight matrix based on collaborative representation model and class label information. Then, a projection to a reduced subspace is obtained by margin maximization between the between-class neighborhood scatter and within-class neighborhood scatter. CRDNP not only characters the inherent geometry relationship of the dataset using L2-graph, but also enhances the between-class submanifold separability. In addition, the discriminating capability of CRDNP is further improved by obtaining the orthogonal projection vectors. Experiment results on public face datasets prove that CRDNP can achieve more accurate results compared with the existing related algorithms.

Robust video summarization using collaborative representation of adjacent frames

With the ever increasing volume of video content, efficient and effective video summarization (VS) techniques are urgently demanded to manage a large amount of video data. Recent developments on sparse representation based approaches have demonstrated promising results for VS. However, these existing approaches treat each frame independently, so the performance can be greatly influenced by each individual frame. In this paper, we formulate the VS problem with a collaborative representation model to take the visual similarity of adjacent frames into consideration. To be specific, during the procedure of reconstruction, both each individual frame and their adjacent frames are reconstructed collaboratively, so the impact of an individual frame can be weakened. In addition, a greedy iterative algorithm is designed for model optimization, where the sparsity and the average percentage of reconstruction (APOR) are adopted to control the iteration. Experimental results on two benchmark datasets with various types of videos demonstrate that the proposed method not only outperforms the state of the art, but also improves the robustness to transitional frames and “outlier” frames.

Hyperspectral Image Classification via Multiscale Joint Collaborative Representation With Locally Adaptive Dictionary

In this letter, a multiscale joint collaborative representation with locally adaptive dictionary (MLJCRC) method is proposed for hyperspectral image classification. Based on the joint collaborative representation model, instead of selecting only a single region scale, MLJCRC incorporates complementary contextual information into classification by multiplying different scales with distinct spatial structures and characteristics. Also, MLJCRC uses a locally adaptive dictionary to reduce the influence of irrelevant pixels on representation, which improves the classification accuracy. The results of experiments on Indian Pines data and Pavia University data demonstrate that the proposed method performs better than support vector machine, sparse representation classification, and other collaborative representation-based classifications.

Collaborative representation with HM-LBP features for palmprint recognition

A novel collaborative representation model with hierarchical multiscale local binary pattern (HM-LBP) for palmprint recognition is proposed in this paper. HM-LBP can retrieve useful information from non-uniform patterns and reduce the influence of gray scale, rotation and illumination. The HM-LBP feature of palmprint is extracted, and its dimension is reduced by principal component analysis. And then, a collaborative classification with HM-LBP is presented to fully exploit the discrimination information. The proposed algorithm is evaluated on the Hong Kong Polytechnic University database (v2) to test its feasibility and performance. The results show that the algorithm can achieve ideal recognition accuracy of 100% and the speediness is able to fit for the real-time palmprint recognition system.

Shape collaborative representation with fuzzy energy based active contour model

This paper presents a fuzzy energy-based active contour model for image segmentation with shape prior based on collaborative representation of training shapes. In the paper, a fuzzy energy functional including a data term and a shape prior term is proposed. The data term relies on image information to guide the evolution of the contour. Meanwhile, the shape prior term constrains the evolving contour with respect to the priori shape to handle background clutter and object occlusion. Especially, in this study, the prior shape is represented as the combination of atoms in the shape dictionary based on collaborative representation. In particular, instead of using ℓ1-norm regularization as in sparse representation, we utilize ℓ2-regularized linear regression scheme which can obtain algebraic solution for the coding coefficients, and significantly reduces the computation time. The proposed model therefore can segment images with background clutter and object occlusion even when the training set includes shapes with large variation. In addition, the proposed shape collaborative representation model also takes less computational time compared to shape sparse representation approach. Experimental results on various images and comparisons with other models show the desired performances of the proposed model.

Dynamic texture recognition with video set based collaborative representation

Efficient feature description and classification of dynamic texture (DT) is an important problem in computer vision and pattern recognition. Recently, the local binary pattern (LBP) based dynamic texture descriptor has been proposed to classify DTs by extending the LBP operator used in static texture analysis to the temporal domain. However, the extended LBP operator cannot characterize the intrinsic motion of dynamic texture well. In this paper, we propose a novel video set based collaborative representation dynamic texture classification method. First, we divide the dynamic texture sequence into subsequences along the temporal axis to form the video set. For each DT, we extract the video set based LBP histogram to describe it. We then propose a regularized collaborative representation model to code the LBP histograms of the query video sets over the LBP histograms of the training video sets. Finally, with the coding coefficients, the distance between the query video set and the training video sets can be calculated for classification. Experimental results on the benchmark dynamic texture datasets demonstrate that the proposed method can yield good performance in terms of both classification accuracy and efficiency.

Improved collaborative representation model with multitask learning using spatial support for target detection in hyperspectral imagery

We propose an improved collaborative representation model with multitask learning using spatial support (ICRTD-MTL) for target detection (TD) in hyperspectral imagery. The proposed model consists of the following aspects. First, multiple features are extracted from the hyperspectral image to represent pixels from different perspectives. Next, we apply these features into the unified CRTD-MTL to acquire a collaborative vector for each feature. To adjust the contribution of each feature, a weight coefficient is included in the optimization problem. Once the collaborative vector is obtained, the class of the test sample can be determined by the characteristics of the collaborative vector on reconstruction. Finally, the spatial correlation and spectral similarity of adjacent neighboring pixels are incorporated into each feature to improve the detection accuracy. The experimental results suggest that the proposed algorithm obtains an excellent performance.

Person Re-Identification Based on Spatiogram Descriptor and Collaborative Representation

Feature and metric designing are two vital aspects in person re-identification. In this letter, we firstly propose a novel spatiogram based person descriptor. Such spatiograms of different image regions from several color channels are calculated and accumulated to create a histogram vector and two distinctive spatial statistical vectors. Secondly, through further investigating the multi-shot set-based metric based on the recent collaborative representation model, we propose an effective and efficient multi-shot metric, which fuses the residual and coding coefficients after collaboratively coding samples on all person classes. Finally, we evaluate the proposed descriptor and metric with other published methods on benchmark datasets. Our methods not only achieve state-of-the-art results but also are novel, straightforward and computationally efficient, which will facilitate the real-time surveillance applications such as pedestrian tracking.

Robust visual ℓ2-regularized least squares tracker with Bayes classifier and coding error

Visual object tracking is a challenging task due to the existence of significant illumination change, appearance variation, and occlusion. We propose a structured collaborative representation-based visual tracking algorithm by using both the generative appearance model and the discriminative model. First, a structured collaborative representation model based on l 2 -regularized least squares is used to exploit both holistic and local information of the target. In the structured collaborative representation framework, a new over-complete dictionary containing both target and background templates is proposed to enhance the robustness of the tracking algorithm. Second, the tracking task is treated as a binary classification problem, and a Bayes classifier is trained online by the use of structured collaborative representation coefficients of positive and negative samples. Furthermore, a residual error score is constructed to improve the detective ability of the tracker. Finally, target templates are updated by combining incremental subspace learning and collaborative representation together, and background templates are subsequently updated by samples around latest results. Compared with four state-of-the-art tracking algorithms in 12 challenging video sequences, the proposed tracking algorithm demonstrates a better performance overall in terms of experimental results.