A dense convolutional neural network for hyperspectral image classification

In this letter, a dense convolutional neural network (DCNN) is proposed for hyperspectral image classification, aiming to improve classification performance by promoting feature reuse and strengthening the flow of features and gradients. In the network, features are learned mainly through designed dense blocks, where feature maps generated in each layer can connect directly to the subsequent layers by a concatenation mode. Experiments are conducted on two well-known hyperspectral image data sets, using the proposed method and four comparable methods. Results demonstrate that overall accuracies of the DCNN reached 97.61 and 99.50% for the respective image data sets, representing an obvious improvement over the accuracies of the compared methods. The study confirms that the DCNN can provide more discriminable features for hyperspectral image classification and can offer higher classification accuracies and smoother classification maps.     

A deep learning framework for hyperspectral image classification using spatial pyramid pooling

In this letter, a new deep learning framework for spectral–spatial classification of hyperspectral images is presented. The proposed framework serves as an engine for merging the spatial and spectral features via suitable deep learning architecture: stacked autoencoders (SAEs) and deep convolutional neural networks (DCNNs) followed by a logistic regression (LR) classifier. In this framework, SAEs is aimed to get useful high-level features for the one-dimensional features which is suitable for the dimension reduction of spectral features, while DCNNs can learn rich features from the training data automatically and has achieved state-of-the-art performance in many image classification databases. Though the DCNNs has shown robustness to distortion, it only extracts features of the same scale, and hence is insufficient to tolerate large-scale variance of object. As a result, spatial pyramid pooling (SPP) is introduced into hyperspectral image classification for the first time by pooling the spatial feature maps of the top convolutional layers into a fixed-length feature. Experimental results with widely used hyperspectral data indicate that classifiers built in this deep learning-based framework provide competitive performance.     

Analysis of the pressor dose response

(1) When incremental infusions of drugs that increase blood pressure are given to human subjects to assess "pressor responsiveness," only the lower part of the sigmoid dose-response curve can be obtained. (2) Fitting a quadratic function does not involve discarding data points, which is usually the case with a linear fit, and it provides a more satisfactory fit to the lower part of a sigmoid dose-response curve. (3) In the presence of a competitive antagonist, a pressor dose-response curve will be shifted to the right. In this situation the dose-response curves obtained before and after treatment with antagonist should be fitted simultaneously to a quadratic model in which the parallel shift is one of the parameters. (4) The use of quadratic fitting is illustrated by reference to clinical experiments to obtain the following three curves for drugs that modify peripheral alpha adrenoceptors: norepinephrine pressor response curves after placebo and doxazosin, an alpha 1 antagonist; norepinephrine pressor response curves after placebo, labetalol, and medroxalol (drugs with combined alpha 1 and beta blocking properties); and phenylephrine pressor response curves before and after prazosin. (5) Fitting a quadratic function is the appropriate initial step in the analysis of pressor dose-response curves in man.     

Spectral–spatial classification of hyperspectral images using deep convolutional neural networks

In this letter, a novel deep learning framework for hyperspectral image classification using both spectral and spatial features is presented. The framework is a hybrid of principal component analysis, deep convolutional neural networks (DCNNs) and logistic regression (LR). The DCNNs for hierarchically extract deep features is introduced into hyperspectral image classification for the first time. The proposed technique consists of two steps. First, feature map generation algorithm is presented to generate the spectral and spatial feature maps. Second, the DCNNs-LR classifier is trained to get useful high-level features and to fine-tune the whole model. Comparative experiments conducted over widely used hyperspectral data indicate that DCNNs-LR classifier built in this proposed deep learning framework provides better classification accuracy than previous hyperspectral classification methods.     

Dual force hybrid detector for hyperspectral images

This letter proposes a novel dual force hybrid detector for hyperspectral image (HSI) target detection. As a hybrid detector, we conduct the HSI spectral unmixing in a non-linear feature space mapped by locally linear embedding, and the image reconstruction error in the manifold embedded feature is considered as the primary force for detection. Then, we introduce the adaptive cosine/coherent estimate detector as the additional force. By combining the aforementioned two parts of detection results together, we obtain the final detection probability map. Based on the public data sets of (1) implanted metal detection in a hyperspectral digital imagery collection experiment image and (2) fabrics and vehicles detection in a Hyperspectral Mapper image, the technical results show that our method achieves superior performance compared to some state-of-the-art hyperspectral detectors.     

A semi-supervised convolutional neural network for hyperspectral image classification

Convolutional neural network (CNN) for hyperspectral image classification can provide excellent performance when the number of labeled samples for training is sufficiently large. Unfortunately, a small number of labeled samples are available for training in hyperspectral images. In this letter, a novel semi-supervised convolutional neural network is proposed for the classification of hyperspectral image. The proposed network can automatically learn features from complex hyperspectral image data structures. Furthermore, skip connection parameters are added between the encoder layer and decoder layer in order to make the network suitable for semi-supervised learning. Semi-supervised method is adopted to solve the problem of limited labeled samples. Finally, the network is trained to simultaneously minimize the sum of supervised and unsupervised cost functions. The proposed network is conducted on a widely used hyperspectral image data. The experimental results demonstrate that the proposed approach provides competitive results to state-of-the-art methods.     

A method based on the adaptive cuckoo search algorithm for endmember extraction from hyperspectral remote sensing images

In this letter, an endmember extraction approach based on an adaptive cuckoo search (ACS) algorithm (ACSEE) is proposed for hyperspectral remote sensing imagery. In the proposed algorithm, the problem of endmember extraction is transformed into combinatorial optimization of the candidate endmembers. The effectiveness of the cuckoo search algorithm is demonstrated by its good balance between exploitation of Lévy flights and random walk, which leads the algorithm to effectively explore the solution space and locate potential solutions to avoid falling into local optima. Furthermore, to improve the convergence characteristics of the original cuckoo search algorithm, a new strategy combined with historical information is proposed to accelerate the search process by adjusting the step size of the Lévy flights. The results of experiments conducted using simulated data and well-known hyperspectral remote sensing data indicate that the proposed ACS algorithm can be used as an alternative tool to solve the problem of endmember extraction on account of its robust stability and guarantee of optimal performance.     

A hybrid neural network for hyperspectral image classification

ABSTRACT

Recent research shows that deep learning-based methods can achieve promissing performance when applied to hyperspectral image (HSI) classification in remote sensing, some challenging issues still exist. For example, after a number of 2D convolutions, each feature map may only correspond to a unique dimension of the hyperspectral image. As a result, the relationship between different feature maps from multiple dimensional hyperspectral image can not be extracted well. Another issue is information in extracted feature maps may be erased by pooling operations. To address these problems, we propose a novel hybrid neural network (HNN) for hyperspectral image classification. The HNN uses a multi-branch architecture to extract hyperspectral image features in order to improve its prediction accuracy. Moreover, we build a deconvolution structure to recover the lost information in the pooling operation. In addition, to improve convergence and prevent overfitting, the HNN applies batch normalization (BN) and parametric rectified linear units (PReLU). In the experiments, two public benchmark HSIs are utilized to evaluate the performance of the proposed method. The experimental results demonstrate the superiority of HNN over several well-known methods.     

A supervised abundance estimation method for hyperspectral unmixing

Abundance estimation is one of the key steps in hyperspectral unmixing. Usually, abundance estimation is based on linear mixing. However, in real hyperspectral image, this assumption is not physically rigorous enough, because nonlinear mixture may be observed. Nonlinear models present an improvement by considering the microscopic interactions. However, in most cases, a nonlinear unmixing method should assume a specific nonlinear mixture model, and the corresponding abundance estimation process is only applicable for this model. Recently, supervised machine learning, especially deep learning methods, have achieved promising performance in hyperspectral image processing. Supervised learning is able to capture the mapping between input and output data. In this letter, a new supervised abundance estimation method is proposed, which aims to learn the mapping between pixels spectra and the fractional abundances. To overcome the difficulty that no groundtruth is available in real hyperspectral images, we propose a training samples generation strategy based on synthetic data. The major contribution of this work is that the proposed method can handle the abundance estimation problem in a uniform framework without assuming specific linear or nonlinear mixing model. Experiments on both synthetic and real data are conducted to validate the effectiveness of the proposed method.     

Spectral-spatial constrained sparse unmixing of hyperspectral imagery using a hybrid spectral library

Spectral unmixing based on sparse regression model has recently attracted the hyperspectral data processing community. The technology has two important characteristics. First, sparseness of the abundance matrix is induced because the number of materials in a mixed pixel is very small when compared with the signatures in spectral library. Second, constraints are applied to improve the reconstruction accuracy of the optimization problem. In this letter, a spectral and spatial constrained sparse unmixing (SSCSUn) method is proposed, in which the total variation (TV) regularization with a spatial weight factor is incorporated to exploit spatial information more effectively. Furthermore, several endmembers from the real image are used to form a new hybrid spectral library and are applied as spectral a priori constraints. Experimental results on simulated and real hyperspectral images demonstrate the effectiveness of the proposed approach.     

Automatic treatment of radioimmunoassay data: an experimental validation of the results.

A computational method for the automatic treatment of radioimmunoassay data has been developed and a computer program has been written in accordance. The main features of the approach used are the following: (a) a constant ratio between the bound fraction and its variance is assumed and estimated in each assay; (b) the points of the dose-response curve are fitted using the three-parameter function y' equals b1/(1 plus b2x(-b)3) where y' equals bound - nonspecific counts and x is the amount of hormone; the fitting is performed using the nonlinear, least-squares technique; (c) the values of the unknown samples are evaluated from the fitted standard curve; their confidence limits are computed taking into account both the variance of the bound replicates and the variance of the parameters of the dose-response curve. Experimental data that support the validity of the assumption on the variance of bound measurement and the suitability of the chosen function to fit the points of the standard curve are presented. A comparison between the confidence limits of the unknowns experimentally obtained and those computed by the program is reported and discussed.     

Spectral-Spatial classification of hyperspectral images using functional data analysis

In this letter, a spectral-spatial classification method using functional data analysis (FDA) is proposed. Since the efficacy of the FDA for hyperspectral image analysis instead of analysis in multivariate analysis framework (MAF) was proved previously, we apply FDA to better extract spectral and spatial information for hyperspectral image classification. Therefore, in the FDA framework a support vector machine (SVM) classifier is used for hyperspectral image classification and a watershed segmentation algorithm is applied in order to extract spatial structures. Several approaches to figure a one-band gradient image, as an input to watershed transformation, are examined and investigated. As a result, the extracted segmentation map is used to improve the pixel-wise classification accuracy on which the classification and the segmentation results are combined together using majority vote approach. The efficiency of the proposed method is evaluated on two hyperspectral data sets. The experimental results show that the proposed spectral-spatial classification method provides better classification accuracies compared to some state-of-the-art spectral-spatial classification methods.     

Spectral-spatial classification of hyperspectral imagery based on recurrent neural networks

Deep neural networks have recently been successfully explored to extract deep features for hyperspectral image classification. Recurrent neural networks (RNNs) are an important branch of the deep learning family, which are widely used for sequence analysis. Indeed, RNNs have been used to model the dependencies between the different spectral bands of hyperspectral image, inspired by the observation that hyperspectral pixels can be considered as spectral sequences. A disadvantage of such methods is that they don’t consider the effect of neighborhood pixels on the final class label. In this letter, a RNN model is proposed for the spectral-spatial classification of hyperspectral image. Specifically, the hyperspectral image cube surrounding a central pixel is considered as a hyperspectral pixels sequence, and a RNN is used to model the dependencies between the different neighborhood pixels. The proposed RNN is conducted on two widely used hyperspectral image datasets. The experimental results demonstrate that the proposed approach provides a better performance than that of conventional methods.     

Block-and-octave constraint SIFT with multi-thread processing for VHR satellite image matching

Very high-resolution satellite image matching is a challenging task due to local distortion, repetitive structures, intensity changes and low efficiency. In this letter, a novel approach based on the block and octave constraint scale-invariant feature transform (SIFT) is proposed for very high-resolution satellite image matching. Details regarding multi-thread processing are provided. First, the corresponding matching block is observed using rational function model (RFM) prediction. The RFM prediction can provide a geometric constraint. In each block matching, keypoints are classified with the feature octave. The keypoints are matched in the group with the same feature octave. The block and octave constraint can narrow the keypoints search scope and be helpful for acquiring additional matches. Since blocks are independent, each corresponding block matching is parallelized to realize multi-thread processing with open multi-processing (OpenMP). Experiments are designed to test the matching performance and runtime. The results indicate the effectiveness and efficiency of the proposed approach.     

PolSAR image classification using a semi-supervised classifier based on hypergraph learning

This letter presents a novel semi-supervised method based on hypergraph learning for polarimetric synthetic aperture radar (PolSAR) image classification. Compared with the classic support vector machine, simple-graph learning, k-nearest neighbour (k-NN) and semi-supervised discriminant analysis (SDA) classifiers, the proposed method achieves better performance with fewer labelled points for PolSAR imagery. A hyperspectral image is used for comparison with use of PolSAR imagery, and the proposed method is found to be inferior to k-NN and SDA for the hyperspectral image. The performance of our method is evaluated in single, dual and full-polarization cases, respectively. The results demonstrate that the performance of our method in the full-polarization case is superior to that in either single or dual-polarization case.     

Constrained nonnegative matrix factorization and hyperspectral image dimensionality reduction

This letter presents a constrained nonnegative matrix factorization (NMF)-based method for hyperspectral image dimensionality reduction. The proposed method combines the NMF and Laplacian Eigenmaps (LE). It overcomes the drawback that NMF does not consider the intrinsic geometric structure of the data space. In LE framework, an affinity graph is constructed to encode the geometrical information. The proposed technique seeks a matrix factorization which considers the graph structure. We also use the smoothness constraint and the sparsity constraint on the lower dimensional matrices. The gradient descent approach is used to find solution of the proposed model. In order to evaluate the developed method, we use the support vector machine and the k-nearest neighbourhood (KNN) approach for hyperspectral image classification. Experiments are done on a hyperspectral image. The results are compared with those obtained using other hyperspectal image dimensionality reduction methods. The classification accuracy using the proposed method is higher than that of the alternative approaches.     

Semi-Supervised hyperspectral image classification using local low-rank representation

In the area of hyperspectral image (HSI) classification, graph-based semi-supervised learning (SSL) has been proved to be highly effective. Constructing a proper graph is critical for graph-based SSL tasks. In HSI, spectral distance is widely used to calculate the weight of graph edge, though it can be influenced by noise and outliers. Meanwhile, links among all the data points are incorporated in the graph, including those from different subspaces. Thus the constructed graph might contain incorrect information. In this letter, a novel semi-supervised HSI classification method using local low-rank representation (SL²R) is proposed. Edge weight calculation will not be affected by noise or outliers thanks to the robustness of low-rank representation (LRR). Since each graph is constructed at local level, where pixels are basically embedded in the same subspace, links among uncorrelated pixels can be removed. Moreover, spatial context is naturally characterized by low-rank constraint on adjacent pixels. Experimental results on two data sets (Indian Pines and Botswana) confirm the effectiveness of the proposed method.     

Application of continuous positive airway pressure to trace static pressure-volume curves of the respiratory system

OBJECTIVE: To evaluate a new technique for pressure-volume curve tracing. DESIGN: Prospective experimental study. SETTING: Animal research laboratory. SUBJECTS: Six anesthetized rats. INTERVENTIONS: Two pressure-volume curves were obtained by means of the super-syringe method (gold standard) and the continuous positive airway pressure (CPAP) method. For the CPAP method, the ventilator was switched to CPAP and the pressure level was raised from 0 to 50 cm H2O in 5 cm H2O steps and then decreased, while we measured lung volume using respiratory inductive plethysmography. Thereafter, lung injury was induced using very high-volume ventilation. Following injury, two further pressure-volume curves were traced. Pressure-volume pairs were fitted to a mathematical model. MEASUREMENTS AND MAIN RESULTS: Pressure-volume curves were equivalent for each method, with intraclass correlation coefficients being higher than.75 for each pressure level measured. Bias and precision for volume values were 0.46 +/- 0.875 mL in basal measurements and 0.31 +/- 0.67 mL in postinjury conditions. Lower and upper inflection points on the inspiratory limb and maximum curvature point on the deflation limb obtained using both methods and measured by regression analysis also were correlated, with intraclass correlation coefficients (95% confidence interval) being.97 (.58,.99),.85 (.55,.95), and.94 (.81,.98) (p <.001 for each one). When inflection points were estimated by observers, the correlation coefficient between methods was.90 (.67,.98) for lower inflection points (p <.001). However, estimations for upper inflection points and maximum curvature point were significantly different. CONCLUSIONS: The CPAP method for tracing pressure-volume curves is equivalent to the super-syringe method. It is easily applicable at the bedside, avoids disconnection from the ventilator, and can be used to obtain both the inspiratory and the deflation limbs of the pressure-volume curve. Use of regression techniques improves determination of inflection points.