Pavia University Scene Research Articles

PGNN-Net: Parallel Graph Neural Networks for Hyperspectral Image Classification Using Multiple Spatial-Spectral Features

Hyperspectral image (HSI) shows great potential for application in remote sensing due to its rich spectral information and fine spatial resolution. However, the high dimensionality, nonlinearity, and complex relationship between spectral and spatial features of HSI pose challenges to its accurate classification. Traditional convolutional neural network (CNN)-based methods suffer from detail loss in feature extraction; Transformer-based methods rely too much on the quantity and quality of HSI; and graph neural network (GNN)-based methods provide a new impetus for HSI classification by virtue of their excellent ability to handle irregular data. To address these challenges and take advantage of GNN, we propose a network of parallel GNNs called PGNN-Net. The network first extracts the key spatial-spectral features of HSI using principal component analysis, followed by preprocessing to obtain two primary features and a normalized adjacency matrix. Then, a parallel architecture is constructed using improved GCN and ChebNet to extract local and global spatial-spectral features, respectively. Finally, the discriminative features obtained through the fusion strategy are input into the classifier to obtain the classification results. In addition, to alleviate the over-fitting problem, the label smoothing technique is embedded in the cross-entropy loss function. The experimental results show that the average overall accuracy obtained by our method on Indian Pines, Kennedy Space Center, Pavia University Scene, and Botswana reaches 97.35%, 99.40%, 99.64%, and 98.46%, respectively, which are better compared to some state-of-the-art methods.

Land Cover Classification from Hyperspectral Images via Weighted Spatial–Spectral Joint Kernel Collaborative Representation Classifier

The continuous changes in Land Use and Land Cover (LULC) produce a significant impact on environmental factors. Highly accurate monitoring and updating of land cover information is essential for environmental protection, sustainable development, and land resource planning and management. Recently, Collaborative Representation (CR)-based methods have been widely used in land cover classification from Hyperspectral Images (HSIs). However, most CR methods consider the spatial information of HSI by taking the average or weighted average of spatial neighboring pixels of each pixel to improve the land cover classification performance, but do not take the spatial structure information for pixels into account. To address this problem, a novel Weighted Spatial–Spectral Joint CR Classification (WSSJCRC) method is proposed in this paper. WSSJCRC only performs spatial filtering on HSI through a weighted spatial filtering operator to alleviate the spectral shift caused by adjacency effect, but also utilizes the labeled training pixels to simultaneously represent each test pixel and its spatial neighborhood pixels to consider the spatial structure information of each test pixel to assist the classification of the test pixel. On this basis, the kernel version of WSSJCRC (i.e., WSSJKCRC) is also proposed, which projects the hyperspectral data into the kernel-induced high-dimensional feature space to enhance the separability of nonlinear samples. The experimental results on three real hyperspectral scenes show that the proposed WSSJKCRC method achieves the best land cover classification performance among all the compared methods. Specifically, the Overall Accuracy (OA), Average Accuracy (AA), and Kappa statistic (Kappa) of WSSJKCRC reach 96.21%, 96.20%, and 0.9555 for the Indian Pines scene, 97.02%, 96.64%, and 0.9605 for the Pavia University scene, and 95.55%, 97.97%, and 0.9504 for the Salinas scene, respectively. Moreover, the proposed WSSJKCRC method obtains the promising accuracy with OA over 95% on the three hyperspectral scenes under the situation of small-scale labeled samples, thus effectively reducing the labeling cost for HSI.

Spatial–spectral hyperspectral images classification based on Krill Herd band selection and edge-preserving transform domain recursive filter

Hyperspectral images (HSIs) have recently been exploited in several aspects as HSIs contain many contiguous and narrow discriminative spectral bands. The problem of dimensionality is a significant dilemma for HSIs due to there being plenty of irrelevant and redundant spectral bands and highly correlated bands that lead to Hughes phenomenon. To this end, we present an approach to selecting the most informative and relevant spectral bands for HSI dimensionality reduction using the Krill Herd (KH) algorithm. Moreover, KH is a heuristic search method that seeks to reach the optimum global solution within the search space and effectively evade falling into the local optima. Then an edge-preserving filter was employed to extract the spatial features while reducing noise and obtaining a suitable smoothing that improves the classification performance. Finally, the support vector machine classifier was performed at the pixel level for HSI classification. Furthermore, the proposed work was compared with the harmony search, genetic algorithm, bat algorithm, particle swarm optimization, and firefly algorithm. The experimental results demonstrated outstanding performance with an overall accuracy equal to 96.54%, 98.93%, 99.78%, and 98.66% on four hyperspectral datasets: Indian Pines scene, Pavia University scene, Salinas scene, and Botswana scene, respectively.

Nonlinear extended blind end-member and abundance extraction for hyperspectral images

Hyperspectral images had become an essential tool in different application frameworks, such as mineral exploration, food inspection, and medical assessment, among others. However, the interpretability of these images involves an initial processing stage to model the optical interaction and analyze the spectral information. In this work, we study nonlinear unmixing of hyperspectral images by a multilinear mixture model (MMM). In this sense, we propose a nonlinear version of the extended blind end-member and abundance extraction (NEBEAE) method for blind unmixing, i.e. estimation of the end-members, their abundances, and the nonlinear interaction levels. In the problem formulation, we include a normalization step in the hyperspectral measurements for the end-members and abundances to improve robustness. The blind unmixing process can be separated into three estimation subproblems for each component in the model, which are solved by a cyclic coordinate descent algorithm and quadratic constrained optimizations. Each problem is mathematically formulated and derived to construct the general nonlinear iterative unmixing technique. We evaluated our proposal with synthetic and experimental datasets from the remote sensing literature (Cuprite, Urban and Pavia University scene datasets) and a biomedical hyperspectral imaging application. In the validation stage, we compared NEBEAE with three state-of-the-art methods to show its advantages in terms of precision and computational time.

Hyperspectral Data Classification Algorithm considering Spatial Texture Features

As a cutting-edge technology, hyperspectral remote sensing has been widely applied in many fields, including agricultural production, mineral identification, target detection, disaster warning, military reconnaissance, and urban planning. The collected hyperspectral data have high spectral resolution and spatial resolution and are characterized by a large amount of information, redundancy, and high dimension. At the same time, there is a strong correlation between the bands. Therefore, hyperspectral data not only provides rich information but also brings great challenges for subsequent processing. Hyperspectral image classification is a hot issue in remote sensing information processing. Traditional hyperspectral remote sensing image classification methods only use the spectral features of the image without considering the spatial features of each pixel in the hyperspectral remote sensing image. In this paper, a hyperspectral image classification method is proposed not only considering spectral features but also considering texture features. This method jointly considers both these features. Firstly, six texture features contributing a lot to each pixel of hyperspectral remote sensing image are extracted by using a gray level cooccurrence matrix, and then, the spectral features of each pixel in neighbor are combined to form the texture-spectral features. Finally, the classification experiment of the Indian Pines and Pavia University scene is carried out based on a support vector machine and extreme random tree algorithm, and the obtained results show that the proposed method achieves higher classification performance than the traditional method.

A hybrid optimization-based deep belief neural network for the classification of vegetation area in multi-spectral satellite image

Over the last few decades, multiple advances have been done for the classification of vegetation area through land cover, and land use. However, classification problem is one of the most complicated and contradicting problems that has received considerable attention. Therefore, to tackle this problem, this paper proposes a new Firefly-Harmony search based Deep Belief Neural Network method (FHS-DBN) for the classification of land cover, and land use. The segmentation process is done using Bayesian Fuzzy Clustering,and the feature matrix is developed. The feature matrix is given to the proposed FHS-DBN method that distinguishes the land coverfrom the land use in the multispectral satellite images, for analyzing the vegetation area. The proposed FHS-DBN method is designedby training the DBN using the FHS algorithm, which is developed by the combination of Firefly Algorithm (FA) and Harmony Search (HS) algorithm. The performance of the FHS-DBN model is evaluated using three metrics, such as Accuracy, True Positive Rate (TPR), and False Positive Rate (FPR). From the experimental analysis, it is concludedthat the proposed FHS-DBN model achieves ahigh classification accuracy of 0.9381, 0.9488, 0.9497, and 0.9477 usingIndian Pine, Salinas scene, Pavia Centre and university, and Pavia University scene dataset.

A 3D-2D Convolutional Neural Network and Transfer Learning for Hyperspectral Image Classification.

As one of the fast evolution of remote sensing and spectral imagery techniques, hyperspectral image (HSI) classification has attracted considerable attention in various fields, including land survey, resource monitoring, and among others. Nonetheless, due to a lack of distinctiveness in the hyperspectral pixels of separate classes, there is a recurrent inseparability obstacle in the primary space. Additionally, an open challenge stems from examining efficient techniques that can speedily classify and interpret the spectral-spatial data bands within a more precise computational time. Hence, in this work, we propose a 3D-2D convolutional neural network and transfer learning model where the early layers of the model exploit 3D convolutions to modeling spectral-spatial information. On top of it are 2D convolutional layers to handle semantic abstraction mainly. Toward simplicity and a highly modularized network for image classification, we leverage the ResNeXt-50 block for our model. Furthermore, improving the separability among classes and balance of the interclass and intraclass criteria, we engaged principal component analysis (PCA) for the best orthogonal vectors for representing information from HSIs before feeding to the network. The experimental result shows that our model can efficiently improve the hyperspectral imagery classification, including an instantaneous representation of the spectral-spatial information. Our model evaluation on five publicly available hyperspectral datasets, Indian Pines (IP), Pavia University Scene (PU), Salinas Scene (SA), Botswana (BS), and Kennedy Space Center (KSC), was performed with a high classification accuracy of 99.85%, 99.98%, 100%, 99.82%, and 99.71%, respectively. Quantitative results demonstrated that it outperformed several state-of-the-arts (SOTA), deep neural network-based approaches, and standard classifiers. Thus, it has provided more insight into hyperspectral image classification.

Three-dimensional convolutional neural network model for tree species classification using airborne hyperspectral images

Airborne hyperspectral remote sensing data with both rich spectral and spatial features can effectively improve the classification accuracy of vegetation species. However, the spectral data of hundreds of bands brings about problems such as dimensional explosion, which poses a huge challenge for hyperspectral remote sensing classification based on classical parameters models. Deep learning methods have been used for remotely sensed images classification in recent years, but the popular HSI datasets including Kennedy Space Center, Indian Pines, Pavia University scene and Salinas scene, have low spatial resolution, significant differences between categories, and regular boundaries. When applied to the classification of forestry tree species, the accuracy often decreases because the spectral response of different plants of the same family and genus are very similar, especially under the fragmented species distribution, complex topography and the occluded canopy. So we collect new data sets, selected Gaofeng State Owned Forest Farm in Guangxi province in south China as the research area and adopted the airborne hyperspectral data obtained by the LiCHy system of the Chinese Academy of Forestry to explore an improved three-dimensional convolutional neural network(3D-CNN) model for tree species classification. The proposed model uses raw data as input without dimension reduction or feature screening, and simultaneously extracts spectral and spatial features. After the 3D convolutional layer, the captured high-level semantic concept is a joint spatial spectral feature representation, so we can turn it into a one-dimensional feature as a new input to learn a more abstract level of expression. The widely used earlystop method is also used to prevent overfitting. The proposed model is a lightweight, generalized, and fast convergence classification model, by which the short-time and large-area of multiple tree species classification with high-precision can be realized. The result shows that the 3D-1D CNN model can shorten the training time of the 3D CNN model by 60% and achieve a classification accuracy of 93.14% within 50 ha in 6.37 min, which provides a basis for the classification of tree species, the mapping of forest form and the inventory of forest resources.

Classification of hyperspectral imagery with a 3D convolutional neural network and J-M distance

A three-dimensional convolutional neural network (3D-CNN) is proposed and applied to the identification of land types from hyperspectral images. Due to the advantages of the combined use of the spectral-spatial features of hyperspectral imagery, high accuracy identification of most objects can be realized. However, too many wavelengths increase information redundancy between adjacent bands and interfere with classification accuracy to some extent, complicating achievement of better results for the identification of ground objects with relatively small differences in spectral and spatial domains. To solve such problems, in this paper, the Jeffries-Matusita (J-M) distance is introduced to select effective bands to reduce the redundancy of spectral information for identifying objects with similar features. This method is based on a 3D-CNN considering pixel spectral and spatial information. The optimal band combination algorithm based on the J-M distance is first used to extract spectral features of hyperspectral data while reducing feature dimensions. Then, the 3D-CNN is applied to mine spectral and spatial features from the hyperspectral images. Finally, a Softmax classifier is used to classify land types based on the high-level features learned by the 3D-CNN. Experiments were carried out on data from an area of north-western Indiana and a Pavia university scene. In such images, some objects have very similar spectral and spatial features. The results were compared with the current 3D-CNN land type classification and show that both methods can achieve high-precision identification of most land types, but for objects with similar features, the method proposed in this article has obvious advantages.

Sequential spectral clustering of hyperspectral remote sensing image over bipartite graph

Unsupervised classification is a crucial step in remote sensing hyperspectral image analysis where producing labeled data is a laborious task. Spectral Clustering is an appealing graph-partitioning technique with outstanding performance on data with non-linear dependencies. However, Spectral Clustering is restricted to small-scale data and neither has been effectively applied to hyperspectral image analysis. In this paper, the unsupervised classification of hyperspectral images is addressed through a sequential spectral clustering that can be extended to the large-scale hyperspectral image. To this end, this paper utilizes a bipartite graph representation along with a sequential singular value decomposition and mini-batch K-means for unsupervised classification of hyperspectral imagery. We evaluate the proposed algorithm with several benchmark hyperspectral datasets including Botswana, Salinas, Indian Pines, Pavia Center Scene and Pavia University Scene. The experimental results show significant improvements made by the proposed algorithm compared to the state-of-art clustering algorithms.

Dimensionality-varied deep convolutional neural network for spectral–spatial classification of hyperspectral data

Many methods of hyperspectral image classification have been proposed recently, and the convolutional neural network (CNN) achieves outstanding performance. However, spectral–spatial classification of CNN requires an excessively large model, tremendous computations, and complex network, and CNN is generally unable to use the noisy bands caused by water-vapor absorption. A dimensionality-varied CNN (DV-CNN) is proposed to address these issues. There are four stages in DV-CNN and the dimensionalities of spectral–spatial feature maps vary with the stages. DV-CNN can reduce the computation and simplify the structure of the network. All feature maps are processed by more kernels in higher stages to extract more precise features. DV-CNN also improves the classification accuracy and enhances the robustness to water-vapor absorption bands. The experiments are performed on data sets of Indian Pines and Pavia University scene. The classification performance of DV-CNN is compared with state-of-the-art methods, which contain the variations of CNN, traditional, and other deep learning methods. The experiment of performance analysis about DV-CNN itself is also carried out. The experimental results demonstrate that DV-CNN outperforms state-of-the-art methods for spectral–spatial classification and it is also robust to water-vapor absorption bands. Moreover, reasonable parameters selection is effective to improve classification accuracy.

Improving hyperspectral band selection by constructing an estimated reference map

We investigate band selection for hyperspectral image classification. Mutual information (MI) measures the statistical dependence between two random variables. By modeling the reference map as one of the two random variables, MI can, therefore, be used to select the bands that are more useful for image classification. A new method is proposed to estimate the MI using an optimally constructed reference map, reducing reliance on ground-truth information. To reduce the interferences from noise and clutters, the reference map is constructed by averaging a subset of spectral bands that are chosen with the best capability to approximate the ground truth. To automatically find these bands, we develop a searching strategy consisting of differentiable MI, gradient ascending algorithm, and random-start optimization. Experiments on AVIRIS 92AV3C dataset and Pavia University scene dataset show that the proposed method outperformed the benchmark methods. In AVIRIS 92AV3C dataset, up to 55% of bands can be removed without significant loss of classification accuracy, compared to the 40% from that using the reference map accompanied with the dataset. Meanwhile, its performance is much more robust to accuracy degradation when bands are cut off beyond 60%, revealing a better agreement in the MI calculation. In Pavia University scene dataset, using 45 bands achieved 86.18% classification accuracy, which is only 1.5% lower than that using all the 103 bands.