Study of Spatial–Spectral Feature Extraction Frameworks With 3-D Convolutional Neural Network for Robust Hyperspectral Imagery Classification

Praveen, Bishwas; Menon, Vineetha

doi:10.1109/jstars.2020.3046414

Cited by 27 publications

(30 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Figures 8,9,10,11,12,13 shows the wheat seeds dataset clustering for the algorithms k-means, nested mini-batch k-means, birch [26], mini-batch k-means, agglomerative clustering, DPCGS (our method) respectively. Figures 14,15,16,17,18,19,20 shows the iris dataset clustering for the algorithms k-means, nested minibatch k-means, birch, mini-batch k-means, agglomerative clustering, affinity propogation [3], DPCGS (our method) respectively. Figures 21,22,23,24,25,26,27 shows the breast cancer dataset clustering for the algorithms k-means, nested mini-batch k-means, birch, mini-batch k-means, agglomerative clustering, affinity propogation, DPCGS (our method) respectively.…”

Section: Results and Detailed Analysismentioning

confidence: 99%

“…In supervised learning [18], the machine learning model is trained with the labeled dataset. During the training process, the features [19] that contribute to the data classification are found out, and the network's loss is calculated against the target class. Accordingly, the weights of the network are backpropagated until the loss of predicted vs. target is minimized.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Data Points Clustering via Gumbel Softmax

Acharya

Zhang

2021

SN COMPUT. SCI.

View full text Add to dashboard Cite

Finding useful patterns in the dataset has been a fascinating topic, and one of the most researched problems in this area is identifying the cluster groups within the dataset. This research paper introduces a "new data clustering method" called Data Points Clustering via Gumbel Softmax (DPCGS) and demonstrates that it is suitable for clustering the data points datasets. We evaluate DPCGS efficiency and clustering quality through several experiments. Experiments show that statistically relevant clustering structures can be identified with our method, depending on the dataset. We also present a performance comparison table where we use datasets such as Wine, Wheat seeds, Iris, Wisconsin breast cancer and compare the DPCGS results with different benchmarking and recently proposed clustering algorithms such as Birch, K-Means, Affinity propagation, Agglomerative clustering, and Mini-batch K-Means and Nested mini-batch K-Means. Our method DPCGS performs better than most of the previously and recently proposed clustering algorithms.

show abstract

Section: Results and Detailed Analysismentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Data Points Clustering via Gumbel Softmax

Acharya

Zhang

2021

SN COMPUT. SCI.

View full text Add to dashboard Cite

show abstract

“…where b is the bias of the convolutional layer, f is the nonlinear activation function, (i, j) represents the coordinate of a given pixel, and X l-1 represents the set of input sensors (and is also the set of output in the l-1th layer [41]).…”

Section: B Cnnmentioning

confidence: 99%

A Novel Multitemporal Deep Fusion Network (MDFN) for Short-Term Multitemporal HR Images Classification

Zheng¹,

Liu²,

Du³

et al. 2021

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

High-resolution (HR) satellite images, due to the technical constraints on spectral and spatial resolutions, usually contain only several broad spectral bands but with a very high spatial resolution. This provides rich spatial details of the objects on the Earth surface, while their spectral discrimination is relatively low. Recently, the increase of the satellite revisit times made it possible to acquire more frequent data coverage for finer classification. In this paper, we proposed a novel multitemporal deep fusion network (MDFN) for short-term multitemporal HR images classification. Specifically, a two-branch structure of MDFN is designed, which includes a long short-term memory (LSTM) and a convolutional neural network (CNN). The LSTM branch is mainly used to learn the joint expression of different temporal-spectral features. For the CNN branch, the 3D convolution is firstly applied along the temporal and spectral dimensions to jointly learn the temporal-spatial and spectral-spatial information, respectively, and then the 2D convolution is performed along the spatial dimension to further extract the spatial context information. Finally, features generated from the two different branches are fused to obtain the discriminative high-level semantic information for classification. Experimental results carried on two real multitemporal HR remote sensing data sets demonstrate that the proposed MDFN provides better classification performance over the state-of-the-art methods, and it also shows the potentiality to use short-term multitemporal HR images for more accurate Land Use/Land Cover (LULC) mapping.

show abstract

“…In hyperspectral remote sensing, deep learning algorithms have been widely applied to hyperspectral imaging classification processing tasks. For example, in [16], a spatialspectral feature extraction framework for robust hyperspectral images classification was proposed to combine a 3D convolutional neural network. Testing overall classification accuracies was 4.23% higher than SVM on Pavia data sets and Pines data sets.…”

Section: Introductionmentioning

confidence: 99%

Spectral Classification Based on Deep Learning Algorithms

Xie

Cai

et al. 2021

Electronics

View full text Add to dashboard Cite

Convolutional neural networks (CNN) can achieve accurate image classification, indicating the current best performance of deep learning algorithms. However, the complexity of spectral data limits the performance of many CNN models. Due to the potential redundancy and noise of the spectral data, the standard CNN model is usually unable to perform correct spectral classification. Furthermore, deeper CNN architectures also face some difficulties when other network layers are added, which hinders the network convergence and produces low classification accuracy. To alleviate these problems, we proposed a new CNN architecture specially designed for 2D spectral data. Firstly, we collected the reflectance spectra of five samples using a portable optical fiber spectrometer and converted them into 2D matrix data to adapt to the deep learning algorithms’ feature extraction. Secondly, the number of convolutional layers and pooling layers were adjusted according to the characteristics of the spectral data to enhance the feature extraction ability. Finally, the discard rate selection principle of the dropout layer was determined by visual analysis to improve the classification accuracy. Experimental results demonstrate our CNN system, which has advantages over the traditional AlexNet, Unet, and support vector machine (SVM)-based approaches in many aspects, such as easy implementation, short time, higher accuracy, and strong robustness.

show abstract

Study of Spatial–Spectral Feature Extraction Frameworks With 3-D Convolutional Neural Network for Robust Hyperspectral Imagery Classification

Cited by 27 publications

References 34 publications

Data Points Clustering via Gumbel Softmax

Data Points Clustering via Gumbel Softmax

A Novel Multitemporal Deep Fusion Network (MDFN) for Short-Term Multitemporal HR Images Classification

Spectral Classification Based on Deep Learning Algorithms

Contact Info

Product

Resources

About