SVM- and MRF-Based Method for Accurate Classification of Hyperspectral Images
read more
Citations
Hyperspectral Unmixing Overview: Geometrical, Statistical, and Sparse Regression-Based Approaches
Deep Feature Extraction and Classification of Hyperspectral Images Based on Convolutional Neural Networks
Hyperspectral Unmixing Overview: Geometrical, Statistical, and Sparse Regression-Based Approaches
Hyperspectral Remote Sensing Data Analysis and Future Challenges
Spectral–Spatial Residual Network for Hyperspectral Image Classification: A 3-D Deep Learning Framework
References
Equation of state calculations by fast computing machines
Statistical learning theory
Stochastic Relaxation, Gibbs Distributions, and the Bayesian Restoration of Images
Probability Estimates for Multi-class Classification by Pairwise Coupling
Kernel-based methods for hyperspectral image classification
Related Papers (5)
Classification of hyperspectral remote sensing images with support vector machines
Frequently Asked Questions (11)
Q2. What is the first step of the proposed procedure?
The first step of the proposed procedure consists in performing a probabilistic SVM pixelwise classification of the hyperspectral image [4], [13].
Q3. How long does it take to perform the SVMMRF-NE method?
While the SVM classifier is a computationally demanding algorithm, other considered methods require at maximum 3% more time to be executed.
Q4. How long did the SVMMRF-NE method take to process the data?
The processing times in seconds were 3339 for the SVM method, 3353 for the WH + MV method, 3351 for the SVMMSF + MV method, 3444 for the SVMMRF-NE method, and 3450 for the SVMMRF-E method.
Q5. How robust is the proposed method for classifying hyperspectral images?
Experimental results have demonstrated that the proposed method yields accurate classification maps within a short time interval and is sufficiently robust for classifying different kinds of images.
Q6. What is the proposed method for computing the edge map?
Instead of computing the edge map, the authors propose to define the following “fuzzy no-edge/edge function”:ε (xj) = 1 − ρjα + ρj (4)where α is a parameter controlling the approximate edge threshold.
Q7. How many iterations did the algorithm take?
After every 106 (order of the number of pixels in an image) iterations, the temperature for the next iteration (k + 1) was recomputed as T k+1 = 0.98T k.
Q8. What is the MAP estimate of a given site?
The authors propose to compute the local energy of a given site associated with a pixel xi asU(xi) = Uspectral(xi) + Uspatial(xi) (1)where Uspectral(xi) is the spectral energy function from the observed data and Uspatial(xi) is the spatial energy term computed over the local neighborhood Ni.
Q9. What is the classification method for the University of Pavia?
For this data set, the SVMMSF + MV classifier gives the best accuracies, and the SVMMRF-E method outperforms the SVMMRF-NE technique in terms of accuracies.
Q10. What is the optimal value of the SVMMRF-NE algorithm?
The initial temperature was set to T 1 = 2 (a relatively low value of the initial temperature results in a faster execution of the algorithm).
Q11. What is the classification method for the University of Pavia image?
the proposed method is efficient only in the case if there is no large misclassified region in the initial pixelwise classification map (this assumption often holds).