Laser & Optoelectronics Progress, Volume. 48, Issue 9, 91001(2011)
Mutual Information Bands Selection and Empirical Mode Decomposition Based Support Vector Machines for Hyperspectral Data High-Accuracy Classification
In remote-sensing data processing research, redundant information and noise of high-dimensional hyperspectral data affect the classification accuracy of hyperspectral data seriously. To solve this problem, we propose an algorithm of hyperspectral data classification based on band selection with mutual information and empirical mode decomposition (MI-EMD-SVM). Band selection based on mutual information is used to achieve redundant information processing, and empirical mode decomposition (EMD) is used to achieve feature extraction. And the obtained hyperspectral data X″ has been processed. The support vector machines (SVM) classification of the data is classified, which has been processed. Experimental results of the AVIRIS data indicate that the proposed approach improves the classification accuracy of hyperspectral data, significantly reduces the number of support vector, and improves the speed of hyperspectral data classification.
Get Citation
Copy Citation Text
Shen Yi, Zhang Min, Zhang Miao. Mutual Information Bands Selection and Empirical Mode Decomposition Based Support Vector Machines for Hyperspectral Data High-Accuracy Classification[J]. Laser & Optoelectronics Progress, 2011, 48(9): 91001
Category: Image Processing
Received: Feb. 16, 2011
Accepted: --
Published Online: Jul. 25, 2011
The Author Email: Yi Shen (shen@hit.edu.cn)