Joint Multiview Fused ELM Learning with Propagation Filter for Hyperspectral Image Classification
In this paper, we propose a fused extreme learning machine (ELM) method with multiview learning for hyperspectral imagery. The proposed approach consists of the following aspects. First, multiple views of spectral-spatial features are generated from the hyperspectral image by using a multiscale spectral-spatial context aware propagation filter. We next apply the weighted-based probabilistic ELM to these multiple feature views to obtain a robust supervised classification results with high accuracy. The advantages of the proposed method are twofold: (1) the multiscale local spectral-spatial contexts of the image are able to be exploited to improve the classification performance significantly; and (2) the algorithm is simple but very robust to the small size of training labeled samples. The experimental results suggest that the proposed algorithm obtains a competitive performance and outperforms other state-of-the-art ELM-based classifiers and the classical SVM classifier.
KeywordsExtreme Learning Machine Hyperspectral Image Propagation Filter Overall Accuracy Decision Level Fusion
This work was supported in part by the National Key Research and Development Program of China under Grant No. 2016YFF0103604, by the National Natural Science Foundation of China under Grant No. 61571230, Grant No. 11431015 and Grant No. 61171165; by the National Scientific Equipment Developing Project of China under Grant No. 2012YQ050250; and by the Natural Science Foundation of Jiangsu Province under Grant No. BK20161500.
- 2.Camps-Valls, G., Rodrigo-González, A., Muñoz-Marí, J., Gómez-Chova, L., Calpe-Maravilla, J: Hyperspectral image classification with mahalanobis relevance vector machines. In: IEEE International Geoscience and Remote Sensing Symposium (IGARSS 2007), Barcelona, Spain (2007)Google Scholar
- 13.Chang, J.-H.R., Wang, Y.-C.F.: Propagated image filtering. In: IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2015), Boston, USA (2015)Google Scholar
- 14.Lu, T., Li, S.: Gradient-guided sparse representation for hyperspectral image denoising. In: IEEE International Geoscience and Remote Sensing Symposium (IGARSS 2015), Italy, Milan (2015)Google Scholar
- 15.Wen-Bo, Z., Hong-Bing, J.: Fusion of extreme learning machines. J. Electron. Inform. Technol. 35(11), 2278–2279 (2013)Google Scholar