Remote Sensing and Classification

I am participating in the research study done by W. Feng who is challenging the classification of imbalance hyperspectral data. Increased performance on some real datasets are obtained by combining in a fruitful manner, existing techniques, such as: using ensemble classification (namely random forest), increasing learning datasets with synthetic data, randomly selecting a varying number of samples with weights set upon modified unsupervised margins.

Publication

[1] W. Feng, W. Huang, G. Dauphin, J. Xia, Y. Quan, H. Ye, and Y. Dong. Ensemble margin based semi-supervised random forest for the classification of hyperspectral image with limited training data. In International Geoscience and Remote Sensing Symposium, IGARSS, pages 1–4, 28 July - 2 August 2019, Yokohama, Japan.

[2] W. Feng, G. Dauphin, W. Huang, Y. Quan, W. Bao, M. Wu, and Q. Li. Dynamic synthetic minority over-sampling technique-based rotation forest for the classification of imbalanced hyperspectral data. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, pages 1–11, 2019.