I have been working for
Dr. Melba Crawford as a Research Assistant in Center for Space Research since '02, while
Dr. Joydeep Ghosh of Dept. of Electrical and Computer Engineering is my co-supervisor. My main research focus is on developing classification algorithms that can be applied on remote sensing data. My approach starts with ensemble methods that includes bagging, boosting, random subspace methods and eventually random forest method. Those ensemble method approaches utilize a base classifier called Binary Hierarchical Classifier (BHC), developed by Shailesh Kumar. Ensemble methods gives good results but they are also computationally costly due to the large number of classifiers required in the ensemble.
My research interest move from using BHC as the base classifier to Support Vector Machines (SVM) a binary classifier later on. Basic studies shows that SVM support high classification accuracies on remote sensing data if the parameters are properly selected. A new class decomposition algorithm- HSVM is presented in this web site and IGARSS '04 conference to show that SVM can be easily applied with a minimum amount of tuning while give high classification accuracies and good generalization. Since HSVM is working great on both remote sensing data and UCI machine learning repository data, I moved on.
The new study that I am working on right now is multi-sensor problems. Originally working on hyperspectral sensors that give reflectance of the ground truth. New LIDAR support a new set of information about the structure of the ground truth like vegetation hight, canopy size.. etc. My new approach will extract the structure information from LIDAR and utilize it jointly with hyperspectral data.