Dr. Pei-Gee Ho dissertation: MULTIVARIATE TIME SERIES MODEL BASED SUPPORT VECTOR MACHINE FOR MULTICLASS REMOTE SENSING IMAGE CLASSIFICATION AND REGION SEGMENTATION

Satellite and airborne Remote Sensing for observing the earth surface, land monitoring and geographical information systems control are issues in world?s daily life. The source of information was primarily acquired by imaging sensors and spectroradiometer in remote sensing multi-spectral image stack format. The contextual information between pixels or pixel vectors is characterized by a time series model for image processing in the remote sensing. Due to the nature of remote sensing images such as SAR and TM which are mostly in multi-spectral image stack format, a 2-D Multivariate Vector AR (ARV) time series model with pixel vectors of multiple elements are formulated. To compute the time series ARV system parameter matrix and estimate the error covariance matrix efficiently, a new method based on modern numerical analysis is developed. As for pixel classification, the powerful Support Vector Machine (SVM) kernel based learning machine is applied. The 2-D multivariate time series model is particularly suitable to capture the rich contextual information in single and multiple images at the same time.