Integrated Multitemporal and Multiangular High Spatial Resolution Image Classification

Remote sensing classification procedures have progressed in recent decades to enhance the interpretation of high resolution data. Contemporary research has focused on improving techniques that can analyze the spectral, spatial, temporal, and polarization properties that influence how platforms interpret substance reflectance. However, such methods are focused primarily on examining these properties for single multispectral images and do not consider their importance for evaluating multiimage datasets. To address this gap, the purpose of this research was to develop a systematic classification procedure that analyzes the temporal, spectral, spatial, and anisotropic properties of remote sensing data of a multiimage dataset. This was accomplished by first employing an orthophotography approach to correctly register multiple images of the same study sites that were collected at different dates. The resulting multitemporal image was evaluated using image filtering techniques, factor analysis, principal component analysis, and stepwise discriminant function analysis to analyze the contribution of all bands from each original dataset. The datasets consisted of high spatial resolution images acquired at several dates of a forest environment with mountain pine beetle forest health concerns. The systematic classification procedures developed in this study were compared to results from a traditional maximum likelihood supervised classification on the same dataset. The results show that there is a significant benefit for integrating multitemporal datasets rather than using a single image for classification. This finding is a step forward in augmenting human image interpretation by computerized image analysis.