Change Detection in Multitemporal Hyperspectral Images

Multitemporal hyperspectral images provide very detailed spectral information that directly relates to land surface composition. This results in the potential detection of more spectral changes than those visible in the traditional multispectral images. However, the process of extracting changes from hyperspectral images is very complex. This chapter addresses the multiple-change detection problem in multitemporal hyperspectral remote sensing images by analyzing the complexity of this task. An analysis of the concept of “change” is given from the perspective of pixel spectral behaviors, in order to formalize the considered problem. A hierarchical change-detection approach is presented, which aims to identify the possible changes occurred between a pair of hyperspectral images. Changes having discriminable spectral behaviors in hyperspectral images are identified hierarchically by following a top-down structure in an unsupervised way. Experimental results obtained on simulated and real bi-temporal images confirm the validity of the proposed hierarchical change detection approach.