Traditional change detection systems yield a binary description: change or nochange. However, changes in remote sensing imagery can arise from a multitude of causes such as illumination variation, occlusions by clouds or man-made objects, land cover change or seasonal spectral variation. Yet all of these causes, whether of interest or not to a user, are described in traditional systems as either change or no-change. Thus, to promote better change understanding we implement a statistical model selection framework which describes changes as belonging to particular cause models. The causes: illumination variation, occlusion, and motion, are modeled on a pixel level. For the motion model we implement an occlusion-adaptive content-based multi-spectral algorithm which utilizes meshes. This model extends the video compression work of Altunbasak to multi-band data, implementing more robust occlusion adaptivity and an improved motion estimation algorithm utilizing an hexagonal matching-based algorithm[2] extended for arbitrary mesh topologies. This work serves as a foundation for an object-based change understanding system and addresses trusts R2B, R2C and R2D.


Poster presented at the 2006 Thrust R2D Image Understanding and Sensor Fusion Methods Conference


Multi Band Remote Sensing Imagery, topology

Subject Categories

Remote sensing




Bernard M. Gordon Center for Subsurface Sensing and Imaging Systems (Gordon-CenSSIS)

Publication Date


Rights Holder

Bernard M. Gordon Center for Subsurface Sensing and Imaging Systems (Gordon-CenSSIS)

Click button above to open, or right-click to save.

Included in

Engineering Commons