An approach to data fusion, based on the overlapping of regions from two or more different channels and on a fuzzy reasoning about the result of the overlapping process, is introduced. Resulting regions are labeled as slivers (regions created by imperfect boundary superimposition) and kernels (all remaining regions). Kernels are classified by a matching with multisensor models, while for slivers a more complex reasoning is necessary to take into account their source (misregistration, segmentation error, or temporal change). The region-overlapping technique is described, and its potentialities are assessed. Experimental results on multitemporal TM (Thermatic Mapper) images are shown, including the multichannel segmented image produced by overlapping, and kernel/sliver distinction.
Remote sensing data fusion by means of a region-overlapping technique
ROLI, FABIO;
1991-01-01
Abstract
An approach to data fusion, based on the overlapping of regions from two or more different channels and on a fuzzy reasoning about the result of the overlapping process, is introduced. Resulting regions are labeled as slivers (regions created by imperfect boundary superimposition) and kernels (all remaining regions). Kernels are classified by a matching with multisensor models, while for slivers a more complex reasoning is necessary to take into account their source (misregistration, segmentation error, or temporal change). The region-overlapping technique is described, and its potentialities are assessed. Experimental results on multitemporal TM (Thermatic Mapper) images are shown, including the multichannel segmented image produced by overlapping, and kernel/sliver distinction.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
