Stereological Estimation of Uncertain and Changing Objects from Remote Sensing Image Mining
This article combines stereology with image mining. Image mining identifies and models objects from a series of remote sensing images and communicates this information to stakeholders. Stereology is the science of deriving properties of objects from lower dimensional features. This article first applies stereology to quantify properties of crisp objects on single images. Next it addresses the development of an object in space and time. Finally, it quantifies uncertainty of fuzzy objects. The article is illustrated with a case study from Cambodia, where flooding regularly occurs along the Mekong River. Nine Landsat images have been mined to assess the size of the flooding in space and time. Areal estimates obtained with stereology from single images show a high precision. Estimates of a space–time volume of the size of flooding in space and time include uncertainty estimates that could be ascribed to atmospheric distortion and limited resolution. Finally stereology is applied to estimate the effects on area estimates of fuzzified boundaries. All information can be transferred to stakeholders, e.g. to quantify the size of a disaster. The article concludes that stereology successfully and concisely quantifies phenomena and their uncertainties in a remote sensing image mining context.
Document Type: Research Article
Affiliations: Geoinformation CentreGeological Survey of Ethiopia
Publication date: 2010-08-01