Stereology for multitemporal images with an application to flooding

Abstract

This paper presents stereology for flooded areas observed on a multitemporal remote sensing image. Stereology is a mathematical method to quantify objects at one dimension from simulated objects at a lower dimension. It was initially developed for geological and soil objects. Here it is applied to objects on multitemporal remote sensing images, i.e. for image mining. Image mining considers the chain from object identification from remote sensing images through modeling, tracking a series of images and prediction, towards communication to stakeholders. The paper introduces the estimation of the area size of the same object observed at various moments in time. It is illustrated with a case study on flooding of the Tongle Sap lake in from Cambodia.