Analysis of Multi-Temporal Series and Change Detection

Change detection refers to the process of highlighting and placing the alterations of the state of a scene's surface, having been observed in several images. This chapter describes and characterizes techniques allowing us to detect these changes. Registration between several images is based on distance measurements. The integrity and the compatibility of the data before their comparison also need to be considered on a radiometric level. It is related to the calibration stage of the data, which can address atmospheric correction, directional correction or inter-calibration. The thematic context inherent to 2-date change detection is therefore associated with security, crisis management and rapid cartography. The methodological requirements therefore induce the definition of measurement tools and relatively simple changes from an algorithmic viewpoint. Besides the orthogonal empirical functions (OEFs), there are few methods that have resorted to the spectral analysis of the time series.