The analysis of the changes occurring on the ground is one of the main research field in remote sensing. Change detection is widely exploited in different application domains related to the human activities such as urban planning and environmental monitoring (human settlements, deforestation). Natural phenomena such as earthquakes, volcanic eruptions, floods, are continuously happening and the analysis of the changes by using remote sensing data can give a huge support in applications such as damage assessments and updating map.
Change detection (or multi-temporal) analysis can be defined as the process of identifying and quantifying temporal differences in the state of an object or phenomenon (Singh, 1989). The aim of the research is to develop efficient techniques for multi-temporal data analysis, assessing at the same time issues that arise from the analysis, i.e., irregular temporal sampling, seasonal effects and imperfect registration, which occurs even with the most accurate co-registration techniques. In a typical approach for time-series analysis of EO data, a feature is extracted from each image then the temporal trajectory of the feature is analyzed. The goal is to identify multi-temporal signatures of features or processes that can be used to characterize a variable of interest. In general, there are two broad approaches to the change detection problem, namely, unsupervised and supervised. The unsupervised techniques are based on the statistical methods, regression methods or by comparison of certain class-specific indices such as the normalized difference vegetation index (NDVI). Supervised change detection is based on supervised classification of the individual images and the post classification comparison of the classes. While supervised approaches have the advantage that they are capable of explicitly recognizing the land use or land cover changes, they require accurate training set which is normally a very difficult and expensive job.
Spectral and spatial domains are both considered. The new generation of satellite sensors are able to provide images with high geometrical resolution, where structures are represented by groups of pixels. The information related to the context becomes very important in detecting changes related to physical structures, rather than single pixels, providing more meaningful change detection map.