International Journal of Computer Science Issues

More than a traditional journal...

Today, many problems in remote sensing are related to change detection; it consists to the characterization and localization of areas that have evolved between two times or two periods of time from two observation or sequences of observation, on the same scene. The induced changes are of different types and origins and various durations. Many methods for change detection from time series of satellite images have been proposed in the literature. These methods can be grouped into two families: the bi-temporal methods and methods of analysis of temporal profiles or multi-temporal methods. We were interested specifically in the bi-temporal method. The latter is applied to detect changes in a geographical area from acquired images at two different times. These methods are based on various algorithms for image processing. We opted to use two different methods which are: the Tasseled Cap transformation and the integration of NDVI (Normalized Difference Vegetation Index) with Tasseled Cap indices. The first approach reduces errors of omission. It consists initially to make a change thresholding on the obtained image after processing. Then, it creates a mask change to extract changes areas and to focus on the classification of these areas. Therefore, the classification accuracy is increased. The integration of the vegetation index NDVI will improve the rate of change in vegetation classes. The reliability of the results is closely related to the method adopted.