A Modeling-Based Threshold Approach To Derive Change/No Change Information Over Vegetation Area

Change detection is an important application of remote sensing in environmental monitoring. A modeling-based threshold approach using the connectivity information of individual pixels (regions) was introduced and compared with the method using standard deviation as the region-based change detection give often better results. This approach was used in combination with four widely used image differencing methods to derive change and no change information over a semi-natural vegetated area in Mediterranean southern France. After comparing the results visually with the aid of a vegetation map and aerial photos, a suitable automatic change detection method was developed: the tasseled cap transformation plus change vector differencing followed by this modeling-based thresholding approach. The dataset used in this study are two Landsat TM/ETM+ scenes of (acquisition dates of the images are 2001 and 1996). A graphical tool is developed in ERDAS/Imagine using EML (Erdas Macro Language) and SML (Spatial Modeling Language) for implementing this automatic change detection algorithm. This study shows that more research is required aiming at integrating digital image processing techniques and the applied use of earth observation