OBJECT-ORIENTED CHANGE DETECTION FROM MULTI-TEMPORAL REMOTELY SENSED IMAGES

An object-oriented change detection algorithm was proposed and experimented to multi-temporal ALOS remotely sensed images. In contrast with conventional pixel-based algorithms, this approach processed homogenous blocks generated by object-oriented image segmentation for change detection. Similar pixels were merged into homogeneous objects by image segmentation at first, and each object (polygon) was described using spectral, texture, shape and other features, which were then exported as the .shp files of ArcGIS. Two segmented polygon layers from two-date remote sensed images were overlapped to create a new polygon layer, and attribute operations to each polygon in the overlapped layer were conducted to determine the thresholds and find those changed polygons. This object-oriented change detection approach was used to land cover change detection over the urban area and mining area of Xuzhou city, and the experimental results indicate that the proposed method is effective to change detection from multi-temporal ALOS satellite images.