Active grid for feature extraction in remote sensing imagery

Yu Zhang, Xiaodong Zhang · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2009

In this paper, an enhanced active contour, namely active grid, is proposed and its application in extracting boundaries from targets which have adjacency topology, such as buildings, fields, road and river etc, is further introduced in more details. In fact, boundaries extraction by general active contour may be regarded as one kind of simplified active grid, which contains only single region. The basic principle of active grid is: an initial grid given by either manual or automatic methods in order to provide the correct topological structure and a starting grid. Then, based on an iterative algorithm, this grid can evolve according to the information of local image. Nodes (or boundaries) of grid are driven by forces of statistics from different grid region in target image. The evolvement of active grid can be represented by an energy function, as a result, the extreme value of energy is corresponding to the correct extraction of target. The experiment results indicate that the active grid is very effective in target extraction from imagery.

Read the paper · More papers on PaperTik