Research On Targets Extraction In Remote Sensing Based On Active Contour Model

Active contour model(ACM)has become a hot spot field in computer vision,motion tracking,medical image recognition.Unlike Marr's theroy,ACM is a fully using image processing method for advanced information,that image segmentation,target detection and knowledge used for identify are in a unified framework to discuss.Compared to general image,high resolution remote sensing image contains more information,have target region without uniform strength,background is more complex,use of earlier geometry ACM has some issues,such as easy to lost target edge,fall into local optimal,it has poor anti-noise,convergence results are sensitive to the initial curve.so there is a need for research,improved algorithm of image processing for remote sensing.This paper study on the starting from the parameters ACM,to deformation contour can't converge to deep recesses of the target,introduceing gradient vector flow(GVF)field is necessery.GVF can expands the scope of a curve captured by the gradient force,so the splines far from the target can converge to the target edge.Rearsch on contour extraction for C-V model,when target and the background gray value is similar,images force are no longer available,and contour is easy to fall into local optimal.This paper propose to added to target median values for region statistic.,and combined geotesic contour model,improve the edge stop function based on edge gradient,construct a globally signed regional pressure function The new model has a strong adaptability for the image.is robust for contour initialization and ROI with weak edge.This model doesn't casue segmentation errors even though the target region contains a variety of strength class.Meanwhile,add an energy functional based on shape prior knowledge into C-V model,so the new C-V model can extract the desired one of targets with variety strength class in images.Experiments show that new sign pressure function improves dynamic performance of contour convergence,improves efficiency;by incorporating prior knowledge energy,improve the extraction accuracy.