Abstract: Using polar coordinates to optimize the fractal model, a novel segmentation algorithm is proposed to deal with extended target under complex environment. First, analyses is introduced for characteristics of morphology operator with variable algebra structures in polar coordinates, corresponding the dilating operations progress to construct a series of covering surfaces. Secondly the fractal feature, named Scale Error, is extracted by the least square method. Then eroding filters eliminate clutter texture. After the course of edge linking, the region with the biggest area is preserved as a target based on transcendental knowledge. Abundance experiments under complex environment test its validity and the contour of extended target is kept intact and perfect.