Research On Road Extraction Algorithm In SAR Image

Synthetic Aperture Radar (SAR) is a kind of high-resolution imaging radar and has abilities of all-weather, multi-polarization which has development rapidly recent years. It has huge signification to auto extract information of linear terrestrial object such as road, bridge, river and coastline from SAR image using computer in SAR image analysis. It hasn't had a set of perfect functional system to accomplish the wish by reason of the increasing complexity which caused by durative extending of city.This thesis proposes a new auto road net detecting method based on the research of the characteristics of the SAR image data according serious of experimentations. The methods which combines MRF with iterative algorithms basically go through the steps such as speckle denoising, detection of road central line, grouping of road segments etc. (1) An amelioration algorithm which based on double- variable model is proposed. In this method, logarithmically transformation applied to original speckled SAR image, followed by redundant wavelet transformation. Then the wavelet coefficient is estimated using the relationship between the father and son high frequency parts. According to experiment, it can achieve preferable effect both on the smoothness of even area and the holding of brim and preparing well for the subsequent work.(2) A new method is proposed which has the steps of speckle denoising– segmentation of the image– grouping of road segments. It uses the MRF-MAP modeling and Iterated Conditional Modes to get the road central line, followed dealing with the morphologic method to get the vector description. Then group the image again using simulated annealing algorithm to get road network. The method which is feasible and adaptable can achieve legible road segment.