Road Detection In Sar Image Study

Synthetic Aperture Radar (SAR) is a kind of high-resolution imaging Radar and has abilities of all-weather, multipolarization, multi-look, multi-angle of depression. It can penetrate through surface features such as smog, cloud and forest and it has developed very fast in recent years. Since SAR has wider observation range than other sensors and the observed object has clearer contour than those images, SAR image can distinguish the near targets better and determine the size of target section more precisely. With the development of modern technology, the high-resolution SAR has syncretized modern computer technology and advanced digital signal technology and is increasingly widely used in many fields such as battlefield scouting, aero-photographing, terrestrial remote sensing, terrestrial resource exploration, mapping, disaster forecast, military command & observation and other national economic fields. In SAR image analysis, some objects such as road, bridge, river and coastline have linear structures. It is significant to autoextract information of linear terrestrial object from SAR image using computer. Linear structure detection can be used in multi-sensor, image-matching, mapping applications, image division and target recognition, etc. It has almost been past twenty years in research on road detection in SAR image home and abroad. The published papers are countless, but it hasn't yet had a set of perfect functional system to accomplish the wish, however, after past twenty years of research on this field home and abroad, it has congregated the pith of several subjects of knowledge and many innovations in theory and technology have come forth. This thesis tries to combine the traditional semiautomatic road detection methods with the new theory such as genetic algorithm and makes great efforts to achieve automatic road network detection. This thesis has researched the road network extraction methods in spaceborne synthetic aperture radar image in detail aiming at the characteristics of the SAR image data. The methods basically go through the steps such as detection of road central line, extraction of road segments, grouping of road segments etc. and finally we get a relatively perfect concrete implement arrangement:...