An approach of bridge detection over water in high resolution SAR image

A new method to detect bridge over water in high resolution SAR images is presented. Firstly, 2-dimensional fuzzy feature is utilized to segment water regions in SAR image. Secondly, tree and hill shadow false targets are excluded based on 8-neighbour chain-code tracking edge method. Then the edge points of potential bridge ROIs (regions of interest) obtained by searching the water area are used to line fitting. In order to avoid the effect of stripe jamming, the fitted line segments are extended to gain the suspected bridge ROIs. Finally, the suspected bridge ROIs are segmented further by double parameter CFAR, and the bridges are detected by the principle axis direction and perimeter extracted. Experimental results show the algorithm can detect the bridges over water accurately, especially the bridges with subtle grey gradient variance or disturbed by stripe jamming in high resolution SAR image.