An Analysis of the Auto Detection Algorithm of High Resolution SAR Image under Fuzzy Semantic Framework
Abstract
High-resolution SAR, as a coherent imaging radar, has been widely used in marine, national
defense, geological and other fields. Target detection of SAR images is a vital part of various
applications. There are many related detection methods. The resolution of SAR images is
gradually increasing, and the background clutter distribution probability density function cannot
be accurately fitted, and the applicability of traditional detection methods is gradually
decreasing. Therefore, the purpose of this paper is to study a high-resolution SAR image vehicle
target detection algorithm that does not require a background model. First, search the light and
dark areas in the scene; second, extract the semantic features by using the fuzzy rate attribute
function to select bright areas that may be strong scattering of the target and dark areas that may
be occluded by the target, based on spatial semantics The relationship is to match the candidate
light and dark regions, and calculate the degree of membership of the same target under the
premise of matching. Finally, the targets with the degree of membership higher than a preset
threshold are combined and output. The experimental results of this research show that the
algorithm has excellent performance in high-resolution SAR image vehicle target detection, and
the false alarm rate is low under the premise of ensuring the detection rate, and the background
clutter distribution probability density function is not required.