SAR图像目标的融合检测方法

A Fusion Method for Target Detection in SAR Image

该文提出了一种利用扩展分形特征和局部对比度特征进行融合的SAR图像目标检测方法。分析了扩展分形特征的尺度敏感性及其在不同目标杂波模型下的二阶统计特性,分析表明扩展分形特征在目标检测中存在负值效应,即在正确检测出目标的同时把一些与目标具有相似形状而灰度值较低的区域也检测出来。而CFAR检测方法只利用了目标的局部对比度信息,不存在负值效应,但在强杂波环境中的检测结果存在很高的虚警。两种方法的融合可以滤除大量杂波虚警而保持目标。实测数据的融合检测结果证明了该方法的有效性。

A method fused by Extended Fractal （EF） feature and local contract feature is proposed for target detection in SAR image. The paper mainly discussed the size sensitivity behavior for the EF feature and the second-order statistics for the EF feature in various target/clutter models, and concluded that the feature is also invariant to negative scalar multiplication of the image in the sense that a deep target-sized shadow can also be detected as well as bright target-sized objects. While the CFAR method only using the local contract information is not symmetric, it has a high false alarm in the strong clutter environment. Fusion of the two features provides an even lower false alarm rate when the targets can be detected. Experiments with real data show the effective of the fusion method.