基于视觉特征的SAR图像评估方法  

A SAR Image Validation Method Based on Visual Features

作　　者：肖舒文[1] 李柏文[1] 张川[1] 陈晓盼[1] 

机构地区：[1]中国国防科技信息中心,北京100142

出　　处：《现代雷达》2016年第4期15-19,49,共6页Modern Radar

摘　　要：为了验证不同算法、模型、软件得到的仿真SAR图像的可靠性,文中提出一种基于专家视觉的SAR图像评估方法,从原始图像中提取视觉关注的目标亮度、轮廓和散射中心分布特征,利用极化映射积分得到差异度量指标,设计合成函数结合三种指标,获得可以准确反映SAR图像差异水平的评估结果。实例验证表明,该方法的性能明显优于现有方法,与专家视觉结果之间的平均误差仅为4.32%,两种结果曲线的相似性高达99.78%。In order to validate the reliability of SAR simulation images from different algorithm; model and software, SAR image validation method is proposed based on experts＇ opinion. Firstly, the target＇s brightness, outline and distribution of scattering centers are extracted from original image. Secondly, these three features are transformed into metrics by polar mapping integral. Finally, the composite function is designed to get the accurate validation results. Experimental results show that the mean difference be- tween this method and the experts＇ opinion is only 4.32 %, and the correlation of two results＇ curves is up to 99.78 %.

关 键 词：雷达目标特性 SAR图像 图像评估方法 视觉特征 极化映射 

分 类 号：TN957.52[电子电信—信号与信息处理]