SAR图像小波域多尺度模型  

Multiscale Model of SAR Imagery in Wavelet Domain

作　　者：袁小红[1] 朱兆达[1] 

机构地区：[1]南京航空航天大学信息科学与技术学院,南京210016

出　　处：《南京航空航天大学学报》2009年第4期510-515,共6页Journal of Nanjing University of Aeronautics & Astronautics

摘　　要：针对面向ATR应用的SAR图像压缩需要将图像压缩与图像自动分析相结合,本文研究小波域SAR图像MAR模型并应用于图像鉴别。与SAR复图像分辨单元相干平均形成多尺度图像序列建立多尺度模型不同,本文对SAR对数检测图像小波变换形成多尺度图像序列建立MAR模型。通过实例辨识了SAR图像中自然杂波(草地)与人造物(战略目标)模型,应用该模型推导了小波域多分辨率判别,对MSTAR图像鉴别实验验证了模型的有效性。其中db2小波域多分辨率判别鉴别性能好而计算复杂度低。小波域多分辨率判别可用在图像压缩有损量化前鉴别出人造物与自然杂波。Muhiscale model of synthetic aperture radar（SAR） imagery in wavelet domain is used to discriminate man-made objects from the natural clutter under consideration of automatic target recognition （ATR）based SAR image compression. It is based on muhiresolution sequence through wavelet transform on SAR log-detected image, different from those based on muhiresolution sequence through coherent averaging on the complex SAR image. Both natural clutter （grass） and man-made objects （tactical targets） models are identified and a muhiresolution discriminant in the wavelet domain is derived. Tests on moving and stationary target acquisition, and recognition（MSTAR） datasets show that the muhiresolution discriminant in the wavelet domain can distinguish man-made objects and natural clutter in particular the one based on db2 wavelet with low computation complexity. Muhiresolution discriminant in the wavelet domain can be utilized to discriminate man-made objects from the natural clutter before quantization in image compression.

关 键 词：合成孔径雷达 多尺度 小波 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]