用于末制导的SAR图像多子区实时匹配算法

A Robust Matching Algorithm for SAR Image with Multiple Sub-areas

为实现用于无人飞行器末制导的SAR实时影像与多源基准影像的匹配,提出一种新的基于子区位置关系约束的SAR图像匹配算法。首先将SAR图像进行对数变换,并进行小波包分解,在高频子图部分对小波包系数进行收缩,抑制噪声,然后利用断面检测法提取边缘。匹配中将SAR实时图分为多个子区,用子区灰度标准差和频谱直方图判别其可匹配性,并更新不可匹配区。将各子区位置分布关系用作匹配结果的约束条件。为保证算法实时性,匹配采用基于提升小波的金字塔分层搜索策略。用星载SAR图像和卫星光学图像进行了大样本匹配实验,实验结果表明该方法是一种能在精度、实时性和可靠性方面都能满足导弹末制导要求的实时匹配算法。

In order to match SAR image to radar reference image and optical one, a novel algorithm of scene matching using multiple sub-area was presented. The image was logarithmicaly transformed and was decomposed into multiple sub-band images with wavelet packet transform. The high frequency coefficients shrink according the relationship between itself and the one nearby, which reduce the speckle noise in SAR data. And the edge feature was then extracted via contour detecting algorithm. The SAR image was divided into multiple sub-areas in matching process. The feasibility for each sub-area was evaluated by computing the gray level standard deviation and the distance of spectral histogram, and the sub-areas can not be matched to reference image were updated. The spatial relationship of the sub-area was used as constraint conditions to correct the result. The multiple-layer search strategy based on hfting wavelet transform was used to improve the performance of the algorithm. Large-sample data experiment has been conducted with satellite SAR reference image and satellite optical one. The experiment result shows that the algorithm is effective, robust and suitable, and it is able to meet these requirements of terminal guidance for missile.