中文摘要：

基于尺度不变特征算法,提出了一种适用于弹载合成孔径雷达图像匹配的改进算法。该算法首先通过在构建的高斯差分尺度空间中搜索特征点,然后利用指数加权均值比算子计算特征点的梯度幅值和方向,并通过简化的特征描述子生成特征向量,最后采用距离比和几何一致性准则剔除错误匹配的特征点对,实现图像匹配。实验结果表明,改进算法对SAR图像在旋转、光照、尺度变化等情况下都有很好的匹配效果,并且降低了计算复杂度,有较强的鲁棒性和实时性。

英文摘要：

An image matching algorithm based on the scale-invariant feature transform for missile-borne synthetic aperture radar （SAR） is proposed. First, the difference of Gaussian （DOG） scale space is constructed, and features are detected in the space. Then the gradient is computed with the ratio of exponentially weighted averages. Finally, the distance ratio and consistency checking are used to remove mismatch points and establish matches. Experiments on SAR images with various rotation, light and scale demonstrate the applicability of the improved algorithm to find feature matches for missile-borne synthetic aperture radar image matching, and also reduce the time complexity. The results show that the improved algorithm has strong robustness and real-time performance.