方向矩异源图像匹配算法

全天候高精度导航与制导需要可靠的可见光与红外、可见光与合成孔径雷达(Synthetic Apeture Radar,SAR)的异源图像匹配算法。提出了一种基于方向矩的异源图像匹配算法:利用像素层次设计异源不变特征,描述像素点与其邻域各方向相似程度,通过统计比较各像素的相似性测度,实现不同传感器图像间的可靠匹配。进行了批量图像匹配实验,在可见光与红外、可见光与SAR异源图像间取得了超过90%的匹配正确率。较传统异源匹配算法,该算法大大提高了匹配正确率,在导航与制导领域有广阔的应用前景。

Multimodal image registration based on orientation-moment

Multimodal image registration between visible, infrared and SAR images is the key technology to all-weather high-precision navigation and guidance. Based on an invariant feature designed at the pixel level, the orientation-moment, which describes the similarity of a pixel with its neighbors in different orientation, a multimodal image registration algorithm, that matches multisensor images by analyzing such similarities, is proposed. In matching tests, including visible, infrared and SAR images, it has reached a success rate of more than 90%. Comparing with traditional multimodal image registration algorithms, this method greatly improved the success rate, and had broad application prospects in navigation and guidance.