基于不变矩特征匹配的目标定位方法(英文)

Target location method based on invariable moment feature matching

红外成像匹配辅助导航系统中由于基准图像和实时图像是从不同的观测角度,在不同时间、不同高度条件下获取的,实时图像与基准图像之间会产生仿射变换与像质差异,增加了成像匹配的难度。为了提高实时图像与基准图像的匹配正确率,提出了基于改进不变矩特征的成像匹配方法。从红外成像的基本原理出发,重点讨论了实时红外图像地面分辨率不同对统计不变矩特征不变性的影响,并从理论上推导出了6维不变矩特征向量的表达式。综合得出,在一定的条件下,该特征向量对平移、旋转、加噪、分辨率变化等多种复合变换下的图像具有不变性,可以作为成像匹配定位应用的特征量。文中选用Camberra距离作为实时图像和基准图像匹配测度,仿真实验部分验证了方法的可行性和正确性。

In infrared imaging matching navigating system, the image distorted transformations between reference images and real-time images can increase the difficulty of imaging matching, for two kinds of images are taken unsimultaneously at different heights and different view points. Based on the basic principles of infrared imaging, effect of the difference in the ground resolution of infrared imaging on the invariability of invariable moment features is discussed in this paper. The expression of six-dimension invariable moment feature vectors of images in different ground resolutions is derived. Discussed results show that the feature vector is invariable for images in complex transformations on translation, rotation, noise adding and differences in resolution, and can be used as the features in the application of imaging matching location. The Camberra distance as the image matching metric between real-time image and reference image is proposed. The method is proved correct and feasible in experiment part.