在红外目标识别过程中,针对系统实时性要求高的场合,提出了一种基于角点的特征不变量,证明了该不变量具有平移不变性、旋转不变性和比例不变性。将该不变量作为匹配因子,运用动态规划的方法把目标的不变量与模型中的不变量进行匹配,并与三角形边长匹配法进行了比较。实验证明,该方法有效地降低了匹配误差,提高了识别的实时性,效果较好。

Real time property is very important in situations like infrared target recognition. In order to enhance the real time performance, a new feature invariant derived from corner was proposed, which was proved to be invariant to rotation, translation and scaling. The invariant was used in matching infrared images with dynamic programming, and was compared to method with side length of triangle. The experiment result shows it can eliminate errors and enhance real time performance, and the algorithm is effective.