论文部分内容阅读
针对逐像元处理的高光谱图像实时线性约束最小方差（Linearly Constrained Minimum Variance，LCMV）算法计算量大，运行速度过慢的问题，在LCMV检测与分类算法的基础上，提出了二种逐行的实时LCMV目标检测与分类算法。首先对LCMV算法进行了因果化，提出了逐行处理的实时因果LCMV（Causal Real-timeLCMV，CR-LCMV）检测与分类算法，再利用Woodbury引理，推导出了逐行处理的实时递归因果LCMV（RecursiveCR-LCMV，RCR-LCMV）检测与分类算法。实验结果表明:（1）与LCMV检测与分类算法相比，两种新型实时算法均能在不影响检测精度的情况下实时地检测目标与对目标进行分类，且所需的数据存储空间大大降低。（2）与逐像元处理的实时LCMV算法相比，两种新型实时算法可获得几乎与之相同的检测精度，计算复杂度大大降低，实时处理能力更强，算法在运行时间上具有明显的优越性。
According to LCMV detection and classification algorithm, two kinds of linearly constrained Minimum Variance (LCMV) algorithms for image-by-pixel hyperspectral image processing are calculated in large quantities and run slowly. Real-time LCMV target detection and classification algorithm. Firstly, the LCMV algorithm has been made into a causal model. A real-time causal LCMV (Causal Real-time LCMV) detection and classification algorithm is proposed. By using Woodbury’s lemma, a real-time recursive causal LCMV (RecursiveCR-LCMV, RCR-LCMV) detection and classification algorithm. The experimental results show that: (1) Compared with the LCMV detection and classification algorithms, the two new real-time algorithms can both detect and classify targets in real time without affecting the detection accuracy, and the required data storage space is greatly reduced . (2) Compared with the real-time LCMV algorithm, the two new real-time algorithms can obtain almost the same detection accuracy, the computational complexity is greatly reduced, the real-time processing capability is stronger, and the algorithm has obvious Superiority.