Linearly Constrained Minimum Variance-Based Real-Time Hyperspectral Image Target Detection Algorithm

摘要

In this paper, a real-time target detection of hyperspectral image processing algorithm is developed, to be called real-time linearly constrained minimum variance (LCMV) algorithm, which not only can be used in hyperspectral image target detection but also can be used in target classification. The only required knowledge of LCMV algorithm is the spectral signatures of targets of interest. It extends the well-known CEM algorithm to detecting multiple targets by a constrained vector. First, the original LCMV was developed to a causal LCMV then to a real-time LCMV, by using Woodbury's equation to updates correlation matrix to solve the problem of re-calculation of new correlation matrix, thus slow down calculation time rapidly. This paper investigates these three versions of LCMV, C-LCMV, and RT-LCMV and also conducts experiments and analysis via synthetic and real image experiments. From the results of the experiments, LCMV classifier in real-time processing can be carried out point by point. Experiments of simulated data and real data proved the results of real-time LCMV.