The real-time linearly constrained minimum variance（LCMV）detection and classification method for hyperspectral imagery is based on the pixel-by-pixel processing, which has the problems of large amount of computation and slow running speed. Two novel real-time LCMV detection and classification methods based on the LCMV detection and classification method are proposed. Firstly, the LCMV algorithm is carried out causality, a causal real-time LCMV (CR-LCMV) detection and classification method based on the line-by-line processing is proposed. Then, by using Woodbury lemma, a recursive causal real-time LCMV (RCR-LCMV) detection and classification method based on the line-by-line processing is derived. Experimental results show that compared with the traditional LCMV detection and classification algorithm, the two novel real-time algorithms can detect and classify targets in real-time without affecting the detection accuracy, and the required data storage space is greatly reduced. Compared with the real-time LCMV algorithm based on the pixel-by-pixel processing, the real-time processing ability of the two novel real-time algorithms is much strong without affecting the classification accuracy, which has obvious superiority in running time.