Provided are a feature extraction and classification method of a hyperspectral remote sensing image. The method comprises steps of: a sampling step, a binary step, a coding step, a counting step, a serially connecting step and a classification step. According to the invention, a two-dimensional LBP is expanded to a three-dimensional LBP, a space-spectral context relation in the hyperspectral remote sensing image is fully used, and a loosen threshold value distinguishing operation is introduced, therefore good robustness against noise is achieved. For the related three-dimensional LBP model, substantive features of the hyperspectral remote sensing image are considered, and the feature extraction and classification method and the system are advantaged by high pertinence, simple operation and high calculating efficiency.
展开▼