The invention discloses a linearity common recognition integration fusion classification method based on space spectrum multi-characteristic limit learning. The method comprises steps that spectrum characteristics, space characteristics and geometric structure characteristics of a high spectral image are extracted; a condition type probability limit learning machine is taken as a classifier, the spectrum characteristics, the space characteristics and the geometric structure characteristics are respectively processed to acquire a set of classification labels and probability distribution; a linearity common recognition integration model is established, and the classification result acquired based on different characteristics is analyzed to acquire the final classification result. The methodis advantaged in that the spectrum characteristics, the space characteristics and the geometric structure characteristics are utilized, the condition type probability limit learning machine is definedas the classifier to acquire the classification labels and probability distribution, the linear common recognition integration fusion model is lastly established, the action effects of different characteristics on different regional picture elements are fully considered, and the better classification result can be acquired.