The invention discloses a method for carrying out terrain classification and monomer extraction on oblique photography results. For an oblique photography result, the method is combined with an unsupervised deep feature learning method to fully automatically identify ground objects and extract ground object features to quickly classify the ground objects, and accurate monomer extraction of an oblique photography live-action three-dimensional model is realized by utilizing an accurate classification result. The method has the advantages that the existing oblique photography technology and achievements are fully utilized, image information of multiple view angles is obtained from the characteristics of geometric structure characteristics, texture characteristics, spatial information and thelike, and the conditions of a characteristic extraction layer are enriched; by combining with an unsupervised deep feature learning method, the processing speed does not depend on the size of an imageany more, but can fully utilize computer hardware resources to realize high-speed parallel calculation of mass data, so that the workload of manual participation is greatly weakened, the calculationand matching accuracy is improved, and automatic and quick ground object classification and precise monomer extraction are realized.