Topological graph scene recognition method and device based on density filtering and landmark saliency

The invention discloses a topological graph scene recognition method and device based on density filtering and landmark saliency, effectively solves the scene recognition problem under the view angle change, extracts SIFT key points from an obtained landmark on the basis of a target detection algorithm, makes full use of the characteristic that the SIFT key points have robustness to the view angle change, and improves the scene recognition efficiency. The method comprises the following steps of: acquiring landmarks with view angle invariance by adopting a density filtering algorithm, then performing cross authentication on depth global descriptors of the landmarks with view angle invariance in a query frame and a reference frame and comparing shape scores of the landmarks to obtain mutually matched landmarks in the two frames, and considering that the extracted landmarks only represent a small part of an image; some landmarks with low identification degree may cause confusion, thereby generating negative effe