An unmanned aerial vehicle vision navigation landing system based on a multi-scale marker comprises the multi-scale marker, an identification and attitude resolution algorithm and a navigation system. The multi-scale marker uses an image center as a world coordinate system origin (0,0,0). The side length of the origin is one pixel. The origin extends in four directions around the pixel. In every outward extension, the side length of a black matrix becomes two times of the original side length. The multi-scale marker is automatically generated according to the pixel dimension and the integral dimension of the marker. Supposing the world coordinate system is at the central point of the marker, the z-axis coordinate of the point on the marker is zero and the central coordinate of each matrix is known; a rotational translation matrix of a camera relative to the marker is calculated. The navigation system comprises a coordinate system calibration module, an image acquisition module and an operational processing module. The operational processing module is used for target identification, attitude resolution, navigation control and communication with an unmanned aerial vehicle flight controller. The unmanned aerial vehicle vision navigation landing system has advantages of high dimension change adaptability, high attitude resolution capability and relatively high expandability.