The Yutu 2 lunar rover has been engaged in scientific research on the lunar surface for more than 3 years. In order to maximize the efficiency of planning and driving for the Yutu 2, the distance between the beginning and end stations of each movement is large. This makes significant scale differences and complex affine transformations between the images in adjacent stations, which usually causes failure of image matching and navigation. To address the above problems, current methods use the single-dimensional pixel spatial resolution model based on imaging geometry to analyze the reasons of failed or successful images matching. However, they are not adaptable to the images with complex affine transformation. In this paper, a general two-dimensional pixel spatial resolution model based on the algebraic derivation of homography transformation is proposed. With the proposed model, the scale difference between images of adjacent stations can be quantitative analyzed, and rate of successful matching between images which have complex affine transformation and scale differences is increased. This will benefit the engineering application such as visual localization and navigation. Experiments under various conditions such as the horizontal and inclined lunar surface are performed to verify the effectiveness of the proposed method.