A method and system for sub-pixel positioning of small-scale surface features based on image aggregation degree. The method comprises: S1, randomly allocating the abundance value calculated accordingto a mixed pixel decomposition model to a sub-pixel class in a test remote sensing image to obtain a sub-pixel initialization result; S2, establishing an image aggregation degree objective function based on that prior information and the image aggregation degree of the sub-pixel of the sub-pixel position result; S3, optimizing the objective function in S2; S4, traversing all that mixed pixel in the sub-pixel initialization result, and executing step S3 for each mixed pixel; S5, using an iterative algorithm and repeating the step S3-the step S4. The positioning accuracy of sub-pixels is higherthan that of a K-mean hard classification method and the mapping accuracy and effect of a pixel exchange sub-pixel positioning model are both improved, and a new way is provided to break through thelimitation of spatial resolution of remote sensing image.