Abstract: Based on the fundamental principle of feature recognition of water flow images, an emerging non-contact river surface flow velocity estimation method is developed in this study by integrating image acquisition, image pre-processing, implicit mapping between class labels and flow velocity, and data analysis. This method is verified using a group sparse representation classier with feature constraints (GSCFC), a new parameter introduced by the authors for image analysis and estimation of the surface flow velocity of the Jiepai River. And we evaluate GSCFC by conducting comparative experiments with the classical sparse representation based classifier (SRC) and the regularized robust coding classifier (RRC). Results show that GSCFC is a robust discriminative classifier that has an excellent performance for water flow image and outperforms SRC and RRC.