In order to speed up the mark matching under the condition of rotating and scaling,a method for shape matching based on sparse model is proposed using Hausdorff distance as the matching distance. This method has two stages:a sparse model is used in rough matching with a large distance threshold; a full model is used in precise matching with dual thresholds which are a moderately small distance threshold and a large coincidence rate threshold. An experiment is conducted,in which marks imposed on a map are tested,and shorter matching time is obtained with low false negative rate and low false positive rate. This method has been successfully used in military mark recognition in maps.