Most modern surveillance systems make use of pan-tilt-zoom (PTZ) cameras.Howev er,due to the influence of many factors,the (pan and tilt) coordinates reported by the PT Z cameras become inaccurate after many hours of operation.As the existing parameter correction method is not adaptable enough for large scale images and provides results with low accuracy,a parameter correction algor ithm based on parameter refinement strategy and hierarchical feature matching is presented in this paper .In the offli ne stage,the parameter refinement strategy is introduced to compute the pose angle of each 3D point in the surveillance scene, which can effectively improve the accuracy of parameter correction.In the onlin e stage,a hierarchical matching and correspondence propagation method is designed to generate pairwise correspondence between observed image and pose angle set,which can effectively improve the adaptive ca pacity of parameter correction algorithm for different scale images.Experimental results demonstrate that the proposed method can accurately correct the parameters of the PTZ camera,and it has better accur acy compared with the classical method.