A blind watermark detecting model was proposed to reduce the false alarm rate and the miss alarm rate on detecting. With Matlab software,image samples were detected by 7 different digital watermark detecting mothods one by one, whose detected results were analysed by data fusion method to discriminate the watermark information in the images.Compared by a single algorithm, the pan methods discussed can shorten testing time and improve dectecting efficiency. When 17 digital watermark embedded systems were detected in the simulation experiment, the false alarm rate was 5.495%，the miss alarm rate was 1.648%，the validity rate reached 92.857%. In addition, the methods discussed is less vulnerable to attacks of certain. In a word, the methods discussed has more feasibility than a single algorithm.
