The work aims to propose an adaptive signature mark detection method based on phase correlation regarding the problems of excessively high labor intensity caused by the traditional artificial examination for the normal book association and the error-prone situations. By means of the phase correlation registration technology, the phase correlation between the signature mark video frame captured by the CCD camera and the signature mark template was computed. Moreover, the 2-frame image registration was conducted, and then the frame difference of 2-frame image was analyzed, thus telling whether the association goes wrong. The experimental results showed that the proposed method could reduce the false detection rate for the normal association and it could also detect the abnormal association. The proposed method can improve the robustness of association signature mark.