The invention discloses a PCANet image-change-detection method based on significance analysis. The method mainly solves problems of too many training samples and too long processing time caused by anexisting Automatic PCANet method and impacts generated by scattering noises of SAR images on classification results, and includes the realization steps of: acquiring a difference graph for two-time-phase SAR images; carrying out significance analysis on the difference graph; utilizing a threshold method to extract a significant region, and reusing the threshold method to obtain positive samples, negative samples and undetermined pixels; and utilizing PCANet to extract features corresponding to the positive and negative samples and the undetermined pixels, using features, which correspond to the positive and negative samples, to train a support vector machine (SVM) module, and then inputting the features of the undetermined pixels into the SVM for final classification. Compared with the existing Automatic PCANet method, the method of the invention is high in detection precision and operation efficiency and good in anti-noise performance, and can be used for SAR-image change detection.