To utilize GF-6 WFV data more efficiently, the cloud detection algorithm, which is based on cloud detection algorithm-generating (CDAG) algorithm, is investigated in this study. The proposed method can effectively realize high-precision cloud detection of multi-spectral satellite sensors by completely mining the spectral difference information of the cloud and the typical surface in visible and near-infrared bands. Considering that the spectral range of GF-6 WFV is relatively narrow, and the recognition ability of the cloud and the bright surface is relatively weak, we add the dispersion index and bright surface index, and use more band combinations to further analyze the differences between cloud and clear pixels so as to improve the recognition accuracy of typical surface and cloud. Cloud detection results from different sub-regions are varified through remote visual interpretation, which suggests that the overall accuracy reaches 85.16%, 14.84% of clouds are not identified, and 2.39% of the surface is incorrectly identified as clouds, thereby demonstrating the proposed method can achieve high recognition accuracy.