Clouds are important in ocean remote sensing detection, and cloud detection accuracy is significant to the inversion of cloud microphysical characteristics over ocean and the observation of ocean waterbody. By using the on-orbit imaging data of the atmospheric aerosol multi-angle polarization detector (directional polarimetric camera, DPC) onboard GF-5, a detection method for cloud cover over the ocean based on multi-angle polarization radiation information is proposed. Firstly, the sunlight angle discrimination method is used to distinguish the ocean glint area and the non-glint area. Then the multi-angle polarization information is used to detect cloud in the ocean glint area, and on the other hand,the methods of reflectance threshold, the visible-near infrared reflectance ratio and the polarization information inspection are used to detect cloud in the non-glint aeras. Taking the Indian Ocean and the Atlantic Ocean as experiment aeras, cloud detection over the ocean was carried out based on DPC data, then the results were compared with MODIS (Moderate-Resolution Imaging Spectroradiometer) cloud mask data, and the consistency reaches 91.39% and 94.73% respectively. In addition, the data of FrenchPOLDER3 (polarization and directionality of the earth′s reflectances) is also used to verify the effectiveness of the proposed algorithm in this work, and the consistency between the cloud detection results and the official POLDER3 cloud detection productsreaches 90.40%. It is believed that the multi-angle polarization radiation threshold cloud detection algorithm proposed in this work can provide effective cloud detection data for satellite observation of cloud characteristics over the ocean and ocean water observation.