Ocean rainfall has an important impact on the global atmospheric cycle and local climate. Monitoring rain cells from remote sensing images is vital for ocean weather prediction. The ability of Synthetic Aperture Radar (SAR) to probe with a wide swath and high spatial resolution makes it an effective observation approach for rain cells with a scale of 10—30 km. This study uses the fusion-feature-based Broad Learning System (BLS) to detect the rain cells. The SAR images dataset composed of nine sea surface phenomena obtained by Sentinel-1 wave mode is also used. Results show that the detection accuracy is 98.51%, and the recall rate is 95.24%. These values are equivalent to those of the ResNet50 pretrained model. However, the training time of ResNet50 is 20 times that of BLS under the same calculation conditions. Compared with the structure of a deep learning network, that of BLS is flexible. That is, the model can be optimized and updated by adding nodes or input data. The experiments show that the node incremental learning of BLS can update the model without retraining the whole model. Following the advantages of the incremental learning and retraining schemes, this study proposed a hybrid model-updating scheme for the model-updating task caused by the expansion of the training dataset. This new scheme can ensure the high accuracy of the model and significantly reduce the time cost for model updating.