In high-latitude oceans, extreme weather conditions and complex environments make it difficult to obtain large-scale ocean color data through in situ measurements, such as buoy and shipboard measurements, especially during the winter season. Ocean color satellites provide the only means of monitoring marine ecological environments at regional and global scales. However, due to the reduced accuracy of atmospheric correction models under large solar zenith angles, publicly-available datasets currently lack ocean color data for the high-latitude oceans during the winter season. In this study, we developed a dataset of ocean color products for the high-latitude oceans of the Southern Hemisphere (≥50°S) from 2003 to 2020 based on a newly-developed atmospheric correction model. We validated the accuracy of the retrieved chlorophyll concentration products using in situ datasets from the NASA Bio-Optical Marine Algorithm Dataset (NOMAD) and the Chinese Antarctic Research Expedition (CHINARE). The results indicated that our dataset had more reliable and accurate records compared to the products distributed by NASA. For instance, in the 412 and 667 nm bands, the NN-LAT50 dataset exhibited absolute percentage deviation (APD) values of 24.1% and 38.1%, respectively, whereas the NIR model had APD values of 33.1% and 52.5%, respectively. Furthermore, the NN-LAT50 dataset demonstrated an effective data volume exceeding 16% of that of NASA’s dataset in multiple wavelength bands. This study has established a dataset that, for the first time, covers a long time series (2003-2020) of ocean color products during the winter season in the high-latitude oceans of the Southern Hemisphere, which should be helpful for the study of the marine ecological environments of the Antarctic and sub-Antarctic oceans.