Plateau lakes were of great significance in reflecting changes of regional natural environment under the background of global climate change. Based on the new homemade remote sensing data source, Tiangong-2 multi-spectral data, an automatic extraction algorithm for frozen lake was proposed by used object oriented method combined with water index and elevation information. The algorithm took full account of the characteristics of different forms of water, which can simultaneously extract frozen and unfrozen lakes, and exclude the influence of glaciers and rivers. For the selected 7 typical regions, automatic extraction algorithm was used to carry out lake extraction test and accuracy verification. The overall accuracy of lake extraction is 99.10% and F-score is 0.982. The results show that Tiangong-2 multi-spectral data has a significant application potential in the extraction of plateau lakes. As an effective data source, it can be applied to the study of lake extraction and change in the Tibetan Plateau, and provides data support for the study of regional climate change.