In the face of the problem that the accuracy of cyanobacteria bloom area in lakes estimated with low and medium resolution images is poor due to the low monitoring scale, this paper took GOCI-2 as an example, selected the Taihu Lake as research area, compared the difference in cyanobacteria bloom area extracted from GOCI-2 and Sentinel-2, researched the relationship between the NDVI of GOCI-2 and the proportion of cyanobacteria bloom area in mixed pixels and conducted regression analysis. Based on this, a corrected model of the cyanobacteria bloom area for GOCI-2 was established. The accuracy of the model was compared and analyzed. The results showed that: a non-linear positive correlation was found between the proportion of cyanobacterial blooms area and NDVI value in mixed pixels, and the area of cyanobacterial blooms extracted directly using the NDVI threshold method was larger than the actual value. After correction by this model, the average accuracy of GOCI-2 cyanobacterial blooms area monitoring was improved from 67.8% to 90.0%. And the model in this article was not sensitive to the changes in NDVI threshold settings. NDVI was found to better reflect the proportion of cyanobacterial blooms in mixed pixels, compared with the algorithm of EVI and AFAI which were used to extract cyanobacterial bloom. The research results of this paper can provide valuable reference for the application of GOCI-2 image in the field of cyanobacteria bloom monitoring.