In previous studies on the extraction of cyanobacteria from Lake Taihu, the domestic satellite images of the GF series are seldom used as data sources. So this paper used the random forest(RF) method, normalized vegetation index(NDVI) threshold method and multispectral green tide index(MGTI) threshold method to extract cyanobacteria from Lake Taihu based on GF-6 data during different outbreak periods to seek the best method to extract cyanobacteria from Lake Taihu in 2019. On this basis, the applicability of the RF method based on different input variables extract cyanobacteria were explored. The results showed that cyanobacteria blooms and eutrophication were serious near the southern shore of Lake Taihu and Lake Meiliang in 2019. Some mild cyanobacteria were identified by the NDVI threshold method. The identification ability of moderate and severe cyanobacteria by the MGTI threshold method was not as good as that by the NDVI method, while the RF method could effectively extract cyanobacteria from Lake Taihu. The RF method with normalized vegetation index and normalized water index as input variables was the most accurate method for extracting cyanobacteria from Lake Taihu during the medium outbreak stage. Its overall classification accuracy and Kappa coefficient are 99% and 0.97, respectively. The research on the extraction method of cyanobacteria provides technical support for controlling the environment of Lake Taihu and a scientific basis for monitoring cyanobacteria by remote sensing in other lakes.