When extracting coastal zone tidal flats using remote sensing transient images, the influence of tides greatly limits the accuracy of tidal flat spatial distribution extraction. In this paper, based on the Google Earth Engine (GEE) cloud platform, a coastal zone tidal flats extraction method by combining the time-series Sentinel-2 image and the tidal flats index is proposed. First, based on the Sentinel-2 time-series image data, we use the quantile synthesis method to generate high- and low-tide images, and then analyze the spectral reflectance characteristics of different land classes on the high- and low-tide images. A NIR-band tidal flat extraction index that excludes the interference of the tidal transient is constructed. Secondly, the image spectral information and the tidal flat extraction index are input into a machine learning algorithm to realize fast and efficient extraction of the tidal flat. Finally, the separability of the tidal flats index and the universality are investigated. The results show that the tidal flats extraction index constructed in this research had a good separability for tidal flats, the overall accuracy of tidal flats extraction was 93.02%, the Kappa coefficient was 0.86, and the proposed method has good applicability to remote sensing images containing near-infrared bands. This method can realize automatic and rapid tidal flat extraction, and provide data support for the sustainable management and protection of coastal zone resources.