Abstract: Satellite remote sensing provides an effective technical means for accurately extracting information from offshore aquaculture areas, and is of great significance to the realization of scientific supervision of the marine aquaculture industry. Existing optical remote sensing methods mostly focus on the use of image spatial features in the extraction of aquaculture information, and the comprehensive utilization of the combination of spectrum and spatial information in the feature expression and discrimination extraction of the aquaculture area needs to be further explored. In this paper, using the remote sensing image of Sentinel-2 satellite, aiming at the characteristics of small sample, multi-dimension, and complex water composition in offshore aquaculture area, this paper proposes an aquaculture area extraction algorithm that combines spectral and spatial information. Firstly, index calculation is performed to enhance the feature expression ability of the target in the aquaculture area; on this basis, the support vector machine model is used to perform the initial classification of the spectral information; finally, combine the Markov random field model to post-process the initial classification results, and comprehensive use of spectral and spatial information to achieve accurate extraction of offshore aquaculture area information. This method has achieved an overall classification accuracy of 94.46% in the study area, Compared with the method that only uses spectral information and texture feature enhancement, the extraction accuracy of various aquaculture areas has been significantly improved. In the extraction results, the mixing phenomenon of seawater and floating rafts has also been effectively improved. This algorithm model provides a new idea for the automatic extraction of offshore aquaculture areas.