Abstract: With the exploration of the complex and wide-area marine environment, the effective identification of sea ice, enteromorpha, oil spill and other marine targets is of great significance for safeguarding maritime rights and interests. The classification of sea ice and other targets is very challenging due to their diverse manifestations, different physicochemical and optical properties, changes in sea surface light and sea water movement. In this regard, more information of the target could be obtained by means of the unique advantage of the combination of the atlas of airborne hyperspectral images. Three classical classification algorithms, spectral angle matching algorithm (SAM), maximum likelihood classification method (MLC) and support vector machine (SVM), were used to classify the sea ice, entera margin and oil spill targets based on the sea background. The qualitative and quantitative comparisons show that the classification results of SAM algorithm for enteromorpha are poor, the Kappa coefficient is 0.67, and that of MLC algorithm for sea ice and oil spill are poor, the boundary of classification area is fuzzy,while the SVM algorithm has good classification effect, tits overall accuracy, average accuracy and Kappa coefficient are above 0.9, which is relatively stable on the whole.