The invention discloses an oil spill hyperspectral image detection method based on a two-way graph U-NET convolutional network, comprising the following steps: Step 1, performing graph structuring on the data of the hyperspectral image that needs to detect oil spill, and obtaining the spectrum Graph structure information and spatial graph structure information; Step 2, send described spectrogram structure information and spatial graph structure information into one way in the two-way graph U-NET convolutional network respectively, obtain the spectrogram of described hyperspectral image respectively feature and spatial map feature; step 3, fusion of the spectrogram feature and the spatial map feature to obtain the space-spectrum feature; step 4, sending the space-spectrum feature into the classifier to obtain the Classification results of hyperspectral images. The invention can map the European image data to non-European data, more effectively represent spectral information and spatial information, further extract the space-spectral feature of the oil spill hyperspectral image, and improve the detection accuracy of the oil spill hyperspectral image.