This chapter introduces a new method to classify the SAR oil spill images. That is Deep Belief Network (DBN). Through the experimental certification, it is shown that the SAR images' information extracted by Gray-Level Co-occurrence Matrix (GLCM) can have a better effect in classification then that extracted by Gabor wavelet features. And using DBN to classify 240 samples including oil slick, looks-like oil slick and seawater, we can reach high total classification accuracy up to 91.25 %. Finally, we get a result that the method of DBN with GLCM features can better meet the needs of the SAR oil spill images classification.
展开▼
