A new oil spill detection algorithm based on Dempster-Shafer evidence theory

Features of oil spills and look-alikes in polarimetric synthetic aperture radar (SAR) images always play an important role spill detection. Many detection algorithms have been implemented based on these features. Although environmental factors such as wind speed are to distinguish look-alikes, some do not consider the factors. To more accurately image features, a new algorithm Dempster-Shafer evidence theory was proposed. The process taking account modeled using subjective Bayesian model. Faster-region convolutional neural networks (RCNN) model used for convolution results two models were fused at decision level theory. establishment test proposed completed our look-alike sample database that contains 1 798 samples information records related samples. analysis evaluation shows good ability detect higher rate, with identification rate greater than 75% false alarm lower 19% from experiments. A total 12 SAR collected validation algorithm. result has performance detecting overall 70%.