科研之友
微信
新浪微博
Facebook
分享链接
摘要
A binary segmentation scheme, based on the Markov random field theory, is presented. In order to obtain a more integrated label field, the simulated annealing schedule is modified for performing a joint conditional estimation of model parameters. To reach a finer detection, the pixel neighborhood system of the a priori model is continuously updated at each cycle of the optimization algorithm. Maximum a posteriori is the central criterion of these algorithms. The proposed processing scheme is applied to a sequence of Envisat/ASAR images of the Deepwater Horizon disaster of the Gulf of Mexico in the spring of 2010. Initial oil spills statistical parameters are extracted by visual analysis, but they are updated during the minimization cycles. The proposed scheme, when compared with a conventional Markov random field one, provides a better detection of fine structures. In addition, facing the complex ocean phenomena reflected in the synthetic aperture radar images, the final label field results are extremely well defined.
- 出版日期2014-9-9
