Oil Identification Research Based On Rough Sets-Genetic BP Network

The impact of oil pollution to the marine environment is more and more serious, real-time and correct monitoring oil spill is great significant. So, after comparing the oil spill recognition algorithms at home and abroad and identifying the reasons of the slower pace and precision on the previous artificial neural network laser fluorescence spectrum, we proposed the improved BP algorithm, that is genetic BP (GABP) algorithm which is based on genetic algorithm, and Rough Set-Genetic BP (RS-GABP) algorithm which is based on rough sets and genetic algorithm.In this paper, we use the error back propagation (BP) network, which is used most extensive in artificial neural network. The point is to build BP, the GABP and RS-GABP laser fluorescence spectrum recognition models, and to use laser spectral data obtained by the remote sensing equipment to test the three experimental models, which are against their recognition speed and accuracy, then get the expected results.After tests, the result comes, that is laser fluorescence spectrum recognition model of BP network's best times for training is 42, and the correct rate of identification is 75%; laser fluorescence spectrum recognition model of GABP network's best training times is 25, and the correct rate of identification is 90%; laser fluorescence spectrum recognition model of RS-GABP network's best training times is 12, and the correct rate of identification is 95%. Whether on the aspects of recognition speed or accuracy, GABP model and RS-GABP model are relatively high, especially the RS-GABP model, but BP network is slightly inferior to the both because of its self-shortcomings.The experimental results show that, RS-GABP laser fluorescence spectrum recognition algorithm's performance is the best; it is a great potential algorithm.