The Application Of The Genetic Neural Network Algorithm For Detecting Littoral Oil-Spills By Laser Remote Sensing

In this paper a Laser Fluorescence hybrid algorithm model, which is based on flexible nonlinear Genetic Neural Network algorithm, is applied for multi-spectral data analysis and differentiating between classes of oil and water surface.By integrating the genetic algorithm with the error back propagation neural network learning algorithm, We make full use of the optimization ability of genetic algorithm in global data area, avoid the problem of being plunged in the part data area and optimize the weights of the neural network . At.the same time, the extensive ability of the neural network decreases the dependence of genetic algorithm on the size of group, the probability of intercrossing and variation, etc. On the other hand, by designing the genetic neural network model, we can accelerate the speed of calculation and increase precision.In the comparative test, we take 30 groups of laser flurorescence samples as the input vector, which are uncertain littoral material of 64 channels. Then we take 10 types of spectrum data of purified materials as target vector, The datas metioned above were used to train BP neural network and Genetic neural network respectively, and gets the precise results that the author expects. From this test, we can get the conclusion: the calculation result of genetic neural network model is much presiciser than that of the general neural network algorithm. GANN algorithm is an effective and advanced method of classifying littoral Oil-Spills.