%0 Journal Article %T 基于RBF网络模型的SAR溢油图像识别方法 %A 周慧 %A 陈澎 %J 大连海事大学学报 %D 2018 %X 利用径向基函数(radial basis function, RBF)神经网络模型区分油膜和类油膜,旨在为溢油事故决策支持提供重要前提.首先,对合成孔径雷达(SAR)图像进行特征提取,获得有效的特征向量,并将特征向量作为输入层参数,建立激励函数;其次,利用SAR图像样本训练RBF神经网络模型,将输出值与实际值之间的误差作为约束条件调整权重因子、径向基中心和宽度,根据输出层的线性激活函数值判断溢油情况.实验结果表明,RBF模型在识别油膜与类油膜图像方面准确率超过90%. 通过比较RBF和BP神经网络在SAR溢油图像分类上的准确率,也证明了RBF的有效性.</br>Radial basis function (RBF) neural network model was used to distinguish oil slicks or look-alikes oil slicks to provide an important prerequisite for oil spill decision. Firstly, the efficient eigenvectors was extracted from synthetic aperture radar (SAR) images to acquire eigenvectors, and eigenvectors were used as input layer parameters to establish excitation function. Secondly,the RBF neural network model was trained by SAR image samples, and the error between output value and actual value was used as a constraint condition to adjust the weight factor, radial basis center and width, and estimate the oil spill situation according to the linear activation function value of output layer. Experimental results show that the accuracy rate of RBF model is more than 90% in recognition of “oil slicks” and “look-alikes oil slicks” image. The results also reveal that the outputs from the RBF neural network are more accurate compared to those from the BP neural network. %K 国家自然科学基金资助项目(51609032) %K 辽宁省教育厅科技项目(L2015057). %U http://xb.dlmu.edu.cn/CN/abstract/abstract514.shtml