论文部分内容阅读
针对油膜和本底海水的光谱特征,提出了一种基于经验模态分解的高光谱数据特征提取方法,并通过BP神经网络构建分类器对油膜和本底海水进行分类识别。该方法首先利用经验模态分解把原始信号在425~2 390 nm谱段范围内分解为若干个固有模态函数(IMF-Intrinsic ModeFunction),然后计算出每一个IMF的能量,选取若干个包含主要特征信息的IMF分量的能量特征参数作为BP神经网络的输入参数来识别海洋表面油膜信息。研究结果表明,该方法能准确、有效地识别出海洋表面微薄的油膜信息。
Aiming at the spectral characteristics of oil film and background seawater, a hyperspectral data feature extraction method based on empirical mode decomposition is proposed. By using BP neural network to construct a classifier, the oil film and background seawater are classified and identified. In this method, the original signal is decomposed into several intrinsic mode functions (IMF-Intrinsic Mode Functions) in the range of 425-2 390 nm by empirical mode decomposition. Then the energy of each IMF is calculated. The energy characteristic parameters of the IMF components of the feature information are used as the input parameters of the BP neural network to identify the oil surface information of the ocean surface. The results show that this method can accurately and effectively identify the meager oil film information on the ocean surface.