摘要
高光谱曲线信息量大,如何从中提取有效信息是一个难点。赤铁矿作为花岗岩型铀矿床蚀变带中的典型蚀变矿物,也是寻找铀矿的标志之一。文章从铀矿床矿物的高光谱曲线中提取光谱特征参数,利用神经网络方法构建光谱特征参数与赤铁矿含量的关联模型,对铀矿床矿物中赤铁矿的含量进行预测。结果表明,和传统的多元线性回归、支持向量机等方法相比,神经网络方法的精度高于现有算法,是一种可行的、有效的算法。
Hyperspectral curve contains a lot of messages and it is difficult to extract the effective information. Hematite is an important indication of uranium, existing in typical alteration mineral in granite-type Uranium deposit. In this paper, spectral characteristic parameters are extracted form the hyperspectral curve, while neural network algorithm is used to build the model between the parameters and content of hematite for predicting the hematite content. Experiment results show that neural network is more accurate than traditional algorithms such as multiply linear regression and is available and effective on hematite content prediction.
出处
《计算机与数字工程》
2011年第12期147-150,共4页
Computer & Digital Engineering
关键词
高光谱
神经网络
回归
hyperspectral, neural network, regression
