树状分类器法用于气体红外光谱数据的识别分类

Tree Discriminator Algorithm Used for Identification and Classification of Gas Infrared Remote Sensing Data

介绍了一种新的分类方法%D树状分类器法 ,它抛弃了传统的线性搜索 ,用一个超平面进行分类的局限 ,引入了多个超平面进行分类 ,从而对在特征空间分布复杂交错的、非线性可分的样本有着较好的分类效果。文中详细地介绍了其原理 ,并利用树状分类器进行算法训练和鉴别分类。通过实例比较说明 ,此方法应用于大气遥感红外光谱数据的复杂样本进行分类和识别时 ,相比于一个超平面是更行之有效的。

A new classification technology tree discriminator is proposed.In which the of traditional linear search using one super surface, replaees by multi surface to classify. When the samples are complex interlaced and cannot be classified by linear method, they can be classified by the new method in the feature space effectively. The article presents its principles in detail.Using the tree discriminator, we have done algorithm training and classification. Experimental results show that this method is more effective than one super surface on identification and classification of complex data of gas infrared remote sensing.