基于多光谱的空间点目标特征提取与识别

A Method Based on the Multispectral to Extract Features and Recognize Space Point Target

基于实际测量的空间目标多光谱数据 ,提出了新的多光谱特征分析方法 ,即把目标红外辐射分为长、中、短三个波段 ,分析提取了三个波段内的辐照度 ,以及长波段、中波段辐照度与短波段辐照度的比值 ,估算出目标温度和有效辐射面积 ,并利用 BP神经网络对四类目标进行识别试验 ,取得了好的识别效果。

Based on the real multispectral data of space targets, a new way of feature-extraction was brought forward. That is to separate the infrared radiation into three bands: long, middle, short. Then analyze and extract the following features: the irradiance of three bands, irradiance's ratio of long, middle to short band. Thus targets' temperature and radiant area are estimated. Good recognition result is achieved in a recognition test of four types of targets with BP neural network.