摘要
多类别识别对于遥感图像分类的实用化具有重大意义。本文提出一种由多层神经网络与无监督分类相结合的复合分类方法。第一步用多层网络对几个大类进行有监督分类，第二步将网络输出作为无监督分类的输入，对遥感图像进行细分，使得可识别的类别数从原来的１０类提高到３０类。对ＳＰＯＴ遥感图像识别的结果表明，该算法能适应多类别识别任务的要求。
Multicategory recoginition is very important to practicality of remotely-sensed image classification。This paper has presented a mixed classification methodintegrating multi-layer neuron network and unsupervised classification algorithm,Atthe first step,a multi-layer neuron network is used and the result serves as inputfor the unsupervised classification at the second step.The number of patterns thatcan be recognized is increased from 10 to 30.Applying this algorithm to SPOTremotely-sensed image recognition shows it can adapt the requiry of multicategoryrecognition.
出处
《环境遥感》
CSCD
1995年第4期298-302,T001,共6页
基金
国家自然科学基金
关键词
复合分类
多层神经网络
遥感
图象
Mixed classification method,Multi-layer neuron network,Supervisedclassification,Unsupervised classification
