南昌工程学院学报Issue(1)：7-11,5.
基于滤波后处理的主动学习高光谱遥感图像分类
Active learning for hyperspectral remote sensing image classification based on filtering postprocessing
摘要
Abstract
The Iimited IabeIed number and high dimensionaI features are the great chaIIenge that hyper-spectraI image cIassification must face. In this paper,combined active Iearning and fiItering for hyperspec-traI image cIassification approach is proposed. It attempts to efficientIy update existing cIassifiers and im-prove the generaIization capabiIity and cIassification accuracy by using fewer IabeIed data,to reduce the number of training sampIes and curse of dimensionaIity. The active Iearning and muItinomiaI Iogistic re-gression are firstIy used to cIassify the hyperspectraI remote sensing images,and then the fiItering is used to deaI with noise. The experiment resuIts show that our aIgorithm significantIy reduced the need of IabeIed sampIe and improve cIassification accuracy.关键词
高光谱/分类/主动学习/多项逻辑回归( MLR)/滤波Key words
hyperspectraI/cIassification/active Iearning/muItinomiaI Iogistic regression( MLR)/fiIter分类
信息技术与安全科学引用本文复制引用
孙宁,邓承志,汪胜前..基于滤波后处理的主动学习高光谱遥感图像分类[J].南昌工程学院学报,2015,(1):7-11,5.基金项目
国家自然科学基金资助项目（61162022，61362036）；江西省高等学校科技落地计划项目（ KJLD12098）；南昌工程学院研究生创新培养基金项目 （）
