西安文理学院学报(自然科学版)
西安文理學院學報(自然科學版)
서안문이학원학보(자연과학판)
Journal of Xi'an University of Arts and Science (Natural Science Edition)
2015年
4期
66-69
,共4页
遥感%深度学习%自学习
遙感%深度學習%自學習
요감%심도학습%자학습
remote sensing%deep learning%self-learning
基于统计的传统无监督机器学习识别分类技术虽经持续改进对于高分遥感图像效果仍不佳,深度学习具备仿人类神经网络多层抽象能力和无监督自学习特点,具有从大量无标签高光谱遥感数据中自主学习和构建其特征的能力,再结合常用分类算法进行识别分类,比传统方法具有相对更高的准确率。
基于統計的傳統無鑑督機器學習識彆分類技術雖經持續改進對于高分遙感圖像效果仍不佳,深度學習具備倣人類神經網絡多層抽象能力和無鑑督自學習特點,具有從大量無標籤高光譜遙感數據中自主學習和構建其特徵的能力,再結閤常用分類算法進行識彆分類,比傳統方法具有相對更高的準確率。
기우통계적전통무감독궤기학습식별분류기술수경지속개진대우고분요감도상효과잉불가,심도학습구비방인류신경망락다층추상능력화무감독자학습특점,구유종대량무표첨고광보요감수거중자주학습화구건기특정적능력,재결합상용분류산법진행식별분류,비전통방법구유상대경고적준학솔。
Although the traditional unsupervised machine learning recognition classification technology based on statistics has been improved continuously, the effect is still poor for high resolution remote sensing images, deep learning has the ability to imitate the multi-level ab-straction and unsupervised self-learning features of the human neural network, and has the abil-ity of autonomous learning and constructing its characteristics from a large number of non-label hyper-spectral remote sensing data. Combined with common classification algorithm for classifi-cation, it has a relatively higher accuracy than that of the traditional method.