Open AccessApplication of deep neural networks in classification of medium resolution remote sensing image
Author(s) -
Wang Dian-lai,
Su Ai-xia,
Wenping Liu
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1682/1/012014
Subject(s) - artificial neural network , support vector machine , artificial intelligence , pattern recognition (psychology) , computer science , cohen's kappa , contextual image classification , remote sensing , image (mathematics) , beijing , land cover , image resolution , machine learning , geography , land use , engineering , civil engineering , archaeology , china
Aiming at the problem that high misclassification rate of remote sensing image occurs due to the phenomenon of “same spectrum different matter” and “same object different spectrum”, the deep neural networks(DNN) is proposed to the classification of medium resolution remote sensing image. The deep neural networkstake advantage of the neural network with multiple hidden layers to learn the characteristics that can describe the essential attributes of data, and has achieved high accuracy in land cover classification in Miyun District, Beijing city. The experimental results indicate that the classification accuracy of this algorithm is the highest compared with SVM and BP neural network, the classification accuracy reaches 96.53%, and the classification accuracy is 10.39% higher than that of SVM. The kappa coefficient of DNN is 0.95, which is also the highest among the comparison algorithm. So, the DNN can apply to the classification of moderate resolution remote sensing image.