Deep neural network for remote-sensing image interpretation: status and perspectives  

作　　者：Jiayi Li Xin Huang Jianya Gong 

机构地区：[1]School of Remote Sensing and Information Engineering,Wuhan University,China [2]State Key Laboratory of Information Engineering in Surveying,Mapping and Remote Sensing,Wuhan University,China

出　　处：《National Science Review》2019年第6期1082-1086,共5页国家科学评论（英文版）

基　　金：supported by the National Natural Science Foundation of China(41771360 and41842035);the National Program for Support of Top-notch Young Professionals;the Hubei Provincial Natural Science Foundation of China(2017CFA029)

摘　　要：Deep neural networks(D N N s)refer to end-to-end mappings(i.e.from data to information)by stacking a large num ber of filters learned from massive samples.By courtesy of the comprehensive Earth observation platforms and convenient data access,remote-sensing practitioners are dealing with very large and ever-growing data volumes,which call for fast and transferrable machine-learning technologies for the large-scale geospatial information mining[l].While some progress has been made,research in deep-learning-based remote-sensing image interpretation is still in its infancy,mainly subject to insufficient annotation samples,high complexity of the model,and lack of in-depth integration between deep learning and remote sensing.Construction of diverse and representative remote-sensing benchmark datasets,further investigation on task-driven deep learning(i.e.the integration of deep learning and remotesensing physical mechanisms)and the efforts towards promoting the practicality of the networks should be considered in the agenda.In this context.

关 键 词：DEEP NEURAL networks 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置] TP183[自动化与计算机技术—控制科学与工程]