破译无人机原始图像的效率与情报支援的速度是密切相关的,但目前针对合成孔径图像与光学图像的自动识别算法还不够成熟,存在模型较大、识别率较低等问题。从提高模型识别率、模型轻量化入手,提出一种可以有效识别合成孔径雷达(Synthetic...破译无人机原始图像的效率与情报支援的速度是密切相关的,但目前针对合成孔径图像与光学图像的自动识别算法还不够成熟,存在模型较大、识别率较低等问题。从提高模型识别率、模型轻量化入手,提出一种可以有效识别合成孔径雷达(Synthetic Aperture Radar,SAR)图像与光学遥感图像的轻量化卷积神经网络算法。首先对残差收缩网络进行改进,构建特征提取模块,用自适应K值的一维卷积取代全连接层,并在网络中加入空间注意力,提高阈值提取效率;然后用特征提取模块构建模型,并用MSTAR(Maing and Stationany Target Acquistion and Recognition)数据集与UC Merced Land-Use Data Set、SIRI-WHU两类光学遥感图像测试模型性能,实验显示模型是有效的。展开更多
Two-dimensional(2D)transition metal chalcogenides(TMCs)are promising for nanoelectronics and energy applications.Among them,the emerging non-layered TMCs are unique due to their unsaturated dangling bonds on the surfa...Two-dimensional(2D)transition metal chalcogenides(TMCs)are promising for nanoelectronics and energy applications.Among them,the emerging non-layered TMCs are unique due to their unsaturated dangling bonds on the surface and strong intralayer and interlayer bonding.However,the synthesis of non-layered 2D TMCs is challenging and this has made it difficult to study their structures and properties at thin thickness limit.Here,we develop a universal dual-metal precursors method to grow non-layered TMCs in which a mixture of a metal and its chloride serves as the metal source.Taking hexagonal Fe_(1-x)S as an example,the thickness of the Fe_(1-x)S flakes is down to 3 nm with a lateral size of over 100 μm.Importantly,we find ordered cation Fe vacancies in Fe_(1-x)S,which is distinct from layered TMCs like MoS_(2) where anion vacancies are commonly observed.Low-temperature transport measurements and theoretical calculations show that 2D Fe_(1-x)S is a stable semiconductor with a narrow bandgap of60 meV.In addition to Fe_(1-x)S,the method is universal in growing various non-layered 2D TMCs containing ordered cation vacancies,including Fe_(1-x)Se,Co_(1-x)S,Cr_(1-x)S,and V_(1-x)S.This work paves the way to grow and exploit properties of non-layered materials at 2D thickness limit.展开更多
文摘破译无人机原始图像的效率与情报支援的速度是密切相关的,但目前针对合成孔径图像与光学图像的自动识别算法还不够成熟,存在模型较大、识别率较低等问题。从提高模型识别率、模型轻量化入手,提出一种可以有效识别合成孔径雷达(Synthetic Aperture Radar,SAR)图像与光学遥感图像的轻量化卷积神经网络算法。首先对残差收缩网络进行改进,构建特征提取模块,用自适应K值的一维卷积取代全连接层,并在网络中加入空间注意力,提高阈值提取效率;然后用特征提取模块构建模型,并用MSTAR(Maing and Stationany Target Acquistion and Recognition)数据集与UC Merced Land-Use Data Set、SIRI-WHU两类光学遥感图像测试模型性能,实验显示模型是有效的。
基金supported by the National Science Fund for Distinguished Young Scholars(52125309)the National Natural Science Foundation of China(51991343,51920105002,51991340,52188101,and 11974156)+3 种基金Guangdong Innovative and Entrepreneurial Research Team Program(2017ZT07C341 and 2019ZT08C044)the Bureau of Industry and Information Technology of Shenzhen for the “2017 Graphene Manufacturing Innovation Center Project”(201901171523)Shenzhen Basic Research Project(JCYJ20200109144616617 and JCYJ20190809180605522)Shenzhen Science and Technology Program(KQTD20190929173815000 and 20200925161102001)。
文摘Two-dimensional(2D)transition metal chalcogenides(TMCs)are promising for nanoelectronics and energy applications.Among them,the emerging non-layered TMCs are unique due to their unsaturated dangling bonds on the surface and strong intralayer and interlayer bonding.However,the synthesis of non-layered 2D TMCs is challenging and this has made it difficult to study their structures and properties at thin thickness limit.Here,we develop a universal dual-metal precursors method to grow non-layered TMCs in which a mixture of a metal and its chloride serves as the metal source.Taking hexagonal Fe_(1-x)S as an example,the thickness of the Fe_(1-x)S flakes is down to 3 nm with a lateral size of over 100 μm.Importantly,we find ordered cation Fe vacancies in Fe_(1-x)S,which is distinct from layered TMCs like MoS_(2) where anion vacancies are commonly observed.Low-temperature transport measurements and theoretical calculations show that 2D Fe_(1-x)S is a stable semiconductor with a narrow bandgap of60 meV.In addition to Fe_(1-x)S,the method is universal in growing various non-layered 2D TMCs containing ordered cation vacancies,including Fe_(1-x)Se,Co_(1-x)S,Cr_(1-x)S,and V_(1-x)S.This work paves the way to grow and exploit properties of non-layered materials at 2D thickness limit.
