盲目反卷积光谱图超分辨复原算法

反卷积是实现光谱图超分辨复原的重要手段,与常规反卷积相比,盲目反卷积具有不需要预先准确获取卷积核函数的优势。着眼于充分利用光谱信号的特点和已有的光谱图反卷积成果,详细讨论了空域迭代盲目反卷积方法用于光谱图反卷积时的算法实现问题,并在分析光谱图卷积退化过程的基础上,针对光谱图反卷积算法特点,提出了光谱图卷积退化简化计算模型和最小二乘高斯拟合模型,以解决算法中相应的计算问题。基于Matlab平台的仿真表明,对于所用的高斯型谱线和点扩散函数,空域迭代盲目反卷积算法效果良好,在信噪比为50 dB时,分辨率提高约30％。     

Repair the faulty TSVs with the improved FNS-CAC codec

Through-silicon via (TSV) is a key enabling technology for the emerging 3-dimension (3D) integrated circuits (ICs). However, the crosstalk between the neighboring TSVs is one of the important sources of the soft faults. To suppress the crosstalk, the Fibonacci-numeral-system-based crosstalk avoidance code ( FNS-CAC) is an effective scheme. Meanwhile, the self-repair schemes are often used to deal with the hard faults, but the repaired results may change the mapping between signals to TSVs, thus may reduce the crosstalk suppression ability of FNS-CAC. A TSV self-repair technique with an improved FNS-CAC codec is proposed in this work. The codec is designed based on the improved Fibonacci numeral system (FNS) adders, which are adaptive to the health states of TSVs. The proposed self-repair technique is able to suppress the crosstalk and repair the faulty TSVs simultaneously. The simulation and analysis results show that the proposed scheme keeps the crosstalk suppression ability of the original FNS-CAC, and it has higher reparability than the local self-repair schemes, such as the signal-switching-based and the signal-shifting-based counterparts.     

微量元素添加对多元Sn-Bi系焊料合金组织与性能的影响

传统焊料合金由于熔点温度高,不能满足部分有机基板、温度敏感器件以及3D封装等多层封装形式的低温封装要求.以Sn-Bi合金为基体,通过添加微量Ag、Cu、Co和Ni元素形成新型多元合金,对多元合金的熔化性能、润湿性能、微观组织和力学性能进行研究.结果 表明:微量元素的添加(质量分数0～1％)对多元Sn-Bi系合金的固相线...     

基于盲数理论的空中目标粗分类

该文首先总结了直升机目标雷达探测和识别的难点,然后分析了直升机目标雷达回波信号频谱调制特征,给出了直升机回波频谱宽度的估算式。以频谱宽度为特征,引入盲数理论,对直升机和固定翼飞机进行初步分类,并给出了仿真试验结果。试验结果证明了该文的理论分析和分类算法。     

基于矢量量化的模糊参数辨识及分辨率增强方法

在实际的图像复原中通常需要预先估计模糊函数。该文将Nakagaki等人提出的基于矢量量化的模糊参数辨识算法进行改进,利用Sobel 算子形成特征矢量,避免了LOG滤波器参数的选择,增强了算法对辨识不同类型图像的模糊函数的鲁棒性,并利用DCT对特征矢量降维,减小了计算量。同时将其应用于超分辨率图像复原中,辨识出多幅低分辨率图像的模糊函数,然后融合具有不同模糊函数和信噪比的低分辨率图像,实现了盲超分辨率图像复原。仿真结果表明了改进算法的有效性和可行性。     

基于迁移学习的原子力显微镜成像恢复方法

受限于探针针尖结构尺寸,用原子力显微镜进行微纳测量时会产生图像边缘失真.提出了一种基于迁移学习的原子力显微镜成像恢复方法,通过迁移学习训练源模型和靶模型实现一维栅格成像恢复.该方法采用数学形态法中的腐蚀算法生成栅格点云数据,通过U-Net网络源模型从点云中提取针尖卷积效应的特征向量,将权重参数迁移至U-Net网络靶模型,靶模型在自适应正则化方法下进行监督学习.实验结果表明,该方法能有效恢复一维栅格的原子力显微镜测量图像,提高横向分辨力,可用于纳米栅格的线宽检测上.     

LTPS TFT层间绝缘层过孔刻蚀的工艺优化

为了适应LTPS TFT LCD显示屏超高分辨率极细布线的趋势,降低LTPS TFT层间绝缘层过孔刻蚀带来的良率损失,提高产品品质,本文研究了LTPS TFT层间绝缘层过孔刻蚀的工艺优化。实验以干法刻蚀为主刻蚀,湿法刻蚀为辅刻蚀的方式,既结合干法刻蚀对侧壁剖面角及刻蚀线宽的精确控制能力,又利用了湿法刻蚀高刻蚀选择比的优良特性,改善了层间绝缘层刻蚀形貌,减少干法刻蚀对器件有源层的损伤,避免有源层被氧化,防止刻蚀副产物污染开孔表面。实验结果表明,干法辅助湿法刻蚀能基本解决刻蚀过程中过刻、残留的问题,使得层间绝缘层过孔不良良率损失减少73%以上,且TFT源漏电极接触电阻减小约90%,器件开态电流提升约15%。干法辅助湿法刻蚀是一种优化刻蚀工艺,提升产品性能的新方法。     

智能天线技术综述

综合、详细地论述了智能天线工作原理、信号模型、波束形成算法和应用现状,并提出了目前在智能天线技术研究方面亟待解决的主要问题。     

Scaling the Aspect Ratio of Nanoscale Closely Packed Silicon Vias by MacEtch: Kinetics of Carrier Generation and Mass Transport

Metal‐assisted chemical etching (MacEtch) has shown tremendous success as an anisotropic wet etching method to produce ultrahigh aspect ratio semiconductor nanowire arrays, where a metal mesh pattern serves as the catalyst. However, producing vertical via arrays using MacEtch, which requires a pattern of discrete metal disks as the catalyst, has often been challenging because of the detouring of individual catalyst disks off the vertical path while descending, especially at submicron scales. Here, the realization of ordered, vertical, and high aspect ratio silicon via arrays by MacEtch is reported, with diameters scaled from 900 all the way down to sub‐100 nm. Systematic variation of the diameter and pitch of the metal catalyst pattern and the etching solution composition allows the extraction of a physical model that, for the first time, clearly reveals the roles of the two fundamental kinetic mechanisms in MacEtch, carrier generation and mass transport. Ordered submicron diameter silicon via arrays with record aspect ratio are produced, which can directly impact the through‐silicon‐via technology, high density storage, photonic crystal membrane, and other related applications.     

基于盲源分离的多幅顺序图像加密方法

根据许多加密方法利用数学难题保障其安全性的思想,盲源分离欠定难题可用于高度安全的多幅图像加密。然而,由于盲源分离存在顺序和幅度模糊性,通过盲源分离得到的解密图像可能发生顺序变化和像素值反转。这在加密多幅顺序图像时可能导致解密错误。针对这一问题,该文利用数字水印技术,加密前在每幅明文图像中嵌入与其对应的顺序信息;解密后,通过在各解密图像中检测该顺序信息而消除其顺序和幅度的模糊性。计算机仿真结果表明,该方法在恢复解密图像顺序的同时也能检测其是否反转,从而有效地解决了多幅顺序图像的盲源分离加密问题。