OLED 驱动器的发展和应用

介绍了OLED/PLED的特点和目前其驱动器的发展状态。应用Solomon公司集行驱动、列驱动和控制器为一体的SSD1301芯片和低电压、低功耗的单片机W78LE58芯片,成功地驱动了96×64点阵的PLED显示屏,给出了硬件接口电路图和软件流程图。     

基于并行附加特征提取网络的SSD地面小目标检测模型

针对SSD原始附加特征提取网络（Original Additional Feature Extraction Network,OAFEN）中stride操作造成图像小目标信息丢失和串联结构产生的多尺度特征之间冗余度较大的问题,提出了一种计算量小、感受野大的深度可分离空洞卷积（Depthwise Separable Dilated Convolution,DSDC）,并利用DSDC设计了一个包含三个独立子网络的并行附加特征提取网络（Parallel Additional Feature Extraction Network,PAFEN）.PAFEN上路用两个DSDC提取尺寸为19*19和3*3的特征图;中路用一个DSDC提取尺寸为10*10的特征图;下路用两个DSDC提取尺寸为5*5和1*1的特征图.实验结果表明,在SSD框架内,PAFEN在mAP和检测时间等方面均优于OAFEN,适用于地面小目标的检测任务.     

含区域信息的核函数物体跟踪算法

＂无穷多解＂问题,是基于核函数的视频物体跟踪算法的主要问题。而＂多极值点＂问题相比＂无穷多解＂而言更为普遍,却尚未被关注。本文对上述问题进行了研究,并且提出了一种含区域信息的核函数物体跟踪算法——区域化核函数跟踪算法（SBT）。该方法主要利用物体分割获得的区域权重信息来消除虚假的极值点,从而使跟踪过程有更高的鲁棒性。此外,该新方法自身天然具有处理物体缩放及旋转的能力。实验证明,SBT算法在跟踪过程中表现出了更好的性能。     

基于改进SSD的合成孔径声呐图像水下多尺度目标轻量化检测模型

针对轻量化目标检测模型SSD-MV2对合成孔径声呐(SAS)图像水下多尺度目标检测精度低的问题,该文提出一种新的卷积核模块-可扩张可选择模块(ESK),ESK具有通道可扩张、通道可选择和模型参数少的优点。与此同时,利用ESK模块重新设计了SSD的基础网络和附加特征提取网络,记作SSD-MV2ESK,并为其选择了合理的扩张系数和多尺度系数。在合成孔径声呐图像水下多尺度目标检测数据集SST-DET上,SSD-MV2ESK在模型参数基本相等的条件下,检测精度比SSD-MV2提升4.71%。实验结果表明,SSD-MV2ESK适用于合成孔径声呐图像水下多尺度目标检测任务。     

结合局部一致性和低频振幅探究躯体症状障碍患者大脑自发性活动的改变

躯体症状障碍（somatic symptom disorder,SSD）是一种常见的医学疾病,致病原因涉及生物学、心理学及社会因素.目前关于SSD的神经机制知之甚少.本研究通过静息态功能磁共振成像（functional magnetic resonance imaging,fMRI）,结合低频振幅（amplitudes of low-frequency fluctuation,ALFF）和局部一致性（regional homogeneity,ReHo）分析探究45位SSD患者和43位健康对照自发性脑活动特征的区别.结果发现：与对照组相比,SSD患者右侧扣带回中部的ReHo值显著升高,而右侧楔前叶、左侧颞下回延伸到左侧颞中回和左侧海马旁回、右侧脑桥的ReHo值显著降低.同时,SSD患者扣带回中部延伸至左侧额中回、右侧脑岛延伸至右侧额下回、左侧额中回延伸至左侧前扣带回的ALFF值均显著升高.这些脑区的脑功能与自我加工、情绪处理、身体知觉等有关,与SSD发病机制有重要联系.     

基于注意力机制的单发多框检测器算法

单发多框检测器SSD是一种在简单、快速和准确性之间有着较好平衡的目标检测器算法.SSD网络结构中检测层单一的利用方式使得特征信息利用不充分,将导致小目标检测不够鲁棒.该文提出一种基于注意力机制的单发多框检测器算法ASSD.ASSD算法首先利用提出的双向特征融合模块进行特征信息融合以获取包含丰富细节和语义信息的特征层,然...     

声表器件用表贴式封装管壳的研制

本文主要阐述了某种声表器件(SSD)用表贴式封装(SMP)管壳的结构设计过程,以 及为了适应大批量生产在工艺上所作的改进及创新。     

Dosimetric comparison of water phantoms,ion chambers,and data acquisition modes for LINAC characterization

PurposeIn this study a dosimetric comparison utilizing continuous data acquisition and discrete data acquisition is examined using IBA Blue Phantom (IBA Dosimetry, Schwarzenbruck, Germany) and PTW (PTW, Freiberg, Germany) MP3-M water tanks. The tanks were compared according to several factors including set up time, ease of use, and data acquisition times. A tertiary objective is to study the response of several ionization chambers in the two tanks examined.MethodsMeasurements made using a Varian 23EX LINAC (Varian Medical Systems, Palo Alto, CA) include PDDs and beam profiles for various field sizes with IBA CC13, PTW Semiflex 31010, PTW Pinpoint N31016, and PTW 31013 ion chambers for photons (6, 18 MV) and electrons (6, 9, 12, 15, and 18 MeV). Radial and transverse profile scans were done at depths of maximum dose, 5 cm, 10 cm, and 20 cm using the same set of tanks and detectors for the photon beams. Radial and transverse profile scans were done at depth of maximum dose for the electron beams on the same tanks and chambers. Data processing and analysis was performed using PTW's MEPHYSTO Navigator software and IBA's OmniPro Accept version 6.6 for the respective water tank systems.ResultsPDD values agree to within 1% and dmax to within 1 mm for the PTW MP3-M tank using PTW 31010 and Blue Phantom using IBA CC13 chamber, respectively and larger discrepancy with the PTW PinPoint N31016 chamber at 6 MV. With respect to setup time the PTW MP3-M and IBA Blue phantom tank took about 20 and 40 min, respectively. Scan times were longer by 5–15 min per field size in the PTW MP3-M tank for the square field sizes from 1 cm to 40 cm as compared to the IBA Blue phantom. However, data processing times were higher by 7 min per field size with the IBA system.ConclusionsTank measurements showed little deviation with the higher energy photons as compared to the lower energy photons with regards to the PDD measurements. Chamber construction as well as tank set up may be causing the slight deviation in data. It is important to identify the exact source of the potential errors to ensure that proper tank usage is performed when making such measurements to ensure that patient safety is in compliance. Beam profiles done with different chambers and tanks showed little to no deviation from one to another. With regards to continuous versus discrete data measurements the main difference was in the data processing technique used. Discrete data obtained required less data processing as compared to the continuous data acquired.     

SSD中一种地址映射算法研究

SSD中地址映射方式有很多种,常用的有三种：页映射、块映射和混合映射。页映射性能好,但是映射表占用空间大;块映射的映射表占用空间小,但是性能差。本文研究了一种通过算法给定的阀值来判定使用页映射还是混合映射,这样动态的转化使系统映射速度快,空间利用率高,同时也降低了系统资源的浪费和映射表的更新频繁等问题。