基于Mask R-CNN的试管-支架系统Data Matrix码识别方法

在试管-支架自动化系统的输入图像中,Data Matrix(DM)码呈现为多个小目标,图像存在成像模糊、边缘干扰严重等问题,使得传统方法难以达到良好的识别效果。为此,提出一种基于深度学习的Data Matrix码识别方法DeepDMCode,以Mask R-CNN模型为基础,通过内容差异化数据合成和同步自动化标注,实现训练数据的增强,提升模型的学习能力。在模型分割结果的基础上,提出一种旋转校正方法,确保可用标准解码库实现DM码的解码。以分辨率为1600×1200、支架容量为96的数据实验表明,由于该方法在前期码定位阶段最大程度地还原码边界信息,准确度可达0.92(mIoU),完成单张图像中所有DM识别的平均速度为5.2 s,优于YOLO、SegNet、CenterNet等主流工业基准算法。

Thermal stability and data retention of resistive random access memory with HfOx/ZnO double layers

As an industry accepted storage scheme, hafnium oxide（HfO_x） based resistive random access memory（RRAM）should further improve its thermal stability and data retention for practical applications. We therefore fabricated RRAMs with HfO_x/ZnO double-layer as the storage medium to study their thermal stability as well as data retention. The HfO_x/ZnO double-layer is capable of reversible bipolar switching under ultralow switching current（〈 3 μA） with a Schottky emission dominant conduction for the high resistance state and a Poole–Frenkel emission governed conduction for the low resistance state. Compared with a drastically increased switching current at 120℃ for the single HfO_x layer RRAM, the HfO_x/ZnO double-layer exhibits excellent thermal stability and maintains neglectful fluctuations in switching current at high temperatures（up to 180℃）, which might be attributed to the increased Schottky barrier height to suppress current at high temperatures. Additionally, the HfO_x/ZnO double-layer exhibits 10-year data retention @85℃ that is helpful for the practical applications in RRAMs.

STT-MRAM存储器数据保持试验方法研究

自旋转移力矩磁随机存储器(STT-MRAM)的基本存储单元结构为磁性隧道结(MTJ),该单元的热稳定性会随着温度的升高而变弱。MTJ这一特性导致传统激活能计算模型无法直接应用于STT-MRAM的高温数据保持特性测试,因此研究STT-MRAM的数据保持特性需探究可替代其激活能的参数。为此,搭建了基于Xilinx Kintex-7系列FPGA的测试系统,进行了多个温度数据保持试验,最终拟合出一个代替激活能来衡量STT-MRAM数据保持能力的参数,即热稳定因子。该多温度数据拟合热稳定因子的方法可有效评估STT-MRAM器件数据保持能力。

Real-time crash prediction on freeways using data mining and emerging techniques

Recent advances in intelligent transportation system allow traffic safety studies to extend from historic data-based analyses to real-time applications. The study presents a new method to predict crash likelihood with traffic data collected by discrete loop detectors as well as the web-crawl weather data. Matched case-control method and support vector machines （SVMs） technique were employed to identify the risk status. The adaptive synthetic over-sampling technique was applied to solve the imbalanced dataset issues. Random forest technique was applied to select the contributing factors and avoid the over-fitting issues. The results indicate that the SVMs classifier could successfully classify 76.32% of the crashes on the test dataset and 87.52% of the crashes on the overall dataset, which were relatively satisfactory compared with the results of the previous studies. Compared with the SVMs classifier without the data, the SVMs classifier with the web-crawl weather data increased the crash prediction accuracy by 1.32% and decreased the false alarm rate by 1.72%, showing the potential value of the massive web weather data. Mean impact value method was employed to evaluate the variable effects, and the results are identical with the results of most of previous studies. The emerging technique based on the discrete traffic data and web weather data proves to be more applicable on real- time safety management on freeways.

Quantifying urbanization levels on the Tibetan Plateau with high-resolution nighttime light data

The Tibetan Plateau(TP)is undergoing rapid urbanization.To improve urban sustainability and construct eco-logical security barriers,it is essential to quantify the spatial patterns of urbanization level on the TP,but the existing studies on the topic have been limited by the lack of socioeconomic data.This study aims to quantify the urbanization level on the TP in 2018 with Luojia1-01(LJ1-01)high-resolution nighttime light(NTL)data.Specifically,the compounded night light index is used to quantify spatial patterns of urbanization level at mul-tiple scales.The results showed that the TP had a low overall urbanization level with a large internal difference.The urbanization level in the northeast,southeast and south of the TP was relatively high,forming three hotspots centered in Xining City,Lhasa City and Shangri-La City,while the urbanization level in the central and western regions was relatively low.The analysis of influencing factors,based on the random forest model,showed that transportation and topography were the main factors affecting the TP’s spatial patterns of urbanization level.The comparison analysis with socioeconomic statistics and traditional NTL data showed that LJ1-01 NTL data can be used to more effectively quantify the urbanization level since it is more advantageous for reflecting the spatial extent of urban land and describing the spatial structure of socioeconomic activities within urban areas.These advantages are attributed to the high spatial resolution of the data,appropriate imaging time and unaf-fected by saturation phenomena.Thus,the proposed LJ1-01 NTL-based urbanization level measurement method has the potential for wide applications around the world,especially in less-developed regions lacking statistical data.Using this method,we refined the measurement of the TP’s urbanization level in 2018 for multiple scales including the region,basin,prefecture and county levels,which provides basic information for the further urban sustainability research on the TP.

An Approach of Tropospheric Correction for VLBI Phase-Referencing using GPS Data

The dominant source of error in VLBI phase-referencing is the troposphere at observing frequencies above 5 GHz. We compare the tropospheric zenith delays derived from VLBI and GPS data at VLBA stations collocated with GPS antennas. The systematic biases and standard deviations both are at the level of sub-centimeter. Based on this agreement, we suggest a new method of tropospheric correction in phase-referencing using combined VLBI and GPS data.

On the Influence Factors of Second-hand House Prices in First-tier Cities——Empirical Analysis Based on Panel Data

With the arrival of the ＂housing stock＂ in first - tier cities, the second - handhousing^market will become the dominant property market. This ardcle aim to the first - tiercities of second - hand housing prices and new home price index for the empirical analysis, thedata related to the cointegration analysis found that the result of the first -tier cities real estatemarket in China, the new home price index is the significant factors influencing the second -hand house price indexi For Beijing, Shanghai second - hand housing and new home price in-dex time series johans test, found that there exists cointegration relationship between two varia-bles,the new city real estate market prices out of a line on the secondary market have clearguide. Therefore, the real estate market regulation aiming at the first -tier cities and the＂housing stock＂ should take the second - hand housing market as the main direction, startingwith the sale price and influencing factors of new houses. At the same time, in different cities,we should adhere to the city＇ s policies, reflect the policy differentiation, promote the reformof the real estate supply side, and promote the return of housing properties.

On the non-Gaussian errors in high-z supernovae type Ia data

The nature of random errors in any data set is Gaussian, which is a well established fact according to the Central Limit Theorem. Supernovae type Ia data have played a crucial role in major discoveries in cosmology. Unlike in laboratory experiments, astronomical measurements cannot be performed in controlled situations. Thus, errors in astronomical data can be more severe in terms of systematics and non-Gaussianity compared to those of laboratory experiments. In this paper, we use the Kolmogorov-Smiruov statistic to test non-Gaussianity in high-z supernovae data. We apply this statistic to four data sets, i.e., Gold data （2004）, Gold data （2007）, the Union2 catalog and the Union2.1 data set for our analysis. Our results show that in all four data sets the errors are consistent with a Gaussian distribution.

Correcting the Contamination of Second-order Spectra:Improving HαMeasurements in Reverberation Mapping Campaigns

Long-term spectroscopic monitoring campaigns on active galactic nuclei(AGNs)provide a wealth of information about its interior structure and kinematics.However,a number of the observations suffer from the contamination of second-order spectra(SOS)which will introduce some undesirable uncertainties at the red side of the spectra.In this paper,we test the effect of SOS and propose a method to correct it in the time domain spectroscopic data using the simultaneously observed comparison stars.Based on the reverberation mapping(RM)data of NGC 5548 in2019,one of the most intensively monitored AGNs by the Lijiang 2.4 m telescope,we find that the scientific object,comparison star,and spectrophotometric standard star can jointly introduce up to~30%SOS for Grism 14.This irregular but smooth SOS significantly affects the flux density and profile of the emission line,while having little effect on the light curve.After applying our method to each spectrum,we find that the SOS can be corrected effectively.The deviation between corrected and intrinsic spectra is~2%,and the impact of SOS on time lag is very minor.This method makes it possible to obtain the HαRM measurements from archival data provided that the spectral shape of the AGN under investigation does not have a large change.

X-Ray Properties of PSR J1811-1925 by Nu STAR

We analyzed the spectral properties and pulse profile of PSR J1811-1925,a pulsar located in the center of composite supernova remnant(SNR)G11.2-0.3,by using high timing resolution archival data from the Nuclear Spectroscopic Telescope Array Mission(NuSTAR).Analysis of archival Chandra data over different regions rules out the SNR shell as the site of the hard X-ray emission while spectral analysis indicates that the NuSTAR photons originate in the pulsar and its nebula.The pulse profile exhibits a broad single peak up to 35 keV.The jointed spectrum by combining NuSTAR and Chandra can be well fitted by a power-law model with a photon index ofΓ=1.58±0.04.The integrated flux of jointed spectrum over 1-10 keV is 3.36×10^(-12)erg cm^(-2)s^(-1).The spectrum of pulsar having photon indexΓ=1.33±0.06 and a 1-10 keV flux of 0.91×10^(-12)erg cm^(-2)s^(-1).We also performed the phase-resolved spectral analysis by splitting the whole pulse-on phase into five phase bins.The photon indices of the bins are all around 1.4,indicating that the photon index does not evolve with the phase.

2014-2018年广州市全血献血者数据分析

目的:分析广州市2014-2018年全血献血者的年龄分布特点和献血量变化情况,探讨全血献血者宣传招募与保留的策略,为进一步加强无偿献血招募工作提供科学依据。方法:从广州血液中心采供血管理系统中获得2014-2018年全血献血者的全数据,分年度统计全血献血者人数、年龄分布、献血量和献血次数。结果:2014-2018年,18~25岁献血人数最高,占当年总献血人数45%~49%,其次分别为26~35岁、36~45岁、46~55岁;2014-2018年,18~25岁、46~55岁及56~60岁献血人数逐年递增,26~35岁及36~45岁献血人数逐年下降。2014-2018年,总献血人数先升后降,年度献血1次的献血者相对占比逐年下降,献血2次的献血者占比逐年增长,3次献血者占比变化不大;献血1次人数占比(87.49%)最高,其次分别为献血2次献血者(12.50%)及3次献血者(0.01%)。2014-2018年,2 U(400 mL)献血者占比最大,其次是1 U(200 mL)和1.5 U(300 mL);2 U献血者比例在2014-2017年由65.88%逐年增长到72.45%,2018年下降至69.89%;1 U献血者比例在2014-2017年由24.12%逐年减少到17.56%,2018年上升至19.83%;2014-2018年,1.5 U献血者占比在9.99%~11.99%波动,变化不大。结论:广州地区的全血献血者队伍总体呈平稳增长趋势,根据献血者的年龄分布应针对年轻人特别制订无偿献血的宣传和招募策略。

湿阻中焦证药膳食疗的组方研究

目的 基于SAS EnterpriseMiner(EM)模块,分析并总结湿阻中焦证药膳方的组方规律,为相关食疗产品的开发奠定基础。方法 将统计学与中医文献学相结合,建立改善湿阻中焦证的药膳方基础数据库,利用频数分析、关联规则分析及聚类分析对药膳方的组方规律进行数据挖掘。结果 录入药膳方144首,共涉及原料162种,中药53种。居前3的中药为薏苡仁、生姜和赤小豆,食材为粳米、鲫鱼和鲤鱼。运用关联规则分析得到药物二阶关联规则23条,基于层次聚类分析挖掘出新组方6个。结论 湿阻中焦证药膳方原料多具有利水渗湿、健脾的功效;挖掘出的新组方原料以温脾燥湿,行气利水为主。该研究对食疗产品的开发具有一定的研究价值,为药食两用的食疗产品开发奠定基础。

基于QAR数据的波音747-400飞机巡航报文解析研究（英文）

为充分挖掘QAR数据中所包含的大量原始监控数据,检验ACARS报文的可靠性,更好地协助航空公司进行飞机和发动机的状态监控及故障诊断,对发动机稳态报文DFD的报文结构进行了解析研究,提出了一种从QAR数据中提取满足触发条件的DFD报文的研究思路和方法,以丰富发动机性能监控的信息量。通过对B747-400飞机实际ACARS报文数据和所提取的报文数据进行计算分析,验证了该方法的可靠性,具有良好的工程应用前景。

移动用户维系挽留系统中K-means算法应用

4G网络技术的成熟使得用户对运营商的业务需求越来越高,如何维系用户和通过对用户属性的研究来迎合用户对业务的需求,建立方便快捷的体验服务手段,建设维系挽留系统是我国通信运营商未来发展的重中之重。首先分析移动用户维系发展现状,提出用户维系发展属性。其次,采用数据挖掘方法建立以用户稳定度和用户价值评价为基础的数据挖掘分析模型,并通过用户数据进行验证。最后,针对存量维系如何进行多渠道精准推送提出进一步展望。

一次性可编程存储器的数据保持特性建模及分析

基于300 mm 0.18μm MS 5 V工艺平台设计并流片了1k×16一次性可编程OTP器件,并对存储单元的结构、工作原理及工艺等可能影响数据保持寿命的因素进行了分析。根据Arrhenius寿命模型对不同样品设置了高温老化实验测试,收集数据并对OTP器件的保持特性进行建模。通过225℃、250℃和275℃条件下的高温老化加速实验,拟合样品最大数据保持时间曲线。在生产过程中可能出现的最差产品条件下,对1/(kT)与数据保持时间曲线进行数学拟合,计算在不同失效条件下的浮栅电荷泄漏的激活能和最大数据保持时间。

Used car price prediction based on XGBoost and retention rate

In order to improve the accuracy of used car price prediction,a machine learning prediction model based on the retention rate is proposed in this paper.Firstly,a random forest algorithm is used to filter the variables in the data.Seven main characteristic variables that affect used car prices,such as new car price,service time,mileage and so on,are filtered out.Then,the linear regression classification method is introduced to classify the test data into high and low retention rate data.After that,the extreme gradient boosting(XGBoost)regression model is built for the two datasets respectively.The prediction results show that the comprehensive evaluation index of the proposed model is 0.548,which is significantly improved compared to 0.488 of the original XGBoost model.Finally,compared with other representative machine learning algorithms,this model shows certain advantages in terms of mean absolute percentage error(MAPE),5%accuracy rate and comprehensive evaluation index.As a result,the retention rate-based machine learning model established in this paper has significant advantages in terms of the accuracy of used car price prediction.

基于PANEL DATA模型的山东省FDI与经济发展关系的实证分析

采用Panel data计量模型和经济学方法,分析了山东省外商直接投资(FDI)和经济增长的关系,得出了它们之间的多个实证模型,对我省在引进外资方面应采取的对策与措施提出了几点建议。