基于压缩感知的实时在线目标追踪

在正负样本区域随机抽取了不同尺度下图像的局部二值模式(LBP)特征,将高维的特征信息投射到低秩的压缩域,并据此建立了表观模型.使用一个随机稀疏测量矩阵来压缩前景和背景目标.将追踪问题转化成为了一个使用朴素贝叶斯分类器的二元分类问题.所提方法可以较快速、实时地在线追踪目标,同时解决了目标尺度变化、遮挡问题.     

基于梯度的帧内自适应深度决策算法

提出了一种基于梯度的帧内候选模式裁剪算法,通过自适应深度划分并利用空间信息来简化帧内预测过程,大大减少了计算复杂度.实验结果显示,所提出的算法对比HM16.7,可以减少35.25%的计算复杂度,且只有极小的编码效率损失.     

Component ACD Model and Its Application in Studying the Price-Related Feedback Effect in Investor Trading Behaviors in Chinese Stock Market

This paper explores the investors’ feedback to the price change by modelling the price-related dynamics of trading intensity. A component decomposition duration modeling approach, called the component autoregressive conditional duration (CACD) model, is proposed to capture the variation of trading intensity across time intervals between price change events. Based on the CACD model, an empirical analysis is carried out on the Chinese stock market that covers different market statuses. The empirical results suggest that the CACD model can capture the price-related dynamics of trading intensity, which supports the existence of the feedback effect and is robust across different market statuses. The authors also study how the investors react to the price change by examining the driven factors of the price-related dynamics of trading intensity. The authors find that the trading can be triggered by the fast rise in the price level and the high trading volume. Besides, investors are more sensitive to the price change direction in the sideways market than in the upward or downward markets.     

A Time-Domain Electromagnetic Transmitter with Great Power and Large Current by Using a Closed-Loop Phase Shifting Control Strategy

A time-domain electromagnetic transmitter (TDET) with great power and large current was designed by using a closed-loop phase shifting control strategy. A clamping voltage source circuit was applied in this system to improve the falling edge of emission current and shorten the turn-off time. The simulated results showed that this TDET exhibited the advantages including a large current, strong capability for rapid turn-off, stably output voltage, preferably dynamic performance, a widely adjustable range in transmitting voltage, accurate measurement precision.     

Novel multiferroicity in orthorhombic SmCrO3

This paper reports the ferroelectricity in antiferromagnetic SmCrO_3 with a reasonably large value of spontaneous electric polarization(70μC/cm~2mC cm~(-2)at 50 K)based on the measurement of pyroelectric current.The experimental results showed that the temperature of the onset of polar order was evidently higher(about220 K)than the Neel temperature(T_N=197 K)in the magnetodielectric studies.The calculated magnetodielectric effect(MDE)value was explicitly negative with maximum|MDE|values of about 3.6%around 230 K.The absence of a convincing signature in MDE at antiferromagnetic ordering temperature T_Nindicated the absence of magnetoelectric coupling around this temperature regime.The results further confirmed that the onset of polar order can be attributed to the structural distortion,rather than the canted noncollinear magnetic structure.     

基于RIA的富媒体播放组件的设计与实现

多媒体在网络中的应用越来越广,然而现有的技术很难达到视频、音频、图片与文字的完美结合.该文采用RIA设计模式,使用FLASH结合XML,JSP,数据库等技术,设计并实现了一个通用的B/S架构下的富媒体播放组件(RM-Player),用于解决上述问题.经实际应用,取得了良好的效果,证明是一种可行的办法.     

获取结合面实用切向阻尼参数的方法

从参数精度和机械系统特性综合的观点讨论了获取结合面切向阻尼参数的方法。分析了结合状态、机械系统中除所研究结合面以外的其它零件以及不同因素处理方法的影响。提出了获取结合面实用切向阻尼参数和变量表达式的方法。这些参数和表达式可以满足实际设计的需要。     

乔木护坡效果物理模型试验研究

随着人们对环境保护的日益重视,植被护坡作为一种生态友好型的边坡支护方式逐步得到了广泛应用,其中针对草本和灌木护坡效果的研究较多,但有关乔木护坡效果的研究较少。为此,通过3D打印技术制作了较为逼真的乔木根系模型,并基于自行设计的滑坡物理模型试验平台,开展了乔木护坡效果物理模型试验,对边坡变形破坏全过程中抗滑力、坡体位移和坡体滑动面积进行了分析。结果表明：与无根系护坡试验工况相比,(1)有根系护坡坡体的抗滑力峰值增大、达到峰值所用时间延长,峰后抗滑力衰减程度降低;(2)有根系护坡坡体的坡体位移,随着根系长度和布设排数的增加总体呈现下降趋势,且研究发现,持续增加根系长度并不能持续显著减小坡体位移,而当根系长度较大时,适当减少根系布设排数亦能达到较好的护坡效果;(3)有根系护坡坡体的滑动面积明显减小。上述认识丰富了乔木护坡效果方面的研究,对乔木护坡实际工程应用有一定指导意义。     

多视域下基于机器视觉的索力测试

拉索是缆索承重体系桥梁的关键承力或传力构件,在体系构成中的地位至关重要。为解决现有非接触视觉测量存在的测试视场不足、应用场景复杂多变,致使其测试精度与视域范围相互矛盾无法协调统一的技术难题,提出了多视域下基于机器视觉的索力测试新方法。以全息视觉传感器系统获取斜拉索全视域-多视域下的空间几何构型数据,建立了像素映射及图像增强算法计算模型;通过数字空间基准平面的信息转化,对复杂测试场景下视觉传感器光轴射线与被测物体所在承影面非正交情况进行测试参数修正。为避免实测拉索全息性态特征因偏离标准悬链线所导致的索力测量误差,以单一视域下的几何构型拟合降噪后的全息曲线并精确定拉索上下锚固点坐标,进而可由拉索全息性态特征参数精确测算斜拉索索力。试验结果表明：所提方法可在复杂测试场景下量化分析拉索索力计算参数,相较于常规索力测试方法,最大相对误差为9.2%,均方根误差为2.79%,满足工程实践测试精度及稳定性要求,为复杂测试场景下斜拉桥索力监测提供了一种新途径。     

基于CNN-LSTM-lightGBM组合的超短期风电功率预测方法

近年来,风电装机规模逐年增加,风电数据采集量呈现规模化增长,面对海量多维、强波动的风电数据,风电功率预测精度仍面临一定的挑战。为提高风电功率预测精度,提出了基于卷积神经网络(convolutional neural networks, CNN)-长短期记忆网络(long short-term memory, LSTM)和梯度提升学习(light gradient boosting machine, lightGBM)组合的超短期风电功率预测方法。首先,分别建立CNN-LSTM和lightGBM的风电功率超短期预测模型。其中,CNN-LSTM模型采用CNN对风电数据集进行特征处理,并将其作为LSTM模型的数据输入,从而建立CNN-LSTM融合的预测模型;然后,采用误差倒数法对CNN-LSTM和lightGBM的预测数据进行加权组合,建立CNN-LSTM-lightGBM组合的预测模型;最后,采用张北曹碾沟风电场的风电数据集,以未来4 h风电功率为预测目标,验证了组合模型的有效性。预测结果表明：相较于其他3种单一模型,组合模型具有更高的预测精度。