基于外观统计特征融合的人体目标再识别

人体目标再识别是视频监控等应用的关键问题之一。该文从外观统计特征融合的角度,利用人体的颜色和结构信息,基于空间直方图和区域协方差两种优秀的统计描述方法,研究了再识别问题的特征构建和测度选择等内容。构建特征时从图像多个层次的统计区域中提取了多类互补性较好的统计向量,设计测度时使用了简单的1l距离进行加权组合。两类统计方式融合而成的再识别方法不需要进行预处理和监督性训练过程。该文进行了广泛的实验比较和分析,验证了该文方法优异的识别性能和较强的实用性能。

基于部位分割的人体再识别方法

人体再识别是非重叠视域多摄像机视频分析过程中需解决的难点问题。提出一种基于部位分割的人体再识别方法,首先基于深度骨骼点实现人体部位分割,对同一人体多帧图像分割后的所有部位使用评分的策略选择最优帧,将全局颜色特征和HOG特征分配不同的权重,融合特征建立人体目标模型,使用EMD(Earth Move’s Distance)距离度量目标间的相似性。在Kinect REID和BIWI RGBD-ID数据库上实验表明,该算法具有较高的鲁棒性和识别率。

颜色特征和超像素特征融合的人体目标再识别

在非重叠视域的多摄像机监控系统中,人体目标再识别有着重要的应用。针对再识别过程中面临的光照变化、视角变化、姿态变化、遮挡等问题,提出了融合全局颜色特征和超像素特征的方法,对颜色特征和超像素特征分配不同的权重,进行人体目标间的相似性度量。超像素特征是将前景图像分割成多个超像素,采用密集采样SIFT特征结合单词包(Bag-of-Words)框架对每个超像素进行描述。将得到的超像素特征和全局颜色特征结合建立人体目标模型,分别使用EMD(Earth Mover’s Distance)距离和巴氏距离度量目标间的相似性。对多个数据库进行实验,结果证明,该算法具有较高的识别率。

基于物理光学法的再入等离子体包覆体空间散射特性分析

建立十一组元模型,数值模拟再入体化学反应绕流流场,得到流场参数随再入高度、马赫数变化的分布特性。基于绕流流场,采用等效输入阻抗方法及分层介质模型,应用物理光学法计算再入等离子体包覆体雷达散射截面。分析流场特性变化对散射特性的影响。对电磁波在等离子体鞘套内传播特性及电磁波与等离子体相互作用机理进行了初步分析。计算结果对再入体目标识别、等离子体隐身及再入通信具有一定的参考价值。

基于加权欧氏距离度量的目标再识别算法

针对传统欧氏距离在特征相似性度量中存在区分能力弱的缺陷,提出了基于加权欧氏距离度量的目标再识别算法.首先,针对现有目标再识别算法中目标分割易受衣着和背景颜色干扰的缺陷以及忽略人体头部特征的现象,提出了一种简单的比例分割方法,即根据VIPeR和i-LIDS数据集上目标各部件的比例统计将目标按比例分割成3部分.然后提取各部件的多种互补特征来增加其对光照变化等因素的鲁棒性.在部件特征描述过程中,文中提出了以显著性因子为权重的显著性局部二值模式(SLBP)特征来增加局部二值模式(LBP)特征对目标显著性的描述.最后综合各部件的相似性度量结果来判断目标是否匹配.在VIPeR和i-LIDS数据集上的对比实验结果显示,文中算法的目标再识别准确率优于其他算法.

多特征融合的人体目标再识别

在非重叠的多摄像机监控系统中,人体目标再识别是需要解决的主要问题之一。针对当前人体目标再识别使用目标的外观统计特征或者通过训练获取目标特征时存在的问题,提出一种无需训练,对视角、光照变化和姿态变化具有较强鲁棒性的基于多特征的人体目标再识别算法。首先根据空间直方图建立目标整体外观表现模型对目标进行粗识别,之后将人体目标分为3部分,忽略头部信息,分别提取躯干和腿部的主色区域的局部颜色和形状特征,并通过EMD(earth movers distance)距离进行目标精识别。实验结果表明,本文算法具有较高的识别率,且不受遮挡和背景粘连的影响。

整合全局——局部度量学习的人体目标再识别

目的人体目标再识别的任务是匹配不同摄像机在不同时间、地点拍摄的人体目标。受光照条件、背景、遮挡、视角和姿态等因素影响,不同摄相机下的同一目标表观差异较大。目前研究主要集中在特征表示和度量学习两方面。很多度量学习方法在人体目标再识别问题上了取得了较好的效果,但对于多样化的数据集,单一的全局度量很难适应差异化的特征。对此,有研究者提出了局部度量学习,但这些方法通常需要求解复杂的凸优化问题,计算繁琐。方法利用局部度量学习思想,结合近几年提出的XQDA(cross-view quadratic discriminant analysis)和MLAPG(metric learning by accelerated proximal gradient)等全局度量学习方法,提出了一种整合全局和局部度量学习框架。利用高斯混合模型对训练样本进行聚类,在每个聚类内分别进行局部度量学习;同时在全部训练样本集上进行全局度量学习。对于测试样本,根据样本在高斯混合模型各个成分下的后验概率将局部和全局度量矩阵加权结合,作为衡量相似性的依据。特别地,对于MLAPG算法,利用样本在各个高斯成分下的后验概率,改进目标损失函数中不同样本的损失权重,进一步提高该方法的性能。结果在VIPeR、PRID 450S和QMUL GRID数据集上的实验结果验证了提出的整合全局—局部度量学习方法的有效性。相比于XQDA和MLAPG等全局方法,在VIPeR数据集上的匹配准确率提高2.0%左右,在其他数据集上的性能也有不同程度的提高。另外,利用不同的特征表示对提出的方法进行实验验证,相比于全局方法,匹配准确率提高1.3%3.4%左右。结论有效地整合了全局和局部度量学习方法,既能对多种全局度量学习算法的性能做出改进,又能避免局部度量学习算法复杂的计算过程。实验结果表明,对于使用不同的特征表示,提出的整合全局—局部度量学习框架均可对全局度量学习方法做出改进。

The research on the re-employment of elderly from the perspective of holism

China face on the serious situation about the population aging. The e-employment of elderly has become one of the approach to positively response to population aging. This paper attempts to put the elderly into the whole social system under the standpoint ofholism, then, analyze the reason of re-employment of elderly for integration and Macro from different perspective, thus, look for the significance of re-employment of elderly. Based on this, the paper put forward the suggestion for the development of the re-employment of elderly.

Construction of Gpm6a/ReelinGFPCreERT2 by BAC recombination using a specific gene in hepatic mesothelial or stellate cells

AIM To prepare a Gpm6a/Reelin^(GFPCreERT2) construct with a rapid and reliable strategy using a bacterial artificial chromosome(BAC). METHODS Gpm6 a and Reelin BACs were purified and transformed into SW102 E. coli by electroporation. The GFPCreE RT2 fragment was prepared from a shuttle vector and transformed into SW102 E. coli carrying a BAC. Homologous recombination was induced in SW102 E. coli. Recombinant clones were screened and confirmed by PCR and restriction enzyme digestion. Recombinant clones were transformed into SW102 E. coli to remove the kanamycin unit.RESULTS A complete BAC was successfully transformed into SW102 E. coli by electroporation because BAC purified from SW102 E. coli showed the same pattern as the original BAC with Bam H I digestion. The GFPCre ERT2 fragment was deemed to have been prepared successfully because we obtained the same size fragment as expected. Homologous recombination was induced, and GFPCre ERT2 was deemed to have been inserted into the correct site of the BAC because we found the band change was the same as the expected pattern after restriction enzyme digestion. The kanamycin unit was deemed to have been removed successfully because we obtained different sizes of bands that were consistent with the results expected by PCR with different primers. CONCLUSION The construct of Gpm6 a^(GFPCreERT2) or Reelin^(GFPCreERT2) was prepared successfully, which will establish a foundation for tracing the hepatic stellate cell lineage and studying its function.

Degradable, absorbable or resorbable——what is the best grammatical modifier for an implant that is eventually absorbed by the body?

The adoption of grammatical modifier for im- plants or other kinds of biomaterials eventually absorbed by the body has been a long-standing confusing issue, and there are diverse terms in the large fields of research, which not only causes the difficulties when searching on the Internet, but also blurs the meaning and boundaries for researchers. Prior unification attempts at laws/standards set the basis for such research fields towards researching, labeling, marketing and instructions for use. Considering this, the typical gram- matical modifiers ＂biodegradable＂, ＂resorbable＂ ＂absorbable＂, along with their noun forms used in the decades of scien- tific research have been reviewed and explained, interdiscipli- nary in chemistry, ecology, materials science, biology, micro- biology, medicine, and based on usage customs, laws, stan- dards and markets. The term ＂biodegradable＂ has been not only used in biomaterials but also in ecology waste manage- ment, biomedicine and even natural environment. Mean- while, the term ＂resorbable＂ has long been used in biological reaction （osteoclast driven bone resorption）, but is inappro- priate for implants that do not carry the potential to grow back into their original form. The term ＂absorbable＂ focuses more on the host metabolism to the foreign biodegradation products of the implanted material/device compared with the term ＂degradable/biodegradable＂. Meanwhile the coherence and normalization of the term ＂absorbable＂ carried by its own in laws and standards contributes as well. In general, the au- thors consider the term ＂absorbable＂ to be the best grammat- ical modifier with respect to other adjectives which share the same inherence. A further internationally unified usage is proposed by us.