融合特征增强模块的小样本农业害虫识别

基于深度学习的图像识别技术在具体应用前必须先经过大量带标签样本的训练,然而在实际场景中目标域样本可能非常稀缺,小样本图像识别技术应运而生.为了提升小样本场景下的图像识别准确率,本文提出一个通用的两阶段训练模型以融合现行主流方法并增强其表现.首先,针对训练时不同害虫种类背景相似度过高的问题提出融合双注意力机制的特征加强模块;其次,针对小样本情况下预测可能产生的过拟合问题提出基于高斯分布的特征生成模块以提高泛化能力;最后,将三种典型小样本识别方法统一成两阶段训练模型以融入提出的方法.将该思路及改进首次应用于传统害虫分类数据集IP102,识别准确率可以在基准方法上取得2.11%到6.87%的提升.为了进一步验证本文方法的有效性,在小样本领域公开数据集Mini Imagenet也进行了相应的实验,提升效果同样显著.     

Weighting and selection of variables for cluster analysis

One of the thorniest aspects of cluster analysis continues to be the weighting and selection of variables. This paper reports on the performance of nine methods on eight leading case simulated and real sets of data. The results demonstrate shortcomings of weighting based on the standard deviation or range as well as other more complex schemes in the literature. Weighting schemes based upon carefully chosen estimates of within-cluster and between-cluster variability are generally more effective. These estimates do not require knowledge of the cluster structure. Additional research is essential: worry-free approaches do not yet exist.     

蒙古族长调《圣》韵律声学特征研究

通过采集蒙古族长调《圣主的两匹骏马》的语音信号,运用实验语音声学分析方法,进行了信号标注和参数提取.文章是通过基频和能量参数来分析长调节拍的基本韵律特征,分析歌唱共振峰和颤音的声学表现,探讨其发音原理和演唱技巧,以数字化方式更为深入和直观地研究蒙古族长调的风格特色.     

国家教学标本子平台的功能与使用

本文首先回顾了我国标本平台的建设的发展进程,然后从主要功能、使用方法和平台特色三个方面对国家教学标本子平台进行了介绍,最后对平台的建设进行了展望。     

藏文字库设计中OpenType特征标记的应用研究

基于国际标准实现藏字处理一直是藏文信息处理的难点,OpenType字库的实现解决了该难点,微软Vista操作系统的推出,把藏文等几种少数民族文字的处理推到了一个新的阶段,但Vista自身只带了一个藏文白体字库,不能满足用户对字体的需求。文章分析了OpenType特征标记在藏文字库设计中的应用,有助于设计OpenType藏文字库。     

mproving the Input of Classified Neural Networks Through Feature Construction

SOFTWARE,ALGORITHM AND SIMULATION1. Pas~EClassification, as an important program of data mining, is to build a classified function or model that is able tomap the item Of database to one of given classes. The classification algorithm in common use mainly includebayesian, decision trees, mule induction and neural networks [1].The classification algorithm based on decision trees has been in maturity earlier in 1980s', typical of whichis ID3 by J. R. Quinlan [21, it introduced the …     

Enhancement of tribological properties in siliconized graphite via hierarchically hybrid SiC/C composite

This study aims to investigate tribological properties in a siliconized graphite(СГ-П-0.5) and hierarchically hybrid Si C/C composites at the load ranging from 10-40 N under dry air/water lubrication conditions. A feasible liquid silicon infiltration(LSI) was used to fabricate the Si C/C composites. Morphology and phase identification of two materials were characterized and analyzed by scanning electron microscope(SEM) and energy dispersive spectrometer(EDS). An image feature extraction was used to post-process the data of wear test, further to gain the wear properties of the samples. The results demonstrated that a lower coefficient of friction(COF) and wear rate can be achieved in the LSI Si C composites than that in the sample of СГ-П-0.5. Both COF decreased under waterlubricated conditions, compared with that under the dry air lubrication. The excellent tribological properties in two materials can be due mainly to the microstructure of the СГ-П-0.5 sample, showing much more graphite and Si dispersed on the matrix. The СГ-П-0.5 sample indicated the typical oxidative wear mechanism, whereas, the sample of LSI Si C demonstrated the combined oxidative and fatigue wear. This finding can provide a promising new siliconized graphite materials for ceramic rolling bearings.     

太白国家森林公园的特征及功能

通过实地考察和分析．论述了太白国家森林公园的基本特征与功能．并时其建设与管理提出了具体的建议和措施。     

原油产能预测中色谱指纹的特征选取方法研究

通过动态监测油井分油层的产能可以掌握油井的生产状况,监测产能在油田生产中具有重要的意义.目前基于色谱指纹的产能预测越来越受到油田相关部门的重视,但是目前色谱指纹的选取还是依靠专业技术人员进行选取,这导致特征选取存在一定的主观性.到目前为止还没有研究者对色谱指纹特征的选择问题进行系统的研究.为了对色谱指纹特征选取问题进行分析,本文分别采用主成分分析(PCA)、线性相关以及随机森林中的重要性分析理论进行特征选取,最后提出了一种新的联合特征选取方法.对来自南海某油田的配比数据进行了测试,实验结果表明,本文提出的方法取得了很好的试验效果,在原油产能预测领域有较大的应用价值.     

特征峰值投影技术应用于鱼粉蛋白近红外波长优选

采用近红外光谱线性分析技术实现对鱼粉蛋白的快速检测,选择合适的波长变量是提高模型预测精度的关键。主要目的是建立一种稳健、简单的多元线性回归(MLR)模型,通过研究基于特征峰值的投影技术实现参与建模的波长优选。特征物质在近红外光谱区域的吸收特征,以鱼粉一阶导数光谱的峰谷波长点作为出发点,依次采用逐步多元线性回归(SMLR)和连续投影线性回归(SPA-MLR)方法完成两度特征信息波长筛选,进一步对备选的波长变量执行显著性检验,最终确定近红外线性分析的特征信息波长组合。结果表明,近红外长波区域中优选出53个特征信息波长变量,能够提高鱼粉蛋白近红外定量模型的预测精度,简化了模型,从而提高了模型的适用性和稳健性。