基于视角置信度和注意力的暴力行为识别

暴力行为容易出现遮挡情况, 识别准确率较低. 目前, 一些算法加入多视角视频输入来解决遮挡问题, 以等量权重将所有视角数据融合, 但是不同视角的视频因拍摄距离和遮挡情况本身就对识别存在差异性. 针对该问题, 本文提出一种基于视角置信度和注意力的暴力行为识别方法, 提高暴力识别的准确率. 本文将时序差分模块TDM的输入扩展成多视角, 将通道注意力机制运用在片段维度来增强TDM中跨段特征提取能力, 通过背景抑制方法突显移动目标的纹理特征并计算出每个视角图像的置信度, 引入双线性池化方法融合多视角视频特征, 根据视角置信度分配每个视角局部特征的权重. 本文在公开数据集CASIA-Action和自制数据集上进行了验证. 实验表明, 本文提出的视角置信度方法优于改进前的双线性池化方法, 暴力行为准确率相较于现有的行为识别方法取得了更好的效果.     

基于范畴论的多TBox整合研究

在语义Web环境下,单一和分散的知识库会引发数据的冗余和不一致性,且严重影响知识库智能化查询,降低知识库中数据的重用性和可操作性.文中提出一种基于范畴论的多TBox整合方法来整合知识库,旨在解决上述者类问题.此方法具有高度抽象性和强大表达能力的范畴论是多知识库整合的理想工具.将TBox作为操作对象,范畴论中"态射"功能可实现TBox的映射,"外推"功能可实现TBox的整合,最后通过实验验证方法的可行性和有效性,实验结果基本符合要求.     

SZ绞成缆设备电控系统的构成及变频器故障分析

SZ绞设备是将单元绞制和缆芯绞制在一道工序完成的一种专业电缆生产设备。SZ绞设备采用了先进的变频调速技术和PLC控制技术,有的还采用了伺服驱动,使其控制快速、精确和可靠。本文就SZ绞设备的电控系统构成及在调试运行中遇到的变频器故障进行了分析。     

语义驱动的关联规则挖掘算法研究

为了使传统的关联规则挖掘算法在结合到具体领域时具有更强的适应性,提出了DS-Apriori算法。该算法建立在语义本体的基础上,根据项集内部的语义相关度动态的确定该项集的最小支持度,并采用了项集语义相关度的增量计算方法。实验结果表明,DS-Apriori算法在很大程度上提高了关联规则挖掘算法的效率和效果。     

基于灰色关联分析法的软件缺陷类型预测

灰色关联分析法是一种描述元素之间影响程度的分析法,适合于小项目数据集。小项目数据集制约着传统的软件缺陷类型的预测方法,使得预测的结果往往不够准确和可靠。因此在灰色关联分析法的基础上提出了特征子集选择、异常工程检测以及软件缺陷类型预测3种方法。通过实验的分析对比,验证了在灰色关联分析法的基础上,提出的软件缺陷类型预测方法的准确性和可靠性。     

嵌入式系统中设备控制函数实现的分析

本文简述了在嵌入式Linux系统开发中驱动程序设计的重要性,以及驱动程序的基本结构和实现原理,通过实例详述了设备控制函数实现的过程与机制.     

知识获取技术及其在网络智能故障诊断中的应用

基于遗传算法提出了一种新型的机器学习方法,它利用遗传算法的全局寻优手段,为解决知识获取这一瓶颈问题另辟捷径。论述了该方法的原理及其实现过程,并且给出了网络故障诊断应用实例,证明了它的有效性。     

指纹图象预处理

采用块方向图滤波的方法实现了沿指纹纹线方向的图象增强及图象的二值化,用细化模板匹配的迭代方法实现了纹线的细化,解决了指纹纹线复杂难以提取的问题,具体实现了指纹图象的预处理。经过预处理后提取出了纹线,且保留了纹线的关键信息,有利于下一步的特征提取工作。     

Characteristics of entropy generation and heat transfer in a microchannel with fan-shaped reentrant cavities and internal ribs

Three-dimensional laminar fluid flow and heat transfer in a micro heat sink with cavities and internal ribs are investigated us-ing numerical methods. Moreover, according to the second law of thermodynamics, the model of entropy generation is also established for variable cross section of the microchannel. The simulation encompasses Reynolds number of 198-600, relative cavity height e1/Dhvalues of 0-0.65, relative rib height e2/Dhrange of 0-0.2167. The results show that the effect of relative rib height on entropy generation is significant, while the relative cavity height has little effect on it. The combined effect of cavities and ribs in the microchannel has better performance of heat transfer than the smooth microchannel under similar conditions. Extensive simulations are conducted to collect data on the characteristics of heat transfer and fluid flow in a micro heat sink with cavities and internal ribs. Using these data, correlations for Nusselt number and friction factor in terms of Reynolds number and the geometry of cavity and rib have been developed.     

Effect of austenite grain size and accelerated cooling start temperature on the transformation behaviors of multi-phase steel

The transformation behaviors and microstructures of a low carbon multi-phase steel were investigated by the simulation of deformation-relaxation-accelerated cooling processing,using a Gleeble 3500 thermal-mechanical simulator.A pre-treatment of solid solution at 1200°C was implemented to minimize the influence on transformation from solid solution/precipitation qualities of 0.08%Nb in this steel.On this basis,the effect of austenite grain size and accelerated cooling start temperature were studied individually.The results indicated that the transformation of ferrite in multi-phase steel could be significantly promoted by the refinement of austenite grains and the increase of relaxation time,while the hard phase,such as lath bainite or martensite,could still be obtained with the following accelerated cooling.In contrast,more uniform lower temperature transformed microstructure could form from coarse grain austenite.The potential benefit of austenite grain size on adjusting the proportion of phases in multiphase steel was also discussed.