传统的视觉词典模型(Bag of Visual Words Model,Bo VWM)中广泛存在视觉单词同义性和歧义性问题.且视觉词典中的一些噪声单词—"视觉停用词",也会降低视觉词典的语义分辨能力.针对这些问题,本文提出了基于近义词自适应软分配...传统的视觉词典模型(Bag of Visual Words Model,Bo VWM)中广泛存在视觉单词同义性和歧义性问题.且视觉词典中的一些噪声单词—"视觉停用词",也会降低视觉词典的语义分辨能力.针对这些问题,本文提出了基于近义词自适应软分配和卡方模型的图像目标分类方法.首先,该方法利用概率潜在语义分析模型(Probabilistic Latent Semantic Analysis,PLSA)分析图像中视觉单词的语义共生概率,挖掘图像隐藏的语义主题,进而得到语义主题在某一视觉单词上的概率分布;其次,引入K-L散度度量视觉单词间的语义相关性,获取语义相关的近义词;然后,结合自适应软分配策略实现SIFT特征点与若干语义相关的近义词之间的软映射;最后,利用卡方模型滤除"视觉停用词",重构视觉词汇分布直方图,并采用SVM分类器完成目标分类.实验结果表明,新方法能够有效克服视觉单词同义性和歧义性问题带来的不利影响,增强视觉词典的语义分辨能力,较好地改善了目标分类性能.展开更多
为了解决传统视觉词典模型(bag of visual words model,Bo VWM)中存在的时间效率低、词典区分性不强的问题,以及由于空间信息的缺失及量化误差等导致的目标检索性能较低的问题。提出一种新的目标检索方法,首先引入精确欧氏位置敏感哈希(...为了解决传统视觉词典模型(bag of visual words model,Bo VWM)中存在的时间效率低、词典区分性不强的问题,以及由于空间信息的缺失及量化误差等导致的目标检索性能较低的问题。提出一种新的目标检索方法,首先引入精确欧氏位置敏感哈希(exact euclidean locality sensitive hashing,E2LSH)过滤训练图像集中的噪声和相似关键点,提高词典生成效率和质量;然后,引入卡方模型(Chi-square model)移除词典中的视觉停用词增强视觉词典的区分性;最后,采用空间一致性度量准则进行目标检索并对初始结果进行K-近邻(K-nearest neighbors,K-NN)重排序。将提出的方法在数据库Oxford5K和Flickr1上进行目标检索,结果表明,新方法在一定程度上改善了视觉词典的质量,增强了视觉语义分辨能力,有效地提高目标检索性能。展开更多
基于视觉词典模型(bag of visual words model,BoVWM)的目标检索存在时间效率低、词典区分性不强的问题,以及由于空间信息的缺失及量化误差等导致的视觉语义分辨力不强的问题.针对这些问题,提出了基于词典优化与空间一致性度量的目标检...基于视觉词典模型(bag of visual words model,BoVWM)的目标检索存在时间效率低、词典区分性不强的问题,以及由于空间信息的缺失及量化误差等导致的视觉语义分辨力不强的问题.针对这些问题,提出了基于词典优化与空间一致性度量的目标检索方法.首先,该方法引入E2LSH(exact Euclidean locality sensitive hashing)过滤图像中的噪声和相似关键点,提高词典生成效率和质量;然后,引入卡方模型(chi-square model,CSM)移除词典中的视觉停用词增强视觉词典的区分性;最后,采用空间一致性度量准则进行目标检索并对初始结果进行K-近邻(K-nearest neighbors,K-NN)重排序.实验结果表明:新方法在一定程度上改善了视觉词典的质量,增强了视觉语义分辨能力,进而有效地提高目标检索性能.展开更多
Natural ventilation is driven by either buoyancy forces or wind pressure forces or their combinations that inherit stochastic variation into ventilation rates. Since the ventilation rate is a nonlinear function of mul...Natural ventilation is driven by either buoyancy forces or wind pressure forces or their combinations that inherit stochastic variation into ventilation rates. Since the ventilation rate is a nonlinear function of multiple variable factors including wind speed, wind direction, internal heat source and building structural thermal mass, the conventional methods for quantifying ventilation rate simply using dominant wind direction and average wind speed may not accurately describe the characteristic performance of natural ventilation. From a new point of view, the natural ventilation performance of a single room building under fluctuating wind speed condition using the Monte-Carlo simulation approach was investigated by incorporating building facade thermal mass effect. Given a same hourly turbulence intensity distribution, the wind speeds with 1 rain frequency fluctuations were generated using a stochastic model, the modified GARCH model. Comparisons of natural ventilation profiles, effective ventilation rates, and air conditioning electricity use for a three-month period show statistically significant differences （for 80% confidence interval） between the new calculations and the traditional methods based on hourly average wind speed.展开更多
文摘传统的视觉词典模型(Bag of Visual Words Model,Bo VWM)中广泛存在视觉单词同义性和歧义性问题.且视觉词典中的一些噪声单词—"视觉停用词",也会降低视觉词典的语义分辨能力.针对这些问题,本文提出了基于近义词自适应软分配和卡方模型的图像目标分类方法.首先,该方法利用概率潜在语义分析模型(Probabilistic Latent Semantic Analysis,PLSA)分析图像中视觉单词的语义共生概率,挖掘图像隐藏的语义主题,进而得到语义主题在某一视觉单词上的概率分布;其次,引入K-L散度度量视觉单词间的语义相关性,获取语义相关的近义词;然后,结合自适应软分配策略实现SIFT特征点与若干语义相关的近义词之间的软映射;最后,利用卡方模型滤除"视觉停用词",重构视觉词汇分布直方图,并采用SVM分类器完成目标分类.实验结果表明,新方法能够有效克服视觉单词同义性和歧义性问题带来的不利影响,增强视觉词典的语义分辨能力,较好地改善了目标分类性能.
文摘Natural ventilation is driven by either buoyancy forces or wind pressure forces or their combinations that inherit stochastic variation into ventilation rates. Since the ventilation rate is a nonlinear function of multiple variable factors including wind speed, wind direction, internal heat source and building structural thermal mass, the conventional methods for quantifying ventilation rate simply using dominant wind direction and average wind speed may not accurately describe the characteristic performance of natural ventilation. From a new point of view, the natural ventilation performance of a single room building under fluctuating wind speed condition using the Monte-Carlo simulation approach was investigated by incorporating building facade thermal mass effect. Given a same hourly turbulence intensity distribution, the wind speeds with 1 rain frequency fluctuations were generated using a stochastic model, the modified GARCH model. Comparisons of natural ventilation profiles, effective ventilation rates, and air conditioning electricity use for a three-month period show statistically significant differences （for 80% confidence interval） between the new calculations and the traditional methods based on hourly average wind speed.
