一种基于高斯混合模型的不均衡分类方法

为应对不均衡分类问题,提高分类准确率,提出了一种基于高斯混合模型的混合采样集成方法GMHSE(Gaussian-Mixture-model-based Hybrid Sampling Ensemble method),首先通过高斯混合模型将数据划分成多个类簇,然后在每个类簇上混合采样获得多个数据子集,最后基于Baggi...     

面向不均衡数据的融合谱聚类的自适应过采样法

分类是模式识别领域中的研究热点,大多数经典的分类器往往默认数据集是分布均衡的,而现实中的数据集往往存在类别不均衡问题,即属于正常/多数类别的数据的数量与属于异常/少数类数据的数量之间的差异很大。若不对数据进行处理往往会导致分类器忽略少数类、偏向多数类,使得分类结果恶化。针对数据的不均衡分布问题,本文提出一种融合谱聚类的综合采样算法。首先采用谱聚类方法对不均衡数据集的少数类样本的分布信息进行分析,再基于分布信息对少数类样本进行过采样,获得相对均衡的样本,用于分类模型训练。在多个不均衡数据集上进行了大量实验,结果表明,所提方法能有效解决数据的不均衡问题,使得分类器对于少数类样本的分类精度得到提升。     

基于变分贝叶斯高斯混合噪声模型的机器人跨模态生成对抗网络

多模态融合对于机械手充分感知外界环境至关重要,单一模态信息会限制机械手对物体的识别、抓取能力,而传统的跨模态数据生成方法生成的图像效果较差,导致多模态融合效果并不理想.为了解决跨模态生成的图像效果差和多模态融合的数据缺乏等问题,本文提出变分贝叶斯高斯混合条件生成对抗网络(BGM-CGAN)的跨模态多样性噪声数据生成式方...     

基于改进高斯混合模型的自适应前景提取

在复杂场景下的运动前景提取是智能视频监控的基础部分。高斯混合模型是常用的背景建模方法,针对高斯混合模型中模型个数固化导致的无谓的系统开销,提出基于单高斯模型成长的动态个数调整形成的高斯混合模型。对模型的更新率根据场景变化的剧烈程度进行实时改变,能较好适应突发场景、光照的变化。对提取的运动前景进行形态学处理,得到最后的提取目标。实验结果表明,该方法背景建模适应性强,提取前景精度有所提升。     

基于变分贝叶斯推断的字典学习算法

传统的字典学习算法在对训练图像进行学习时收敛速率慢,当图像受到噪声干扰时学习效果变差.对此,提出一种基于变分推断的字典学习算法.首先设定模型中各参数的共轭稀疏先验分布;然后基于贝叶斯网络求出所有参数的联合概率密度函数;最后利用变分贝叶斯推断原理计算出各参数的最优边缘分布,训练出自适应学习字典.利用该字典进行图像去噪实验以及压缩感知重构实验,仿真结果表明,所提出的算法可显著提高字典学习效率,对测试图像的去噪效果和重构精度有很大改善.     

基于局部变分贝叶斯推断的分布式交互式多模型估计

针对目前部分多模型算法预先设定运动模型转移概率矩阵对状态估计精度的不利影响,本文提出了一种基于局部变分贝叶斯推断的分布式交互式多模型估计算法.不同于传统交互式多模型估计中运动模型转移概率矩阵为先验已知的假设条件,在分布融合估计框架下,首先基于最小化Kullback-Leibler散度准则的递归优化策略实现对运动模型转移概率矩阵的预测与更新;在此基础上,结合变分贝叶斯推断实现对当前时刻目标状态与模型概率的联合估计;最后依据协方差交叉融合策略完成对局部状态估计融合.仿真结果表明:新算法通过对运动模型转移概率矩阵以及模型概率自适应在线估计,有效提升了机动目标的状态估计精度.     

基于混合高斯模型的目标差分自适应背景模型

提出一种自适应背景更新率的方法。在混合高斯背景模型基础上。通过混合高斯匹配将前后两帧分割出的运动目标作差分。差分后再次采用混合高斯匹配分辨出过去目标区和当前目标区。在背景更新的过程中,加大过去目标区的更新率,让运动目标在出现停滞状态时给背景模型带来的干扰区域得到快速恢复。试验结果表明,该方法有效地提高背景模型的鲁棒性。     

二值probit回归模型的坍缩变分贝叶斯推断算法

给出了二值probit回归模型的坍缩变分贝叶斯推断算法.此算法比变分贝叶斯推断算法能更逼近对数边缘似然,得到更精确的模型参数后验期望值.如果两个算法得到的分类错误一致,则该算法的迭代次数较变分法明显减少.仿真实验结果验证了所提出算法的有效性.     

Bayesian change detection based on spatial sampling and Gaussian mixture model

A Gaussian mixture model (GMM) and Bayesian inferencing based unsupervised change detection algorithm is proposed to achieve change detection on the difference image computed from satellite images of the same scene acquired at different time instances. Each pixel of the difference image is represented by a feature vector constructed from the difference image values of the neighbouring pixels to consider the contextual information. The feature vectors of the difference image are modelled as a GMM. The conditional posterior probabilities of changed and unchanged pixel classes are automatically estimated by partitioning GMM into two distributions by minimizing an objective function. Bayesian inferencing is then employed to segment the difference image into changed and unchanged classes by using the conditional posterior probability of each class. Change detection results are shown on real datasets.     

复杂条件下高斯混合模型的自适应背景更新

针对高斯混合模型背景更新中面临的光照突变和目标与背景相互转化的问题,提出一种分情况分区域的背景自适应更新算法。首先根据当前检测目标的面积大小判别是否发生光照突变情况,采取针对性更新策略,对于未发生光照突变情况再分背景区域和目标区域分别进行背景自适应更新。其中,重点讨论了目标区域的背景更新问题,提出根据目标尺寸、运动速度和匹配次数等特征参数来调整目标区域的背景更新速率。仿真结果表明,该算法在保证了目标检测完整性的同时,提高了模型对背景变化的适应能力。     

Target detection algorithm and data model optimization based on improved Gaussian mixture model

A set of multi-stage image processing algorithms developed to detect the change of image sequence analysis. This means Moving Target Identification (MTI). The difference compared with the traditional method, the new method of spatial diversity. General purpose of object detection, image processing and image understanding of the specific field of research. Hue Saturation and Value (HSV,) colour space algorithm uses spot and shape detection to ensure detection under various conditions of lighting, shade, and distance. The algorithm is tested on the lighting and form detection account where the different variation of the displays. The only computational process of the challenge group is the presence of more than one target of the same colour in finding the correct target under changing lighting conditions. It has been found that the elasticity target people based on localization of these image processing methods for better detection is compared with the target detection. Digital image data contains most of this image data recognition model is optimized by integrating task planning normalization and inertial representation of the remote sensing image classification model based on spatial communication. The immediate message issue is that the Gaussian compound model cannot detect the entire moving object, and is prone to sudden light changes, etc. The advanced algorithm to be performed is proposed based on the Gaussian compound model and the detection method of the three legal difference moving object. Subsequently, a new adaptive selection technique for Gaussian distributions is introduced to reduce processing time and improve detection accuracy.     

不平衡数据中基于权重的边界混合采样

针对单一的不平衡数据分类算法合成样本质量不佳、未考虑类内样本分布等局限性,提出一种不平衡数据中基于权重的边界混合采样(boundary mixed sampling based on weight selection in imbalanced data,BWBMS)。剔除噪声样本并引入边界因子概念,把原样本空间分成边界集和非边界集;考虑类内样本分布,对于边界集中每个少数类样本赋予支持度权重和密度权重并增加采样比重将其划分为两类,对两类样本子集采用不同的过采样算法和过采样倍率;考虑不同区域样本重要性的不同,根据多数类样本距离其最近的k个异类近邻的平均距离来删减部分非边界集多数类样本点。实验结果表明,结合SVM分类器的BWBMS算法在不同数据集上的性能指标得到了提升,验证了其有效性。     

复杂场景下自适应背景减除算法

目的 复杂场景下的背景减除是智能视频监控研究领域的研究重点和热点之一.针对混合高斯模型中高斯分布个数固定和参数初始化粗糙问题,提出一种应用于复杂场景中的基于混合高斯模型的自适应背景减除算法(AMGBS).方法 通过灰度值归类算法自适应调整模型的高斯分布个数,使得背景模型能够适应场景的变化,并且结合在线 K均值(online K-means)算法和在线期望最大化(online EM)算法初始化混合高斯模型参数.结果 针对灰度值统计结果调整高斯分布数,以及采用优化参数初始化过程,实验表明,本文方法的平均查准率和平均查全率比传统的混合高斯算法高出10%左右,比其他改进的混合高斯算法高出2%左右.结论 提出一种新的自适应背景减除算法,针对灰度值统计结果调整高斯分布数,以及采用优化参数初始化过程.实验结果表明,该方法对复杂场景有较强的适应能力,能够有效快速地完成背景减除,进而实现运动目标的提取.     

混合高斯模型的自适应前景提取

复杂场景下的运动前景提取是计算机视觉研究领域的研究重点。为解决复杂场景中的前景目标提取问题,本文提出一种应用于复杂变化场景中的基于混合高斯模型的自适应前景提取方法。本方法可以对视频帧中每个像素的高斯分布数进行动态控制,并且通过在线EM算法对高斯分布的各参数进行学习,此外每个像素的权值更新速率可根据策略进行调整。实验结果表明本方法对复杂变化场景具有较好的适应性,可有效、快速地提取前景目标,提取结果具有较好的查准率和查全率。     

Generative learning for imbalanced data using the Gaussian mixed model

Imbalanced data classification, an important type of classification task, is challenging for standard learning algorithms. There are different strategies to handle the problem, as popular imbalanced learning technologies, data level imbalanced learning methods have elicited ample attention from researchers in recent years. However, most data level approaches linearly generate new instances by using local neighbor information rather than based on overall data distribution. Differing from these algorithms, in this study, we develop a new data level method, namely, generative learning (GL), to deal with imbalanced problems. In GL, we fit the distribution of the original data and generate new data on the basis of the distribution by adopting the Gaussian mixed model. Generated data, including synthetic minority and majority classes, are used to train learning models. The proposed method is validated through experiments performed on real-world data sets. Results show that our approach is competitive and comparable with other methods, such as SMOTE, SMOTE-ENN, SMOTE-TomekLinks, Borderline-SMOTE, and safe-level-SMOTE. Wilcoxon signed rank test is applied, and the testing results show again the significant superiority of our proposal.     

基于混合高斯模型的阴影去除算法

阴影去除是智能视频领域中运动目标识别的一项重要内容,其结果直接影响目标识别的准确性。针对当前基于纹理特征的阴影去除算法的不足,提出一种结合YCbCr颜色空间和混合高斯模型(GMM)的阴影去除算法。首先利用混合高斯模型提取出运动区域;然后通过分析运动区域的前景和背景在YCbCr颜色空间的差值统计特性,建立混合高斯阴影模型;最后根据高斯分布的概率分布规律,得到阴影分布特性,从而实现对阴影的去除。对于实验中的序列图像,所提算法有70%以上的阴影检测率。实验结果表明,所提方法能够在不同的场合快速有效地去除阴影,准确地提取运动目标。     

基于高斯混合模型的脚步声身份识别方法

针对脚步声身份识别数据少、识别准确率低的问题,提出一种以相邻两个脚步声信号作为基本单元,基于高斯混合模型(GMM)实现脚步声身份识别的方法.首先,使用滤波和谱减法相结合的方法对原始脚步声信号进行降噪处理;其次,根据脚步声压级信号和自适应阈值对自然行走状态下连续脚步声信号的起始点进行检测,并将相邻两个脚步声信号作为一个脚步声单元;最后,以声学特征梅尔频率倒谱系数(MFCC)、感知线性预测(PLP)系数作为脚步声单元的特征参数,使用GMM对特征参数进行建模和识别,实现行走人的身份识别.在200人的数据集中进行仿真实验,识别率最高达到98％;同时,在仅用一个脚步声单元进行身份识别时,识别率达到67.5％.仿真结果表明,所提方法能够在数据集人数较多时达到较高的识别率;且在测试数据不足的情况下,识别率显著提高,极大地提高了脚步声身份识别系统的性能.     

A Gaussian mixture model based discretization algorithm for associative classification of medical data

Knowledge-based systems such as expert systems are of particular interest in medical applications as extracted if-then rules can provide interpretable results. Various rule induction algorithms have been proposed to effectively extract knowledge from data, and they can be combined with classification methods to form rule-based classifiers. However, most of the rule-based classifiers can not directly handle numerical data such as blood pressure. A data preprocessing step called discretization is required to convert such numerical data into a categorical format. Existing discretization algorithms do not take into account the multimodal class densities of numerical variables in datasets, which may degrade the performance of rule-based classifiers. In this paper, a new Gaussian Mixture Model based Discretization Algorithm (GMBD) is proposed that preserve the most frequent patterns of the original dataset by taking into account the multimodal distribution of the numerical variables. The effectiveness of GMBD algorithm was verified using six publicly available medical datasets. According to the experimental results, the GMBD algorithm outperformed five other static discretization methods in terms of the number of generated rules and classification accuracy in the associative classification algorithm. Consequently, our proposed approach has a potential to enhance the performance of rule-based classifiers used in clinical expert systems.