层次分析法中基于SVDD群体信息提取方法

针对层次分析法中群决策问题提出一种基于支持向量域描述(SVDD)的集结方法.首先利用生成树的方法把判断矩阵进行一致性剖分;然后利用支持向量域描述的方法排除干扰信息,找出群体公共信息,引入群体相容性,最优解等概念,提出并证明了关于群体信息球的特性;根据最大特征值法把群体信息球中的向量合成为群决策最优解,即关于方案的排序向量.并通过一个具体示例给出该方法的算法步骤同时显示了该方法的可行性和有效性.     

Multi-task learning for video anomaly detection

We propose a multi-task learning framework for video anomaly detection based on a novel pipeline. Our model contains two crossing streams, one stream employs the backbone of Attention-R2U-net for future frame prediction, while the other is designed based on an encoder–decoder network to reconstruct the input optical flow maps. In addition, the latent layers of the two streams are merged together and assigned with a Deep SVDD-based loss at each location individually. Through the combination of these three tasks, the two-stream-crossing pipeline can be trained end-to-end to provide a comprehensive evaluation for the anomaly targets. Experimental results on several popular benchmark datasets show that our model outperforms the state-of-the-art competing models, which can be applied to different types of anomalous targets and meanwhile achieves remarkable precision.     

光纤传感系统振动信号模式识别的研究

提出了一种基于MFCC特征和SVDD训练的光纤传感系统振动信号的识别算法,该算法选择MFCC系数作为模式识别的特征量,利用SVDD算法实现对特征量的训练和模式匹配,同时对SVDD算法进行了扩展,使其能够利用单个事件进行模板库训练和多类别识别。实验证明,该算法在光纤传感系统应用中,能够准确识别常见扰动事件的类别,具有较高的识别率,并简化了模板库的建立过程。     

基于高光谱图像技术和SVDD的玉米种子识别

特征提取的充分性和分类器设计的合理性是影响玉米种子识别精度的两个关键问题。采集了玉米种子的高光谱图像,并提取每粒玉米种子在不同波段下的图像熵作为分类特征;在此基础上,利用支持向量数据描述方法构建每类玉米的分类器模型,对待识别样本的测试精度达到了94.14%,对新类别样本的识别精度达到92.28%。仿真结果表明：新方法可实现玉米种子的准确识别,同时解决了传统分类器对新类别样本的错误分类问题。     

高光谱图像混合像元分解算法

传统的高光谱图像混合像元分解技术包括端元提取和估计每个端元的混合比例.虽然很多模型都能得到可以接受的解混结果,但是一些未知端元的存在使得结果在包含未知端元的像素点处出现偏差.因此,提出了一种基于支持向量数据描述的高光谱图像混合像元分解算法.首先高光谱图像数据被分成类内和类外两部分,类内是完全由已知端元数据混合的像素点,而类外数据是包含未知端元的像素点.两类数据交界处被认为是已知端元和未知端元混合的数据.然后再对这些像素点进行混合像元分解,分别对仿真数据和真实高光谱图像进行实验.结果表明该算法可以有效地解决因存在未知端元对解混精度的影响,而且能给出未知端元的解混分量.该方法的解混结果几乎不受未知端元的影响,优于直接解混结果     

Review of validation and reporting of non-targeted fingerprinting approaches for food authentication

Food fingerprinting approaches are expected to become a very potent tool in authentication processes aiming at a comprehensive characterization of complex food matrices. By non-targeted spectrometric or spectroscopic chemical analysis with a subsequent (multivariate) statistical evaluation of acquired data, food matrices can be investigated in terms of their geographical origin, species variety or possible adulterations. Although many successful research projects have already demonstrated the feasibility of non-targeted fingerprinting approaches, their uptake and implementation into routine analysis and food surveillance is still limited. In many proof-of-principle studies, the prediction ability of only one data set was explored, measured within a limited period of time using one instrument within one laboratory. Thorough validation strategies that guarantee reliability of the respective data basis and that allow conclusion on the applicability of the respective approaches for its fit-for-purpose have not yet been proposed. Within this review, critical steps of the fingerprinting workflow were explored to develop a generic scheme for multivariate model validation. As a result, a proposed scheme for “good practice” shall guide users through validation and reporting of non-targeted fingerprinting results. Furthermore, food fingerprinting studies were selected by a systematic search approach and reviewed with regard to (a) transparency of data processing and (b) validity of study results. Subsequently, the studies were inspected for measures of statistical model validation, analytical method validation and quality assurance measures. In this context, issues and recommendations were found that might be considered as an actual starting point for developing validation standards of non-targeted metabolomics approaches for food authentication in the future. Hence, this review intends to contribute to the harmonization and standardization of food fingerprinting, both required as a prior condition for the authentication of food in routine analysis and official control.     

基于支持向量域描述的雷达地面目标鉴别技术

在高分辨一维距离像目标识别中,有效的对库内目标特征空间进行描述,并且对库外目标进行鉴别是一个关键问题.本文提出了一种基于非均匀特征向量分布的目标鉴别器设计方法,该方法利用训练特征空间的协方差分布情况,选择库内协方差较小,即样本密度较大的区域进行细致描述,有效克服了训练样本分布非均匀造成的特征空间描述偏差,进而保证对库内...     

基于训练特征空间分布的雷达地面目标鉴别器设计

该文对雷达地面目标高分辨1维距离像目标识别中的库外目标鉴别问题,提出一种基于训练特征空间分布的雷达地面目标鉴别器。在训练阶段利用基于相关系数预处理的K-Means聚类方法对库内目标样本特征空间进行区域划分,并采用基于空间分布的支撑向量域描述方法确定样本特征空间的边界与支撑向量,利用样本特征空间边界与加权K近邻原则对目标类别进行判决。该方法解决了库内目标与库外目标的鉴别问题,提高了目标识别系统的总体性能。针对多种不同姿态下目标特征空间非均匀聚合的特点,对训练样本特征空间进行区域划分,减小模板匹配搜索运算规模,保证目标鉴别所需的实时性工作要求。最后通过仿真和实测数据验证了该方法具备优良的鉴别性能与良好的实时处理能力。