Activity classification and anomaly detection using m-mediods based modelling of motion patterns

Techniques for video object motion analysis, behaviour recognition and event detection are becoming increasingly important with the rapid increase in demand for and deployment of video surveillance systems. Motion trajectories provide rich spatiotemporal information about an object's activity. This paper presents a novel technique for classification of motion activity and anomaly detection using object motion trajectory. In the proposed motion learning system, trajectories are treated as time series and modelled using modified DFT-based coefficient feature space representation. A modelling technique, referred to as m-mediods, is proposed that models the class containing n members with m mediods. Once the m-mediods based model for all the classes have been learnt, the classification of new trajectories and anomaly detection can be performed by checking the closeness of said trajectory to the models of known classes. A mechanism based on agglomerative approach is proposed for anomaly detection. Four anomaly detection algorithms using m-mediods based representation of classes are proposed. These includes: (i)global merged anomaly detection (GMAD), (ii) localized merged anomaly detection (LMAD), (iii) global un-merged anomaly detection (GUAD), and (iv) localized un-merged anomaly detection (LUAD). Our proposed techniques are validated using variety of simulated and complex real life trajectory datasets.     

Compositional metric learning for multi-label classification

Multi-label classification aims to assign a set of proper labels for each instance, where distance metric learning can help improve the generalization ability of instance-based multi-label classification models. Existing multi-label metric learning techniques work by utilizing pairwise constraints to enforce that examples with similar label assignments should have close distance in the embedded feature space. In this paper, a novel distance metric learning approach for multi-label classification is proposed by modeling structural interactions between instance space and label space. On one hand, compositional distance metric is employed which adopts the representation of a weighted sum of rank-1 PSD matrices based on component bases. On the other hand, compositional weights are optimized by exploiting triplet similarity constraints derived from both instance and label spaces. Due to the compositional nature of employed distance metric, the resulting problem admits quadratic programming formulation with linear optimization complexity w.r.t. the number of training examples.We also derive the generalization bound for the proposed approach based on algorithmic robustness analysis of the compositional metric. Extensive experiments on sixteen benchmark data sets clearly validate the usefulness of compositional metric in yielding effective distance metric for multi-label classification.     

Computability of probability measures and Martin-Löf randomness over metric spaces

In this paper, we investigate algorithmic randomness on more general spaces than the Cantor space, namely computable metric spaces. To do this, we first develop a unified framework allowing computations with probability measures. We show that any computable metric space with a computable probability measure is isomorphic to the Cantor space in a computable and measure-theoretic sense. We show that any computable metric space admits a universal uniform randomness test (without further assumption).     

Multi-granularity and metric spatial reasoning

Composition reasoning is a basic reasoning task in qualitative spatial reasoning (QSR). It is an important qualitative method for robot navigation, node localization in wireless sensor networks and other fields. The previous composition reasoning works dedicated in single granularity framework. Multi-granularity spatial relation is not rare in real world, and some qualitative spatial relation models are multi-granularity models, such as RCC, STAR_m, CDC_m and OPRA_m. Although multi-granularity composition reasoning is very useful in many applications, it has not been systematically studied before. A special case of multi-granularity composition reasoning, referred to as metric spatial reasoning, is also discussed here. The general frameworks and basic theories for multi-granularity and metric spatial reasoning are put forward here. Furthermore, we redefine the spatial relation models for distance, topology and direction under the proposed multi-granularity and metric frameworks. We add metric representation for the OPRA_m. The multi-granularity and metric reasoning tasks are studied for these four models for the first time. Finally we perform some experiments on OPRA_m with encouraging results to verify our theories. Multi-granularity and metric spatial reasoning tasks are new problems in QSR and quite different from the previous works. Our works can be potentially applied in robot navigation, wireless sensor networks and other applications.     

Motion-based behaviour learning, profiling and classification in the presence of anomalies

Techniques for understanding video object motion activity are becoming increasingly important with the widespread adoption of CCTV surveillance systems. Motion trajectories provide rich spatiotemporal information about an object's activity. This paper presents a novel technique for clustering and classification of motion. In the proposed motion learning system, trajectories are treated as time series and modelled using modified DFT (discrete fourier transform)-based coefficient feature space representation. A framework (iterative HSACT-LVQ (hierarchical semi-agglomerative clustering-learning vector quantization)) is proposed for learning of patterns in the presence of significant number of anomalies in training data. A novel modelling technique, referred to as m-Mediods, is also proposed that models the class containing n members with m Mediods. Once the m-Mediods-based model for all the classes have been learnt, the classification of new trajectories and anomaly detection can be performed by checking the closeness of said trajectory to the models of known classes. A mechanism based on agglomerative approach is proposed for anomaly detection. Our proposed techniques are validated using variety of simulated and complex real life trajectory data sets.     

Fuzzy C-means based clustering for linearly and nonlinearly separable data

In this paper we present a new distance metric that incorporates the distance variation in a cluster to regularize the distance between a data point and the cluster centroid. It is then applied to the conventional fuzzy C-means (FCM) clustering in data space and the kernel fuzzy C-means (KFCM) clustering in a high-dimensional feature space. Experiments on two-dimensional artificial data sets, real data sets from public data libraries and color image segmentation have shown that the proposed FCM and KFCM with the new distance metric generally have better performance on non-spherically distributed data with uneven density for linear and nonlinear separation.     

A joint feature selection framework for multivariate resource usage prediction in cloud servers using stability and prediction performance

Resource provisioning in cloud servers depends on future resource utilization of different jobs. As resource utilization trends vary dynamically, effective resource provisioning requires prediction of future resource utilization. The problem becomes more complicated as performance metrics related to one resource may depend on utilization of other resources also. In this paper, different multivariate frameworks are proposed for improving the future resource metric prediction in cloud. Different techniques for identifying the set of resource metrics relevant for the prediction of desired resource metric are analyzed. The proposed multivariate feature selection and prediction frameworks are validated for CPU utilization prediction in Google cluster trace. Joint analysis based on the prediction performance of the multivariate framework as well as its stability is used for selecting the most suitable feature selection framework. The results of the joint analysis indicate that features selected using the Granger causality technique perform best for multivariate resource usage prediction.     

Training of support vector machine with the use of multivariate normalization

SVM (support vector machines) techniques have recently arrived to complete the wide range of classification methods for complex systems. These classification systems offer similar performances to other classifiers (such as the neuronal networks or classic statistical classifiers) and they are becoming a valuable tool in industry for the resolution of real problems. One of the fundamental elements of this type of classifier is the metric used for determining the distance between samples of the population to be classified. Although the Euclidean distance measure is the most natural metric for solving problems, it presents certain disadvantages when trying to develop classification systems that can be adapted as the characteristics of the sample space change. Our study proposes a means of avoiding this problem using the multivariate normalization of the inputs (both during the training and classification processes). Using experimental results produced from a significant number of populations, the study confirms the improvement achieved in the classification processes. Lastly, the study demonstrates that the multivariate normalization applied to a real SVM is equivalent to the use of a SVM that uses the Mahalanobis distance measure, for non-normalized data.     

Generalized distances in digital geometry

A generalized distance measure called m-neighbor distance in n-D quantized space is presented. Its properties as a metric are examined. It is shown to give the shortest path length between two points in n-D digital space. An algorithm for finding such a shortest path between two points is presented. It is shown that lower dimension (2-D and 3-D) distance measures presently used in digital geometry can easily be derived as special cases. Other properties of m-neighbor distance are also examined.     

Computational Technique for an Efficient Classification of Protein Sequences With Distance‐Based Sequence Encoding Algorithm

Machine learning is being implemented in bioinformatics and computational biology to solve challenging problems emerged in the analysis and modeling of biological data such as DNA, RNA, and protein. The major problems in classifying protein sequences into existing families/superfamilies are the following: the selection of a suitable sequence encoding method, the extraction of an optimized subset of features that possesses significant discriminatory information, and the adaptation of an appropriate learning algorithm that classifies protein sequences with higher classification accuracy. The accurate classification of protein sequence would be helpful in determining the structure and function of novel protein sequences. In this article, we have proposed a distance‐based sequence encoding algorithm that captures the sequence's statistical characteristics along with amino acids sequence order information. A statistical metric‐based feature selection algorithm is then adopted to identify the reduced set of features to represent the original feature space. The performance of the proposed technique is validated using some of the best performing classifiers implemented previously for protein sequence classification. An average classification accuracy of 92% was achieved on the yeast protein sequence data set downloaded from the benchmark UniProtKB database.     

Feature fusion based on Wootters metric

For further enhancing the completeness and conciseness of the existing quantum-inspired feature fusion methods, this paper applies the quantum-related theories of Wootters metric and Fisher linear discriminant to dimension reduction and feature fusion. From the perspective of quantum metric spaces, i.e. phase space and probability space, this paper proposes two different feature fusion methods which take the Wootters statistical distance as the key factor to detect and fuse the duplicate feature data, and are different to the already developed quantum-inspired feature fusion methods. The experimental results reflect the superiority of the proposed feature fusion methods based on the Wootters metric for their better performances on relative completeness and conciseness.     

A cellular automata-based learning method for classification

Over the last few decades, classification applied to numerous applications in science, engineering, business and industries have rapidly been increased, especially for big data. However, classifiers dealing with complicated high dimension problems with non-conforming patterns with high accuracy are rare, especially for bit-level features. It is a challenging research problem. This paper proposed a novel efficient classifier based on cellular automata model, called Cellular Automata-based Classifier (CAC). CAC possesses the promising capability to deal with non-conforming patterns in the bit-level features. It was developed on a new kind of the proposed elementary cellular automata, called Decision Support Elementary Cellular Automata (DS-ECA). The classification capability of DS-ECA is promising since it can describe very complicated decision rule in high dimension problems with less complexity. CAC comprises double rule vectors and a decision function, the structure of which has two layers; the first layer is employed to evolve an input pattern into feature space and the other interprets the patterns in feature space as binary answer through the decision function. It has a time complexity of learning at O(n²), while the classification for one instance is O(1), where n is a number of bit patterns. For classification performance, 12 datasets consisting of binary and non-binary features are empirically implemented in comparison with Support Vector Machines (SVM) using k-fold cross validation. In this respect, CAC outperforms SVM with the best kernel for binary features, and provides the promising results equivalent to SVM on average for non-binary features.     

结合模型集成与特征融合的图像拷贝检测

基于内容的图像拷贝检测关键在于提取的图像特征能够针对不同形式的图像拷贝攻击具有不变性。现实中拷贝攻击手段变化多样,且存在很多相似图像的干扰,目前并没有任何一种图像特征可以对抗所有不同形式的图像攻击。现有方法虽然在图像特征表示上做了很多改进,但都局限于单个特征表示。因此从特征融合的角度对提取特征进行增强,基于卷积神经网络融合图像高层特征以及低层特征以实现特征多样性,集成ImageNet预训练分类模型以及提出的距离度量模型以实现特征互补性。度量模型针对该类问题在预训练模型的基础上通过学习合适的距离度量来对抗由于图像编辑引起的特征差异,拉近拷贝图像与原始图像在特征空间的距离。实验结果表明,结合模型集成和多层深度特征融合的方式可以有效增强特征的鲁棒性,相比单一特征的检测效果提升十分明显。     

Improving malicious URLs detection via feature engineering: Linear and nonlinear space transformation methods

In malicious URLs detection, traditional classifiers are challenged because the data volume is huge, patterns are changing over time, and the correlations among features are complicated. Feature engineering plays an important role in addressing these problems. To better represent the underlying problem and improve the performances of classifiers in identifying malicious URLs, this paper proposed a combination of linear and non-linear space transformation methods. For linear transformation, a two-stage distance metric learning approach was developed: first, singular value decomposition was performed to get an orthogonal space, and then a linear programming was used to solve an optimal distance metric. For nonlinear transformation, we introduced Nyström method for kernel approximation and used the revised distance metric for its radial basis function such that the merits of both linear and non-linear transformations can be utilized. 33,1622 URLs with 62 features were collected to validate the proposed feature engineering methods. The results showed that the proposed methods significantly improved the efficiency and performance of certain classifiers, such as k-Nearest Neighbor, Support Vector Machine, and neural networks. The malicious URLs’ identification rate of k-Nearest Neighbor was increased from 68% to 86%, the rate of linear Support Vector Machine was increased from 58% to 81%, and the rate of Multi-Layer Perceptron was increased from 63% to 82%. We also developed a website to demonstrate a malicious URLs detection system which uses the methods proposed in this paper. The system can be accessed at: http://url.jspfans.com.     

Data-dependent dissimilarity measure: an effective alternative to geometric distance measures

Nearest neighbor search is a core process in many data mining algorithms. Finding reliable closest matches of a test instance is still a challenging task as the effectiveness of many general-purpose distance measures such as \(\ell _p\)-norm decreases as the number of dimensions increases. Their performances vary significantly in different data distributions. This is mainly because they compute the distance between two instances solely based on their geometric positions in the feature space, and data distribution has no influence on the distance measure. This paper presents a simple data-dependent general-purpose dissimilarity measure called ‘\(m_p\)-dissimilarity’. Rather than relying on geometric distance, it measures the dissimilarity between two instances as a probability mass in a region that encloses the two instances in every dimension. It deems two instances in a sparse region to be more similar than two instances of equal inter-point geometric distance in a dense region. Our empirical results in k-NN classification and content-based multimedia information retrieval tasks show that the proposed \(m_p\)-dissimilarity measure produces better task-specific performance than existing widely used general-purpose distance measures such as \(\ell _p\)-norm and cosine distance across a wide range of moderate- to high-dimensional data sets with continuous only, discrete only, and mixed attributes.     

动作切分和流形度量学习的视频动作识别

目的 为了提高视频中动作识别的准确度,提出基于动作切分和流形度量学习的视频动作识别算法。方法 首先利用基于人物肢体伸展程度分析的动作切分方法对视频中的动作进行切分,将动作识别的对象具体化;然后从动作片段中提取归一化之后的全局时域特征和空域特征、光流特征、帧内的局部旋度特征和散度特征,构造一种7×7的协方差矩阵描述子对提取出的多种特征进行融合;最后结合流形度量学习方法有监督式地寻找更优的距离度量算法提高动作的识别分类效果。结果 对Weizmann公共视频集的切分实验统计结果表明本文提出的视频切分方法具有很好的切分能力,能够作好动作识别前的预处理;在Weizmann公共视频数据集上进行了流形度量学习前后的识别效果对比,结果表明利用流形度量学习方法对动作识别效果提升2.8%;在Weizmann和KTH两个公共视频数据集上的平均识别率分别为95.6%和92.3%,与现有方法的比较表明,本文提出的动作识别方法有更好的识别效果。结论 多次实验结果表明本文算法在预处理过程中动作切分效果理想,描述动作所构造协方差矩阵对动作的表达有良好的多特征融合能力,而且光流信息和旋度、散度信息的加入使得人体各部位的运动方向信息具有了更多细节的描述,有效提高了协方差矩阵的描述能力,结合流形度量学习方法对动作识别的准确性有明显提高。     

An efficient orientation distance–based discriminative feature extraction method for multi-classification

Feature extraction is an important step before actual learning. Although many feature extraction methods have been proposed for clustering, classification and regression, very limited work has been done on multi-class classification problems. This paper proposes a novel feature extraction method, called orientation distance–based discriminative (ODD) feature extraction, particularly designed for multi-class classification problems. Our proposed method works in two steps. In the first step, we extend the Fisher Discriminant idea to determine an appropriate kernel function and map the input data with all classes into a feature space where the classes of the data are well separated. In the second step, we put forward two variants of ODD features, i.e., one-vs-all-based ODD and one-vs-one-based ODD features. We first construct hyper-plane (SVM) based on one-vs-all scheme or one-vs-one scheme in the feature space; we then extract one-vs-all-based or one-vs-one-based ODD features between a sample and each hyper-plane. These newly extracted ODD features are treated as the representative features and are thereafter used in the subsequent classification phase. Extensive experiments have been conducted to investigate the performance of one-vs-all-based and one-vs-one-based ODD features for multi-class classification. The statistical results show that the classification accuracy based on ODD features outperforms that of the state-of-the-art feature extraction methods.     

On minimum distribution discrepancy support vector machine for domain adaptation

Domain adaptation learning (DAL) is a novel and effective technique to address pattern classification problems where the prior information for training is unavailable or insufficient. Its effectiveness depends on the discrepancy between the two distributions that respectively generate the training data for the source domain and the testing data for the target domain. However, DAL may not work so well when only the distribution mean discrepancy between source and target domains is considered and minimized. In this paper, we first construct a generalized projected maximum distribution discrepancy (GPMDD) metric for DAL on reproducing kernel Hilbert space (RKHS) based domain distributions by simultaneously considering both the projected maximum distribution mean and the projected maximum distribution scatter discrepancy between the source and the target domain. In the sequel, based on both the structure risk and the GPMDD minimization principle, we propose a novel domain adaptation kernelized support vector machine (DAKSVM) with respect to the classical SVM, and its two extensions called LS-DAKSVM and μ-DAKSVM with respect to the least-square SVM and the v-SVM, respectively. Moreover, our theoretical analysis justified that the proposed GPMDD metric could effectively measure the consistency not only between the RKHS embedding domain distributions but also between the scatter information of source and target domains. Hence, the proposed methods are distinctive in that the more consistency between the scatter information of source and target domains can be achieved by tuning the kernel bandwidth, the better the convergence of GPMDD metric minimization is and thus improving the scalability and generalization capability of the proposed methods for DAL. Experimental results on artificial and real-world problems indicate that the performance of the proposed methods is superior to or at least comparable with existing benchmarking methods.     

Learning distance metric for object contour tracking

Contour tracking can be implemented by measuring the probability distributions (e.g. intensity, color and texture) of both interior and exterior regions of an object contour. Choosing a suitable distance metric for measuring the (dis)similarity between two distributions significantly influences the tracking performance. Most existing contour tracking methods, however, utilize a predefined metric which may not be appropriate for measuring the distributions. This paper presents a novel variational level set framework for contour tracking. The image energy functional is modeled by the distance between the foreground distribution and the given template, divided by the distance between the background distribution and the template. The form of the distance between two distributions is represented by the quadratic distance (Rubner et al. in Int J Comput Vis 40(2):99–121, 2000). To obtain the more robust tracking results, a distance metric learning algorithm is employed to achieve the similarity matrix for the quadratic distance. In addition, a distance between the evolving contour and the zero level set of the reference shape function is adopted as the shape prior to constrain the contour evolution process. Experiments on several video sequences prove the effectiveness and robustness of our method.