Train SVM Research Articles

An Imbalanced Spam Mail Filtering Method

The high false alarm rate appears during the traditional spam recognition method processing the large-scale unbalanced data. A method which transforms the unbalanced issue into the balanced issue is proposed, when the K-means clustering algorithm is improved based on the support vector machine classification model, to obtain the balanced training set. Firstly, the improved K-means clustering algorithm clusters spam and extracts the typical spam，then the training set consists of the typical spam and legitimate messages, and finally the goal of the filtration of spam is realized by trained SVM classification model. Comparing the K-SVM filtration method to standard SVM method through the experiment, the result indicates that the K-SVM filtration method in large-scale unbalance data set can obtain high classified efficiency and the generalization performance.

Support Vector Machine을 이용한 선에코 특성 분석 및 탐지 방법

SVM은 학습 데이터를 두 개의 집단으로 분리시키는 최적의 초평면을 찾는 이진 분류기로서 우수한 성능 때문에 다양한 분야에서 귀납 추론, 이진 분류, 예측 등을 목적으로 사용되는 알고리즘이다. 또한 대표적인 블랙박스 모델 중 하나이기 때문에 학습 후 생성되는 SVM의 해석에 대한 연구도 활발히 진행되고 있다. 본 논문에서는 SVM 알고리즘을 이용하여 기상 레이더의 데이터 내에 비교적 높은 빈도로 발생하여 기상 예보의 정확도를 감소시키는 비강수에코 중 하나인 선에코를 자동으로 탐지하는 방법에 대한 연구를 수행하였다. 학습 데이터로는 평균 반사도, 크기, 발생 형태, 중심 고도 등과 같은 특성을 활용하였는데, 이는 기상 레이더 데이터에 저장된 다양한 데이터 중 반사도 값을 선택한 후 클러스터링 기법을 통해 추출한 것이다. 이와 같이 학습된 SVM 분류기를 실제 사례를 바탕으로 하여 검증하였으며, Decision Tree 알고리즘을 적용하여 생성한 분류기의 해석을 수행하였다. A SVM is a kind of binary classifier in order to find optimal hyperplane which separates training data into two groups. Due to its remarkable performance, the SVM is applied in various fields such as inductive inference, binary classification or making predictions. Also it is a representative black box model; there are plenty of actively discussed researches about analyzing trained SVM classifier. This paper conducts a study on a method that is automatically detecting the line-shaped echoes, sun strobe echo and radial interference echo, using the SVM algorithm because the line-shaped echoes appear relatively often and disturb weather forecasting process. Using a spatial clustering method and corrected reflectivity data in the weather radar, the training data is made up with mean reflectivity, size, appearance, centroid altitude and so forth. With actual occurrence cases of the line-shaped echoes, the trained SVM classifier is verified, and analyzed its characteristics using the decision tree method.

Local contrast phase descriptor for fingerprint liveness detection

We propose a new local descriptor for fingerprint liveness detection. The input image is analyzed both in the spatial and in the frequency domain, in order to extract information on the local amplitude contrast, and on the local behavior of the image, synthesized by considering the phase of some selected transform coefficients. These two pieces of information are used to generate a bi-dimensional contrast-phase histogram, used as feature vector associated with the image. After an appropriate feature selection, a trained linear-kernel SVM classifier makes the final live/fake decision. Experiments on the publicly available LivDet 2011 database, comprising datasets collected from various sensors, prove the proposed method to outperform the state-of-the-art liveness detection techniques.

Research on Combination Optimization of Parameters and Character Choice for SVM Based on Simulated Annealing and Improved QPSO

In order to improve the classification accuracy of SVM, combination optimization of parameters and character choice for SVM was proposed. Improved QPSO algorithm was researched on in this paper. At the same time, simulated annealing and improved QPSO algorithm was adopted to train SVM in this paper in order to improve its training speed and diagnosis precision. At last the method mentioned above was applied to train and validate SVM using UCI data, and the result showed that this method was very good.

Weld Defect Detection of X-ray Images Based on Support Vector Machine

ABSTRACTSupport vector machines (SVM) have received widespread attention for their excellent ability to learn, and have been applied in many fields. The application of SVM in weld defect defection of X-ray images is receiving more and more attention, which has led to the rapid development of weld defect detection automation. In this paper, a method of weld defect detection based on SVM is proposed. First, all potential defects are detected by a novel method based on grey-level profile analysis. Then three feature vectors are extracted from potential weld defects and used to train the SVM. Lastly, the trained SVM is used to distinguish real defects from potential defects. Results show that the proposed method can achieve a high correct defection rate.

Outdoor Navigating Scene Labeling Using Importance Factor Based I-RELIEF and Feature Weighted Support Vector Machines

Feature selection plays an important role in terrain classification for outdoor robot navigation. For terrain classification, the image data usually have a large number of feature dimensions. The better selection of features usually results in higher labeling accuracy. In this work, a novel approach for terrain perception using Importance Factor based I-Relief algorithm and Feature Weighted Support Vector Machines (IFIR-FWSVM) is put forward. Firstly, the weight of each feature for classification is computed by using Importance Factor based I-Relief algorithm (IFIR) and the irrelevant features are eliminated. Then the weighted features are used to compute the kernel functions of SVM and trained the classifier. Finally, the trained SVM is employed to predict the terrain label in the far-field regions. Experimental results based on DARPA datasets show that the proposed method IFIR-FWSVM is superior over traditional SVM.

Study on Detection Method for Crack in Eggs Based on Computer Vision and Support Vector Machine Neural Network

To improve the accuracy of detection and classification of egg with cracks, this paper is to add Support Vector Machine to neural network to automatically identify and classify the eggs with cracks. Firstly process the egg images with light-transmitting were obtained by the computer vision device including denoising, threshold segmentation. Five characteristic parameters of crack areas and noise areas were acquired. Secondly train SVM Neural Network and identify the eggs with cracks by five parameters data as the sample data. The correct discerning rate of grading table eggs is 98.07%. It proves better than traditional method in terms of prediction accuracy and robustness. The generalization ability of SVM Neural Network is strengthened.

Power Supply Quality Analysis Using S-Transform and SVM Classifier

In this paper, a SVM classifier based on S-Transform is presented for power quality disturbances classification. Firstly, seven types of PQ events are created using Matlab simulation. These signals are analyzed to detect and localize PQ events via S-Transform by visual inspection. Then five significant features of the PQ disturbances are extracted from the S-Transform output. Afterwards, PQ disturbance samples with the five features are fed to SVM for training and automatic classification. Besides, particle swarm optimization is implemented to improve the performance of SVM. The results of the classification indicate that SVM classifier is an effective mechanism to detect and classify power quality disturbances.

An adaptive superpixel based hand gesture tracking and recognition system.

We propose an adaptive and robust superpixel based hand gesture tracking system, in which hand gestures drawn in free air are recognized from their motion trajectories. First we employed the motion detection of superpixels and unsupervised image segmentation to detect the moving target hand using the first few frames of the input video sequence. Then the hand appearance model is constructed from its surrounding superpixels. By incorporating the failure recovery and template matching in the tracking process, the target hand is tracked by an adaptive superpixel based tracking algorithm, where the problem of hand deformation, view-dependent appearance invariance, fast motion, and background confusion can be well handled to extract the correct hand motion trajectory. Finally, the hand gesture is recognized by the extracted motion trajectory with a trained SVM classifier. Experimental results show that our proposed system can achieve better performance compared to the existing state-of-the-art methods with the recognition accuracy 99.17% for easy set and 98.57 for hard set.

Detecting Tower Crane with Multi-Features

In this paper, we present a new algorithm for tower crane detection in images with multi-feature analysis. The image will be segmented into several objects first, and then for each object, its color, texture and geometric features are extracted, and at last, all features extracted are fed to a trained SVM to generate the final judgment. Experiments show that our algorithm can detect tower crane in images with high recall and precision.

Open set source camera attribution and device linking

Camera attribution approaches in digital image forensics have most often been evaluated in a closed set context, whereby all devices are known during training and testing time. However, in a real investigation, we must assume that innocuous images from unknown devices will be recovered, which we would like to remove from the pool of evidence. In pattern recognition, this corresponds to what is known as the open set recognition problem. This article introduces new algorithms for open set modes of image source attribution (identifying whether or not an image was captured by a specific digital camera) and device linking (identifying whether or not a pair of images was acquired from the same digital camera without the need for physical access to the device). Both algorithms rely on a new multi-region feature generation strategy, which serves as a projection space for the class of interest and emphasizes its properties, and on decision boundary carving, a novel method that models the decision space of a trained SVM classifier by taking advantage of a few known cameras to adjust the decision boundaries to decrease false matches from unknown classes. Experiments including thousands of unconstrained images collected from the web show a significant advantage for our approaches over the most competitive prior work.

Training sparse SVM on the core sets of fitting-planes

Controlling the sparsity of a classifier is a key to train SVM efficiently on very large scale problems. This paper explores building SVM classifier on the fitting-plane of each class of data, which captures the distributing trend of the corresponding class of data. The newly developed plane-fitting model can be solved by core set methods, and the SVM is trained only on the core sets which are small subsets of the original data. The computing complexity of the proposed algorithm is up bounded by Ο(1/ε). Experimental results show that the new algorithm scales better than SVMperf and CVM/BVM, while their predicting accuracies are almost comparable.

Action Recognition Based on Motion Representing and Reconstructed Phase Spaces Matching of 3D Joint Positions

This paper presents an efficient and novel framework for human action recognition based on representing the motion of human body-joints and the theory of nonlinear dynamical systems. Our work is motivated by the pictorial structures model and advances in human pose estimation. Intuitively, a collective understanding of human joints movements can lead to a better representation and understanding of any human action through quantization in the polar space. We use time-delay embedding on the time series resulting of the evolution of human body-joints variables along time to reconstruct phase portraits. Moreover, we train SVM models for action recognition by comparing the distances between trajectories of human body-joints variables within the reconstructed phase portraits. The proposed framework is evaluated on MSR-Action3D dataset and results compared against several state-of-the-art methods.

Optimal design of an in-situ bioremediation system using support vector machine and particle swarm optimization

A methodology based on support vector machine and particle swarm optimization techniques (SVM-PSO) was used in this study to determine an optimal pumping rate and well location to achieve an optimal cost of an in-situ bioremediation system. In the first stage of the two stage methodology suggested for optimal in-situ bioremediation design, the optimal number of wells and their locations was determined from preselected candidate well locations. The pumping rate and well location in the first stage were subsequently optimized in the second stage of the methodology. The highly nonlinear system of equations governing in-situ bioremediation comprises the equations of flow and solute transport coupled with relevant biodegradation kinetics. A finite difference model was developed to simulate the process of in-situ bioremediation using an Alternate-Direction Implicit technique. This developed model (BIOFDM) yields the spatial and temporal distribution of contaminant concentration for predefined initial and boundary conditions. BIOFDM was later validated by comparing the simulated results with those obtained using BIOPLUME III for the case study of Shieh and Peralta (2005). The results were found to be in close agreement. Moreover, since the solution of the highly nonlinear equation otherwise requires significant computational effort, the computational burden in this study was managed within a practical time frame by replacing the BIOFDM model with a trained SVM model. Support Vector Machine which generates fast solutions in real time was considered to be a universal function approximator in the study. Apart from reducing the computational burden, this technique generates a set of near optimal solutions (instead of a single optimal solution) and creates a re-usable data base that could be used to address many other management problems. Besides this, the search for an optimal pumping pattern was directed by a simple PSO technique and a penalty parameter approach was adopted to handle the constraints in the PSO. The results showed that the costs involved in the proposed management solution were consistent with that resulting from other nontraditional optimization techniques which use embedded/linked bioremediation simulation models. Moreover, an optimal transient pumping strategy resulted in an overall reduction in pumping cost by almost 20% when compared to cases where a steady state pumping strategy was adopted. A considerable reduction in the number of simulations was achieved using the SVM approach.

$L_{p}$ Norm Localized Multiple Kernel Learning via Semi-Definite Programming

Our objective is to train SVM based Localized Multiple Kernel Learning with arbitrary <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex Notation="TeX">$l_{p}$</tex> </formula> -norm constraint using the alternating optimization between the standard SVM solvers with the localized combination of base kernels and associated sample-specific kernel weights. Unfortunately, the latter forms a difficult <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex Notation="TeX">$l_{p}$</tex></formula> -norm constraint quadratic optimization. In this letter, by approximating the <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex Notation="TeX">$l_{p}$</tex></formula> -norm using Taylor expansion, the problem of updating the localized kernel weights is reformulated as a non-convex quadratically constraint quadratic programming, and then solved via associated convex Semi-Definite Programming relaxation. Experiments on ten benchmark machine learning datasets demonstrate the advantages of our approach.

Model-based recognition of human actions by trajectory matching in phase spaces

This paper presents a human action recognition framework based on the theory of nonlinear dynamical systems. The ultimate aim of our method is to recognize actions from multi-view video. We estimate and represent human motion by means of a virtual skeleton model providing the basis for a view-invariant representation of human actions. Actions are modeled as a set of weighted dynamical systems associated to different model variables. We use time-delay embeddings on the time series resulting of the evolution of model variables along time to reconstruct phase portraits of appropriate dimensions. These phase portraits characterize the underlying dynamical systems. We propose a distance to compare trajectories within the reconstructed phase portraits. These distances are used to train SVM models for action recognition. Additionally, we propose an efficient method to learn a set of weights reflecting the discriminative power of a given model variable in a given action class. Our approach presents a good behavior on noisy data, even in cases where action sequences last just for a few frames. Experiments with marker-based and markerless motion capture data show the effectiveness of the proposed method. To the best of our knowledge, this contribution is the first to apply time-delay embeddings on data obtained from multi-view video.

Morphological Filtering Algorithm for Estimation of Tool Wear in CNC Milling Machining

Estimation of tool condition has very important meaning to improve the product quality, continuous machining ability and reliability of the manufacturing system. Based on mathematical morphology, a systematic approach is developed to implement online estimation of tool wear in this paper. As the nonlinear filter, morphological filter is selected to reduce the higher frequency noises before feature values extraction. The feature vector consists of original characteristics of vibration signal and cutting force signal. Then, they are input into SVM for training and testing. Experiments show that this method can achieve tool wear estimation effectively.

The Visual Quality Recognition of Nonwovens Using a Novel Wavelet Based Contourlet Transform

In this paper, a novel wavelet based contourlet transform for texture extraction is presented. In the texture analysis section, we propose a novel wavelet based contourlet transform, which can be considered as a simplified but more sufficient for texture analysis for nonwoven image compared with version of the one introduced by Eslami in theory view. In experiment, nonwoven images of five different visual quality grades, 125 of each grade, are decomposed using wavelet based contourlet transform with ‘PKVA’ filter as the default filter of Laplacian Pyramid (LP) and Directional Filter Bank (DFB) at 3 levels and two energy-based features, norm-1 L1 and norm-2 L2, are calculated from the wavelet coefficients at the first level and contourlet coefficients of each high frequency subband at different levels and directions to train and test SVM. When the nonwoven images are decomposed at 3 levels and 16 L1s are extracted, with 500 samples to train the SVM, the average recognition accuracy of test set is 98.4%, which is superior to the comparative method using wavelet texture analysis.

Selecting training points for one-class support vector machines

This paper proposes a training points selection method for one-class support vector machines. It exploits the feature of a trained one-class SVM, which uses points only residing on the exterior region of data distribution as support vectors. Thus, the proposed training set reduction method selects the so-called extreme points which sit on the boundary of data distribution, through local geometry and k-nearest neighbours. Experimental results demonstrate that the proposed method can reduce training set considerably, while the obtained model maintains generalization capability to the level of a model trained on the full training set, but uses less support vectors and exhibits faster training speed.

Error handling approach using characterization and correction steps for handwritten document analysis

In this paper, we present a framework to handle recognition errors from a N-best list of output phrases given by a handwriting recognition system, with the aim to use the resulting phrases as inputs to a higher-level application. The framework can be decomposed into four main steps: phrase alignment, detection, characterization, and correction of word error hypotheses. First, the N-best phrases are aligned to the top-list phrase, and word posterior probabilities are computed and used as confidence indices to detect word error hypotheses on this top-list phrase (in comparison with a learned threshold). Then, the errors are characterized into predefined types, using the word posterior probabilities of the top-list phrase and other features to feed a trained SVM. Finally, the final output phrase is retrieved, thanks to a correction step that used the characterized error hypotheses and a designed word-to-class backoff language model. First experiments were conducted on the ImadocSen-OnDB handwritten sentence database and on the IAM-OnDB handwritten text database, using two recognizers. We present first results on an implementation of the proposed framework for handling recognition errors on transcripts of handwritten phrases provided by recognition systems.