Objective pilling evaluation of wool fabrics

An objective pilling evaluation method based on the multi-scale two-dimensional dual-tree complex wavelet transform and linear discriminant function of Bayes' Rule was developed. The surface fuzz and pills are identified from the high-frequency noise, fabric textures, fabric surface unevenness, and illuminative variation of a pilled fabric image by the two-dimensional dual-tree complex wavelet decomposition and reconstruction. The energies of the reconstructed subimages in six spatial orientations (±15º, ±45º, ±75º) are calculated as the elements of the pilling feature vector, whose dimension is reduced by principal component analysis. A linear discriminant function of Bayes' Rule was used as a classifier to establish classification rules among the five pilling grades. A new pilled sample with the same physical construction can then be automatically assigned to one of the five pilling grades by the classification rules. A general evaluation of the proposed method was conducted using the SM50 woven, non-woven, and SM54 knitted standard pilling test image sets. The results suggest that the new method can successfully establish classification rules among the five pilling grade groups for each of the three standard pilling test image sets and should be applicable to practical objective pilling evaluation.

Objective grading of fabric pilling with wavelet texture analysis

A new objective fabric pilling grading method based on wavelet texture analysis was developed. The new method created a complex texture feature vector based on the wavelet detail coefficients from all decomposition levels and horizontal, vertical and diagonal orientations, permitting a much richer and more complete representation of pilling texture in the image to be used as a basis for classification. Standard multi-factor classification techniques of principal components analysis and discriminant analysis were then used to classify the pilling samples into five pilling degrees. The preliminary investigation of the method was performed using standard pilling image sets of knitted, woven and non-woven fabrics. The results showed that this method could successfully evaluate the pilling intensity of knitted, woven and non-woven fabrics by selecting the suitable wavelet and associated analysis scale.

Identification of moving obstacles with pyramidal Lucas Kanade optical flow and k means clustering

This paper describes the methodology for identifying moving obstacles by obtaining a reliable and a sparse optical flow from image sequences. Given a sequence of images, basically we can detect two-types of on road vehicles, vehicles traveling in the opposite direction and vehicles traveling in the same direction. For both types, distinct feature points can be detected by Shi and Tomasi corner detector algorithm. Then pyramidal Lucas Kanade method for optical flow calculation is used to match the sparse feature set of one frame on the consecutive frame. By applying k means clustering on four component feature vector, which are to be the coordinates of the feature point and the two components of the optical flow, we can easily calculate the centroids of the clusters and the objects can be easily tracked. The vehicles traveling in the opposite direction produce a diverging vector field, while vehicles traveling in the same direction produce a converging vector field

The robustness of objective fabric pilling evaluation method

Previously, we proposed a new method to identify fabric pilling and objectively measure fabric pilling intensity based on the two-dimensional dual-tree complex wavelet reconstruction and neural network classification. Here we further evaluate the robustness of the method. Our results indicate that the pilling identification method is robust to significant variation in the brightness and contrast of the image, rotation of the image, and 2 i (i is an integer) times dilation of the image. The pilling feature vector developed to characterize the pilling intensity is robust to brightness change but is sensitive to large rotations of the image. As long as all fabric images are adjusted to have the same contrast level and the sample is illuminated from the same direction, the pilling feature vectors are comparable and can be used to classify the pilling intensity.

New methods for objective evaluation of fabric pilling by frequency domain image processing

Fabric pilling is a serious problem for the apparel industry. Resistance to pilling is normally tested by simulated accelerated wear and manual assessment of degree of pilling based on a visual comparison of the sample to a set of test images. A number of automated systems based on image analysis have been developed. The authors propose new methods of image analysis based on the two-dimensional wavelet transform to objectively measure the pilling intensity in sample images. Initial work employed the detail coefficients of the two-dimensional discrete wavelet transform (2DDWT) as a measure of the pilling intensity of woven/knitted fabrics.

This method is shown to be robust to image translation and brightness variation. Using the approximation coefficients of the 2DDWT, the method is extended to non-woven pilling image sets. Wavelet texture analysis (WTA) combined with principal components analysis are shown to produce a richer texture description of pilling for analysis and classification. Finally, employing the two-dimensional dual-tree complex wavelet transform as the basis for the WTA feature vector is shown to produce good automated classification on a range of standard pilling image sets.

Internet traffic classification using machine learning : a token-based approach

Due to the increasing unreliability of traditional port-based methods, Internet traffic classification has attracted a lot of research efforts in recent years. Quite a lot of previous papers have focused on using statistical characteristics as discriminators and applying machine learning techniques to classify the traffic flows. In this paper, we propose a novel machine learning based approach where the features are extracted from packet payload instead of flow statistics. Specifically, every flow is represented by a feature vector, in which each item indicates the occurrence of a particular token, i.e.; a common substring, in the payload. We have applied various machine learning algorithms to evaluate the idea and used different feature selection schemes to identify the critical tokens. Experimental result based on a real-world traffic data set shows that the approach can achieve high accuracy with low overhead.

Malware variant detection using similarity search over sets of control flow graphs

Static detection of polymorphic malware variants plays an important role to improve system security. Control flow has shown to be an effective characteristic that represents polymorphic malware instances. In our research, we propose a similarity search of malware using novel distance metrics of malware signatures. We describe a malware signature by the set of control flow graphs the malware contains. We propose two approaches and use the first to perform pre-filtering. Firstly, we use a distance metric based on the distance between feature vectors. The feature vector is a decomposition of the set of graphs into either fixed size k-sub graphs, or q-gram strings of the high-level source after decompilation. We also propose a more effective but less computationally efficient distance metric based on the minimum matching distance. The minimum matching distance uses the string edit distances between programs' decompiled flow graphs, and the linear sum assignment problem to construct a minimum sum weight matching between two sets of graphs. We implement the distance metrics in a complete malware variant detection system. The evaluation shows that our approach is highly effective in terms of a limited false positive rate and our system detects more malware variants when compared to the detection rates of other algorithms.

Objective surface evaluation of fiber reinforced polymer composites

The mechanical properties of advanced composites are essential for their structural performance, but the surface finish on exterior composite panels is of critical importance for customer satisfaction. This paper describes the application of wavelet texture analysis (WTA) to the task of automatically classifying the surface finish properties of two fiber reinforced polymer (FRP) composite construction types (clear resin and gel-coat) into three quality grades. Samples were imaged and wavelet multi-scale decomposition was used to create a visual texture representation of the sample, capturing image features at different scales and orientations. Principal components analysis was used to reduce the dimensionality of the texture feature vector, permitting successful classification of the samples using only the first principal component. This work extends and further validates the feasibility of this approach as the basis for automated non-contact classification of composite surface finish using image analysis.

Crowd detection from still images

The analysis of human crowds has widespread uses from law enforcement to urban engineering and traffic management. All of these require a crowd to first be detected, which is the problem addressed in this paper. Given an image, the algorithm we propose segments it into crowd and non-crowd regions. The main idea is to capture two key properties of crowds: (i) on a narrow scale, its basic element should look like a human (only weakly so, due to low resolution, occlusion, clothing variation etc.), while (ii) on a larger scale, a crowd inherently contains repetitive appearance elements. Our method exploits this by building a pyramid of sliding windows and quantifying how “crowd-like” each level of the pyramid is using an underlying statistical model based on quantized SIFT features. The two aforementioned crowd properties are captured by the resulting feature vector of window responses, describing the degree of crowd-like appearance around an image location as the surrounding spatial extent is increased.

Recognising user identity in twitter social networks via text mining

Social networks have become a convenient and effective means of communication in recent years. Many people use social networks to communicate, lead, and manage activities, and express their opinions in supporting or opposing different causes. This has brought forward the issue of verifying the owners of social accounts, in order to eliminate the effect of any fake accounts on the people. This study aims to authenticate the genuine accounts versus fake account using writeprint, which is the writing style biometric. We first extract a set of features using text mining techniques. Then, training of a supervised machine learning algorithm to build the knowledge base is conducted. The recognition procedure starts by extracting the relevant features and then measuring the similarity of the feature vector with respect to all feature vectors in the knowledge base. Then, the most similar vector is identified as the verified account.

Control flow-based malware variant detection

Static detection of malware variants plays an important role in system security and control flow has been shown as an effective characteristic that represents polymorphic malware. In our research, we propose a similarity search of malware to detect these variants using novel distance metrics. We describe a malware signature by the set of control flowgraphs the malware contains. We use a distance metric based on the distance between feature vectors of string-based signatures. The feature vector is a decomposition of the set of graphs into either fixed size k-subgraphs, or q-gram strings of the high-level source after decompilation. We use this distance metric to perform pre-filtering. We also propose a more effective but less computationally efficient distance metric based on the minimum matching distance. The minimum matching distance uses the string edit distances between programs' decompiled flowgraphs, and the linear sum assignment problem to construct a minimum sum weight matching between two sets of graphs. We implement the distance metrics in a complete malware variant detection system. The evaluation shows that our approach is highly effective in terms of a limited false positive rate and our system detects more malware variants when compared to the detection rates of other algorithms. © 2013 IEEE.

Condition assessment of timber utility poles based on a hierarchical data fusion model

This paper proposes a novel hierarchical data fusion technique for the non-destructive testing (NDT) and condition assessment of timber utility poles. The new method analyzes stress wave data from multisensor and multiexcitation guided wave testing using a hierarchical data fusion model consisting of feature extraction, data compression, pattern recognition, and decision fusion algorithms. The researchers validate the proposed technique using guided wave tests of a sample of in situ timber poles. The actual health states of these poles are known from autopsies conducted after the testing, forming a ground-truth for supervised classification. In the proposed method, a data fusion level extracts the main features from the sampled stress wave signals using power spectrum density (PSD) estimation, wavelet packet transform (WPT), and empirical mode decomposition (EMD). These features are then compiled to a feature vector via real-number encoding and sent to the next level for further processing. Principal component analysis (PCA) is also adopted for feature compression and to minimize information redundancy and noise interference. In the feature fusion level, two classifiers based on support vector machine (SVM) are applied to sensor separated data of the two excitation types and the pole condition is identified. In the decision making fusion level, the Dempster–Shafer (D-S) evidence theory is employed to integrate the results from the individual sensors obtaining a final decision. The results of the in situ timber pole testing show that the proposed hierarchical data fusion model was able to distinguish between healthy and faulty poles, demonstrating the effectiveness of the new method.

Dynamic feature selection for spam filtering using support vector machine

Spam is commonly defined as unsolicited email messages and the goal of spam filtering is to differentiate spam from legitimate email. Much work have been done to filter spam from legitimate emails using machine learning algorithm and substantial performance has been achieved with some amount of false positive (FP) tradeoffs. In this paper, architecture of spam filtering has been proposed based on support vector machine (SVM,) which will get better accuracy by reducing FP problems. In this architecture an innovative technique for feature selection called dynamic feature selection (DFS) has been proposed which is enhanced the overall performance of the architecture with reduction of FP problems. The experimental result shows that the proposed technique gives better performance compare to similar existing techniques.

Stable feature selection with support vector machines

The support vector machine (SVM) is a popular method for classification, well known for finding the maximum-margin hyperplane. Combining SVM with l1-norm penalty further enables it to simultaneously perform feature selection and margin maximization within a single framework. However, l1-norm SVM shows instability in selecting features in presence of correlated features. We propose a new method to increase the stability of l1-norm SVM by encouraging similarities between feature weights based on feature correlations, which is captured via a feature covariance matrix. Our proposed method can capture both positive and negative correlations between features. We formulate the model as a convex optimization problem and propose a solution based on alternating minimization. Using both synthetic and real-world datasets, we show that our model achieves better stability and classification accuracy compared to several state-of-the-art regularized classification methods.