Mohamed F. Tolba

Fundamental matrix estimation: A study of error criteria

The fundamental matrix (FM) describes the geometric relations that exist between two images of the same scene. Different error criteria are used for estimating FMs from an input set of correspondences. In this paper, the accuracy and efficiency aspects of the different error criteria are studied. We mathematically and experimentally proved that the most popular error criterion, the symmetric epipolar distance, is biased. It was also shown that despite the similarity between the algebraic expressions of the symmetric epipolar distance and Sampson distance, they have different accuracy properties. In addition, a new error criterion, Kanatani distance, was proposed and proved to be the most effective for use during the outlier removal phase from accuracy and efficiency perspectives. To thoroughly test the accuracy of the different error criteria, we proposed a randomized algorithm for Reprojection Error-based Correspondence Generation (RE-CG). As input, RE-CG takes an FM and a desired reprojection error value d. As output, RE-CG generates a random correspondence having that error value. Mathematical analysis of this algorithm revealed that the success probability for any given trial is 1-(2/3)^2 at best and is 1-(6/7)^2 at worst while experiments demonstrated that the algorithm often succeeds after only one trial.

Fiducial feature reduction analysis for electrocardiogram (ECG) based biometric recognition

Although the electrocardiogram (ECG) has been a reliable diagnostic tool for decades, its deployment in the context of biometrics is relatively recent. Its robustness to falsification, the evidence it carries about aliveness and its rich feature space has rendered the deployment of ECG based biometrics an interesting prospect. The rich feature space contains fiducial based information such as characteristic peaks which reflect the underlying physiological properties of the heart. The principal goal of this study is to quantitatively evaluate the information content of the fiducial based feature set in terms of their effect on subject and heart beat classification accuracy (ECG data acquired from the PhysioNet ECG repository). To this end, a comprehensive set of fiducial based features was extracted from a collection of ECG records. This feature set was subsequently reduced using a variety of feature extraction/selection methods such as principle component analysis (PCA), linear discriminant analysis (LDA), information-gain ratio (IGR), and rough sets (in conjunction with the PASH algorithm). The performance of the reduced feature set was examined and the results evaluated with respect to the full feature set in terms of the overall classification accuracy and false (acceptance/rejection) ratios (FAR/FRR). The results of this study indicate that the PASH algorithm, deployed within the context of rough sets, reduced the dimensionality of the feature space maximally, while maintaining maximal classification accuracy.

Arabic sign language continuous sentences recognition using PCNN and graph matching

Many previous researchers have tried developing sign languages recognition systems in general and Arabic sign language specifically. They succeeded to achieve acceptable results for isolated gestures level, but none of them investigated the recognition of connected sequence of gestures. This paper focuses on how to recognize real-time connected sequence of gestures using graph-matching technique, also how the continuous input gestures are segmented and classified. Graphs are a general and powerful data structure useful for the representation of various objects and concepts. This work is a component of a real-time Arabic Sign Language Recognition system that applied pulse-coupled neural network for static posture recognition in its first phase. This work can be adapted and applied to different sign languages and other recognition problems.

A proposed PCNN features quality optimization technique for pose-invariant 3D Arabic sign language recognition

This paper proposes a novel technique to deal with pose variations in 3D object recognition. This technique uses pulse-coupled neural network (PCNN) for image features generation from two different viewing angles. These signatures qualities are then evaluated, using a proposed fitness function. The features evaluation step is taken before any classification steps are performed. The evaluation results dynamic weighting factors for each camera express the features quality from the current viewing angles. The proposed technique uses the two 2D image features and their corresponding calculated weighting factors to construct optimized quality 3D features. An experiment was conducted in Arabic sign language recognition application which multiple views are necessary to distinguish some signs. The proposed technique obtained a 96% recognition accuracy for pose-invariant restrictions with a degree of freedom from 0 to 90.

A wavelet feature extraction method for electrocardiogram (ECG)-based biometric recognition

This paper proposes a discrete wavelet feature extraction method for an electrocardiogram (ECG)-based biometric system. In this method, the RR intervals are extracted and decomposed using discrete biorthogonal wavelet in wavelet coefficient structures. These structures are reduced by excluding the non-informative coefficients, and then, they are fed into a radial basis functions (RBF) neural network for classification. Moreover, the ability of using only the QT or QRS intervals instead of the RR intervals is also investigated. Finally, the results achieved by our method outperformed the auto-correlation (AC)/discrete cosine transform (DCT) method where the DCT coefficients are derived from the AC of ECG segments and fed into the RBF network for classification. The conducted experiments were validated using four Physionet databases. Critical issues like stability overtime, the ability to reject impostors, scalability and generalization to other datasets have also been addressed.

Visualization of large time-varying vector data

In computation flow visualization, integration based geometric flow visualization is often used to explore the flow field structure. A typical time-varying dataset from a Computational Fluid Dynamics (CFD) simulation can easily require hundreds of gigabytes to even terabytes of storage space, which creates challenges for the consequent data-analysis tasks. This paper presents a new technique for path-lines visualization of extremely large time varying vector data using high performance computing. The high level requirements that guided the formulation of the new technique are (a) support for large dataset sizes, (b) support for temporal coherence of the vector data, (c) support for distributed memory high performance computing and (d) optimum utilization of the computing nodes with multi-cores (multi-core processors). The challenge is to design and implement a technique that meets these complex requirements and balances the conflicts between them. The fundamental innovation in this work is developing efficient distributed path-lines visualization for large time varying vector data. The maximum performance was reached through the parallelization of multiple processes on the multi-cores of each computing node. Accuracy of the proposed technique was confirmed compared to the results of the Visualization Tool Kit (VTK). In addition, the proposed technique exhibited acceptable scalability for different data sizes with better scalability for the larger ones. Finally, the utilization of the computing nodes was satisfactory for the considered test cases.

Cattle Identification Using Muzzle Print Images Based on Texture Features Approach

The increasing growth of the world trade and growing concerns of food safety by consumers need a cutting-edge animal identification and traceability systems as the simple recording and reading of tags-based systems are only effective in eradication programs of national disease. Animal biometric-based solutions, e.g. muzzle imaging system, offer an effective and secure, and rapid method of addressing the requirements of animal identification and traceability systems. In this paper, we propose a robust and fast cattle identification approach. This approach makes use of Local Binary Pattern (LBP) to extract local invariant features from muzzle print images. We also applied different classifiers including Nearest Neighbor, Naive Bayes, SVM and KNN for cattle identification. The experimental results showed that our approach is superior than existed works as ours achieves 99,5% identification accuracy. In addition, the results proved that our proposed method achieved this high accuracy even if the testing images are rotated in various angels or occluded with different parts of their sizes.

An enhanced test case generation technique based on activity diagrams

Test case generation is a core phase in any testing process, therefore automating it plays a tremendous role in reducing the time and effort spent during the testing process. This paper proposes an enhanced XML-based automated approach for generating test cases from activity diagrams. The proposed architecture creates a special table called Activity Dependency Table (ADT) for each XML file. The ADT covers all the functionalities in the activity diagram as well as handling the decisions, loops, fork, join, merge, object and conditional threads. Then it automatically generates a directed graph called Activity Dependency Graph (ADG) that is used in conjunction with the ADT to extract all the possible final test cases. The proposed model validates the generated test paths during the generation process to ensure meeting a hybrid coverage criterion. The generated test cases can be sent to any requirements management tool to be traced against the requirements. The proposed model is prototyped on 30 differently sized activity diagrams in different domains An experimental evaluation of the proposed model is done as well. It saves time and effort besides, increases the quality of generated test cases, therefore optimizes the overall performance of the testing process Moreover, the generated test cases can be executed on the system under test using any automatic test execution tool.

Neutrosophic Sets and Fuzzy C-Means Clustering for Improving CT Liver Image Segmentation

In this paper, an improved segmentation approach based on Neutrosophic sets (NS) and fuzzy c-mean clustering (FCM) is proposed. An application of abdominal CT imaging has been chosen and segmentation approach has been applied to see their ability and accuracy to segment abdominal CT images. The abdominal CT image is transformed into NS domain, which is described using three subsets namely; the percentage of truth in a subset T, the percentage of indeterminacy in a subset I, and the percentage of falsity in a subset F. The entropy in NS is defined and employed to evaluate the indeterminacy. Threshold for NS image is adapted using Fuzzy C-mean algorithm. Finally, abdominal CT image is segmented and liver parenchyma is selected using connected component algorithm. The proposed approach denoted as NSFCM and compared with FCM using Jaccard Index and Dice Coefficient. The experimental results demonstrate that the proposed approach is less sensitive to noise and performs better on nonuniform CT images.

Evaluation of unsupervised feature extraction neural networks for face recognition

In this paper, new appearances based on neural networks (NN) algorithms are presented for face recognition. Face recognition is subdivided into two main stages: feature extraction and classifier. The suggested NN algorithms are the unsupervised Sanger principal component neural network (Sanger PCNN) and the self-organizing feature map (SOFM), which will be applied for features extraction of the frontal view of a face image. It is of interest to compare the unsupervised network with the traditional Eigenfaces technique. This paper presents an experimental comparison of the statistical Eigenfaces method for feature extraction and the unsupervised neural networks in order to evaluate the classification accuracies as comparison criteria. The classifier is done by the multilayer perceptron (MLP) neural network. Overcoming of the problem of the finite number of training samples per person is discussed. Experimental results are implemented on the Olivetti Research Laboratory database that contains variability in expression, pose, and facial details. The results show that the proposed method SOFM/MLP neural network is more efficient and robust than the Sanger PCNN/MLP and the Eigenfaces/MLP, when used a few number of training samples per person. As a result, it would be more applicable to utilize the SOFM/MLP NN in order to accomplish a higher level of accuracy within a recognition system.