Robust visual tracking based on online learning sparse representation

Handling appearance variations is a very challenging problem for visual tracking. Existing methods usually solve this problem by relying on an effective appearance model with two features: (1) being capable of discriminating the tracked target from its background, (2) being robust to the target's appearance variations during tracking. Instead of integrating the two requirements into the appearance model, in this paper, we propose a tracking method that deals with these problems separately based on sparse representation in a particle filter framework. Each target candidate defined by a particle is linearly represented by the target and background templates with an additive representation error. Discriminating the target from its background is achieved by activating the target templates or the background templates in the linear system in a competitive manner. The target's appearance variations are directly modeled as the representation error. An online algorithm is used to learn the basis functions that sparsely span the representation error. The linear system is solved via ℓ1 minimization. The candidate with the smallest reconstruction error using the target templates is selected as the tracking result. We test the proposed approach using four sequences with heavy occlusions, large pose variations, drastic illumination changes and low foreground-background contrast. The proposed approach shows excellent performance in comparison with two latest state-of-the-art trackers.

A sequential algorithm for sparse support vector classifiers

Support vector machines (SVMs), though accurate, are not preferred in applications requiring high classification speed or when deployed in systems of limited computational resources, due to the large number of support vectors involved in the model. To overcome this problem we have devised a primal SVM method with the following properties: (1) it solves for the SVM representation without the need to invoke the representer theorem, (2) forward and backward selections are combined to approach the final globally optimal solution, and (3) a criterion is introduced for identification of support vectors leading to a much reduced support vector set. In addition to introducing this method the paper analyzes the complexity of the algorithm and presents test results on three public benchmark problems and a human activity recognition application. These applications demonstrate the effectiveness and efficiency of the proposed algorithm.


--------------------------------------------------------------------------------

A wavelet-based damage detection algorithm based on bridge acceleration response to a vehicle

Previous research based on theoretical simulations has shown the potential of the wavelet transform to detect damage in a beam by analysing the time-deflection response due to a constant moving load. However, its application to identify damage from the response of a bridge to a vehicle raises a number of questions. Firstly, it may be difficult to record the difference in the deflection signal between a healthy and a slightly damaged structure to the required level of accuracy and high scanning frequencies in the field. Secondly, the bridge is going to have a road profile and it will be loaded by a sprung vehicle and time-varying forces rather than a constant load. Therefore, an algorithm based on a plot of wavelet coefficients versus time to detect damage (a singularity in the plot) appears to be very sensitive to noise. This paper addresses these questions by: (a) using the acceleration signal, instead of the deflection signal, (b) employing a vehicle-bridge finite element interaction model, and (c) developing a novel wavelet-based approach using wavelet energy content at each bridge section which proves to be more sensitive to damage than a wavelet coefficient line plot at a given scale as employed by others.

FAULT-TOLERANT ALGORITHM FOR HIGH-AVAILABILITY AND WAFER SCALE SYSTEMS.

In this paper, the authors have presented one approach to configuring a Wafer-Scale Integration Chip. The approach described is called the 'WINNER', in which bus channels and an external controller for configuring the working processors are not required. In addition, the technique is applicable to high availability systems constructed using conventional methods. The technique can also be extended to arrays of arbitrary size and with any degree of fault tolerance simply by using an appropriate number of cells.

Bit-level systolic array implementation of the Winograd Fourier transform algorithm

A bit level systolic array system is proposed for the Winograd Fourier transform algorithm. The design uses bit-serial arithmetic and, in common with other systolic arrays, features nearest-neighbor interconnections, regularity and high throughput. The short interconnections in this method contrast favorably with the long interconnections between butterflies required in the FFT. The structure is well suited to VLSI implementations. It is demonstrated how long transforms can be implemented with components designed to perform a short length transform. These components build into longer transforms preserving the regularity and structure of the short length transform design.

SYSTOLIC IMPLEMENTATION OF THE WINOGRAD FOURIER TRANSFORM ALGORITHM.

A bit-level systolic array system is proposed for the Winograd Fourier transform algorithm. The design uses bit-serial arithmetic and, in common with other systolic arrays, features nearest neighbor interconnections, regularity, and high throughput. The short interconnections in this method contrast favorably with the long interconnections between butterflies required in the FFT. The structure is well suited to VLSI implementations. It is demonstrated how long transforms can be implemented with components designed to perform short-length transforms. These components build into longer transforms, preserving the regularity and structure of the short-length transform design.

QRD and SVD processor design based on an approximate rotations algorithm

A silicon implementation of the Approximate Rotations algorithm capable of carrying the computational load of algorithms such as QRD and SVD, within the real-time realisation of applications such as Adaptive Beamforming, is described. A modification to the original Approximate Rotations algorithm to simplify the method of optimal angle selection is proposed. Analysis shows that fewer iterations of the Approximate Rotations algorithm are required compared with the conventional CORDIC algorithm to achieve similar degrees of accuracy. The silicon design studies undertaken provide direct practical evidence of superior performance with the Approximate Rotations algorithm, requiring approximately 40% of the total computation time of the conventional CORDIC algorithm, for a similar silicon area cost. © 2004 IEEE.

Improved Montgomery modular inverse algorithm

A new, single and unified Montgomery modular inverse algorithm, which performs both classical and Montgomery modular inversion, is proposed. This reduces the number of Montgomery multiplication operations required by 33% when compared with previous algorithms reported in the literature. The use of this in practice has been investigated by implementation of the improved unified algorithm and the previous algorithms on FPGA devices. The unified algorithm implementation shows a significant speed-up and a reduction in silicon area usage.

Architectural synthesis of an image processing algorithm using IRIS

Details are presented of the IRIS synthesis system for high-performance digital signal processing. This tool allows non-specialists to automatically derive VLSI circuit architectures from high-level, algorithmic representations, and provides a quick route to silicon implementation. The applicability of the system is demonstrated using the design example of a one-dimensional Discrete Cosine Transform circuit.

Constant scale factor, on-line CORDIC algorithm in the circular coordinate system

Real time digital signal processing requires the development of high performance arithmetic algorithms suitable for VLSI design. In this paper, a new online, circular coordinate system CORDIC algorithm is described, which has a constant scale factor. This algorithm was developed using a new Angular Representation (AR) model A radix 2 version of the CORDIC algorithm is presented, along with an architecture suitable for VLSI implementation.

A biofuel strategy for Ireland with an emphasis on production of biomethane and minimization of land-take

Increasing energy consumption has exerted great pressure on natural resources; this has led to a move towards sustainable energy resources to improve security of supply and to reduce greenhouse gas emissions. However, the rush to the cure may have been made in haste. Biofuels in particular, have a bad press both in terms of competition with good agricultural land for food, and also in terms of the associated energy balance with the whole life cycle analysis of the biofuel system. The emphasis is now very much on sustainable biofuel production; biofuels from wastes and lignocellulosic material are now seen as good sustainable biofuels that affect significantly better greenhouse gas balances as compared with first generation biofuels. Ireland has a significant resource of organic waste that could be a potential source of energy through anaerobic digestion. Ireland has 8% of the cattle population of the EU with less than 1% of the human population; as a result 91% of agricultural land in Ireland is under grass. Residues such as slurries and slaughter waste together with energy crops such as grass have an excellent potential to produce biogas that may be upgraded to biomethane. This biomethane may be used as a natural gas substitute; bio-compressed natural gas may then be an avenue for a biofuel strategy. It is estimated that a maximum potential of 33% of natural gas may be substituted by 2020 with a practical obtainable level of 7.5% estimated. Together with biodiesel from residues the practical obtainable level of this strategy may effect greater than a 5% substitution by energy of transport. The residues considered in this strategy to produce biofuel (excluding grass) have the potential to save 93,000 ha of agricultural land (23% of Irish arable land) when compared to a rapeseed biodiesel strategy. © 2009 Elsevier Ltd. All rights reserved.

Scaled Heavy-Ball Acceleration of the Richardson-Lucy Algorithm

The Richardson-Lucy algorithm is one of the most important algorithms in the image deconvolution area. However, one of its drawbacks is slow convergence. A very significant acceleration is obtained by the technique proposed by Biggs and Andrews (BA), which is implemented in the deconvlucy function of the Image Processing MATLAB toolbox. The BA method was developed heuristically with no proof of convergence. In this paper, we introduce the Heavy-Ball (H-B) method for Poisson data optimization and extend it to a scaled H-B method, which includes the BA method as a special case. The method has proof of the convergence rate of O(k-2), where k is the number of iterations. We demonstrate the superior convergence performance of the scaled H-B method on both synthetic and real 3D images.

A self-learning algorithm for biased molecular dynamics

A new self-learning algorithm for accelerated dynamics, reconnaissance metadynamics, is proposed that is able to work with a very large number of collective coordinates. Acceleration of the dynamics is achieved by constructing a bias potential in terms of a patchwork of one-dimensional, locally valid collective coordinates. These collective coordinates are obtained from trajectory analyses so that they adapt to any new features encountered during the simulation. We show how this methodology can be used to enhance sampling in real chemical systems citing examples both from the physics of clusters and from the biological sciences.

Sensitivity and specificity of frequency-doubling technology, tendency-oriented perimetry, and Humphrey Swedish interactive threshold algorithm-fast perimetry in a glaucoma practice

PURPOSE: To evaluate the sensitivity and specificity of the screening mode of the Humphrey-Welch Allyn frequency-doubling technology (FDT), Octopus tendency-oriented perimetry (TOP), and the Humphrey Swedish Interactive Threshold Algorithm (SITA)-fast (HSF) in patients with glaucoma. DESIGN: A comparative consecutive case series. METHODS: This was a prospective study which took place in the glaucoma unit of an academic department of ophthalmology. One eye of 70 consecutive glaucoma patients and 28 age-matched normal subjects was studied. Eyes were examined with the program C-20 of FDT, G1-TOP, and 24-2 HSF in one visit and in random order. The gold standard for glaucoma was presence of a typical glaucomatous optic disk appearance on stereoscopic examination, which was judged by a glaucoma expert. The sensitivity and specificity, positive and negative predictive value, and receiver operating characteristic (ROC) curves of two algorithms for the FDT screening test, two algorithms for TOP, and three algorithms for HSF, as defined before the start of this study, were evaluated. The time required for each test was also analyzed. RESULTS: Values for area under the ROC curve ranged from 82.5%-93.9%. The largest area (93.9%) under the ROC curve was obtained with the FDT criteria, defining abnormality as presence of at least one abnormal location. Mean test time was 1.08 ± 0.28 minutes, 2.31 ± 0.28 minutes, and 4.14 ± 0.57 minutes for the FDT, TOP, and HSF, respectively. The difference in testing time was statistically significant (P <.0001). CONCLUSIONS: The C-20 FDT, G1-TOP, and 24-2 HSF appear to be useful tools to diagnose glaucoma. The test C-20 FDT and G1-TOP take approximately 1/4 and 1/2 of the time taken by 24 to 2 HSF. © 2002 by Elsevier Science Inc. All rights reserved.

Association Study of 167 Candidate Genes for Schizophrenia Selected by a Multi-Domain Evidence-Based Prioritization Algorithm and Neurodevelopmental Hypothesis

Integrating evidence from multiple domains is useful in prioritizing disease candidate genes for subsequent testing. We ranked all known human genes (n = 3819) under linkage peaks in the Irish Study of High-Density Schizophrenia Families using three different evidence domains: 1) a meta-analysis of microarray gene expression results using the Stanley Brain collection, 2) a schizophrenia protein-protein interaction network, and 3) a systematic literature search. Each gene was assigned a domain-specific p-value and ranked after evaluating the evidence within each domain. For comparison to this
ranking process, a large-scale candidate gene hypothesis was also tested by including genes with Gene Ontology terms related to neurodevelopment. Subsequently, genotypes of 3725 SNPs in 167 genes from a custom Illumina iSelect array were used to evaluate the top ranked vs. hypothesis selected genes. Seventy-three genes were both highly ranked and involved in neurodevelopment (category 1) while 42 and 52 genes were exclusive to neurodevelopment (category 2) or highly ranked (category 3), respectively. The most significant associations were observed in genes PRKG1, PRKCE, and CNTN4 but no individual SNPs were significant after correction for multiple testing. Comparison of the approaches showed an excess of significant tests using the hypothesis-driven neurodevelopment category. Random selection of similar sized genes from two independent genome-wide association studies (GWAS) of schizophrenia showed the excess was unlikely by chance. In a further meta-analysis of three GWAS datasets, four candidate SNPs reached nominal significance. Although gene ranking using integrated sources of prior information did not enrich for significant results in the current experiment, gene selection using an a priori hypothesis (neurodevelopment) was superior to random selection. As such, further development of gene ranking strategies using more carefully selected sources of information is warranted.