Intelligent trademark recognition and similarity analysis using a two-stage transfer learning approach

The ubiquitous and fast-growing e-marketplaces are causing serious concerns of unauthorized trademark usages, especially their pictorial images. There is a significant need for e-commerce service intermediaries and trading sites to actively check whether product images posted online are not infringing upon others’ intellectual property rights (IPRs). To ensure registered trademark (TM) or copyrighted logo-image (logo) protection, this research develops an intelligent system that can detect, locate, and crop (cut) logos posted online and check whether they are substantially or deceptively similar to registered TM logos. This research integrates two deep learning models to achieve the research goal. The first is the logo detection and localization model for cropping trademark like images from complex online merchandise photos, which could have images of many product views and even worn and displayed on a human model. Only the cropped logo image is needed for comparison to a database of registered trademarks. The second model performs TM similarity analysis using the cropped logos compared to the published TM images. Yolo v4 is adopted as the general logo locator for intelligent logo image cropping. The triplet convolutional neural network model is used to fine-tuned for intelligent trademark similarity analysis. The models are trained with an image dataset, combining image samples from LogoDet-3k and images found through web search. The system performance is quantitatively tested using the FlickrLogos-32 dataset and consumer product images extracted from an e-commerce platform. The test results of the system achieve high precision (0.91) for trademark class matching.     

An eigenspace and residual approach to transfer function matrix synthesis

Based on results from eigenspace and residual assignment a closed-form parameterization of all transfer function matrices obtainable by state feedback is given.     

Shape sensitivity analysis using a fixed basis function finite element approach

An approach is presented for the determination of solution sensitivity to changes in problem domain or shape. A finite element displacement formulation is adopted and the point of view is taken that the finite element basis functions and grid are fixed during the sensitivity analysis; therefore, the method is referred to as a “fixed basis function” finite element shape sensitivity analysis. This approach avoids the requirement of explicit or approximate differentiation of finite element matrices and vectors and the difficulty or errors resulting from such calculations. Effectively, the sensitivity to boundary shape change is determined exactly; thus, the accuracy of the solution sensitivity is dictated only by the finite element mesh used. The evaluation of sensitivity matrices and force vectors requires only modest calculations beyond those of the reference problem finite element analysis; that is, certain boundary integrals and reaction forces on the reference location of the moving boundary are required. In addition, the formulation provides the unique family of element domain changes which completely eliminates the inclusion of grid sensitivity from the shape sensitivity calculation. The work is illustrated for some one-dimensional beam problems and is outlined for a two-dimensional C⁰ problem; the extension to three-dimensional problems is straight-forward. Received December 5, 1999?Revised mansucript received July 6, 2000     

A new approach to design safe-supported HDD against random excitation by using optimization of rubbers spatial parameters

Hard disk drives (HDDs) of laptop personal computers (LPCs) are devices vulnerable in harsh mechanical environments. Hence, they need to be protected against damages due to vibration in order to have better read/write performance. In the present study, a LPC and its HDD are modeled as a system with two degrees of freedom and the nonlinear optimization method is employed to perform a passive control through minimizing the root mean square of HDD absolute acceleration due to a base random excitation. The presented random excitation is considered as a stationary, zero mean process with Gaussian distribution. In addition, eleven inequality constraints are defined based on geometrical limitations and allowable intervals of lumped modal parameters. The target of the optimization is to obtain optimum modal parameters of rubber mounts and rubber feet as design variables and subsequently propose new characteristics of rubber mounts and rubber feet to be manufactured for HDD protection against random excitation. In this paper, a nonlinear optimization problem is separately solved for three widely-used cases of HDD by using modified constrained steepest descent algorithm (PLBA) which was extended based on sequential quadratic programming. Finally, the genetic algorithm is used to verify results of the PLBA algorithm.

     

A transfer function approach to the realisation problem of nonlinear systems

This article studies the nonlinear realisation problem, i.e. the problem of finding the state equations of a nonlinear system from the transfer function representation being easily computable from the higher order input–output differential equation. The realisation in both observer and controller canonical forms is studied. The results demonstrate a clear connection with those from linear theory. In the solution the concept of adjoint polynomials, adjoint transfer function and right factorisation of the transfer function play a key role. Finally, the results are applied for system linearisation up to input–output injection used in the observer design.     

The slope of a transfer function

Exact and approximate equations are derived that relate the magnitude slope of a general transfer function directly to its phase angle. The result is used to derive simple conditions on stabilizability of feedback loops with right half-plane poles and zeros or dead-time.     

HDD: a hypercube division-based algorithm for discretisation

Discretisation, as one of the basic data preparation techniques, has played an important role in data mining. This article introduces a new hypercube division-based (HDD) algorithm for supervised discretisation. The algorithm considers the distribution of both class and continuous attributes and the underlying correlation structure in the data set. It tries to find a minimal set of cut points, which divides the continuous attribute space into a finite number of hypercubes, and the objects within each hypercube belong to the same decision class. Finally, tests are performed on seven mix-mode data sets, and the C5.0 algorithm is used to generate classification rules from the discretised data. Compared with the other three well-known discretisation algorithms, the HDD algorithm can generate a better discretisation scheme, which improves the accuracy of classification and reduces the number of classification rules.     

A constrained extremum‐seeking control approach

This paper proposes an extremum‐seeking control approach for a class of uncertain nonlinear dynamical systems subject to constraints. It is assumed that the objective function along with the constraints are available for measurement. The contribution of the paper is to formulate the extremum‐seeking problem as a time‐varying estimation problem. The constraints are enforced by using an interior point strategy that preserves feasibility of the systems' trajectories. Simulation examples are used to illustrate the effectiveness of the proposed technique. Copyright © 2014 John Wiley & Sons, Ltd.     

Multivariable system design by transfer function synthesis

This paper is concerned with the problem of designing a constant output feedback matrix to synthesize the denominator and certain numerator polynomials of the transfer function matrix of a linear multivariable system. The problem is tackled by first using the special case of unity rank feedback and then removing this restriction to cover the case of unconstrained feedback.

Simple expressions relating closed-loop and open-loop transfer functions of multi-variable systems and unity rank feedback are first established. These are then used to develop a recursive synthesis algorithm employing unity rank output feedback. The algorithm uses the pseudo-inverse concept for obtaining least-squares solutions of sots of simultaneous linear equations. The extension from unity rank feedback to unrestricted rank feedback is carried out by considering the feedback matrix as a sum of unity rank matrices. Using this concept and the expressions derived earlier for unity rank feedback, a recursive algorithm is developed for calculating unrestricted rank output feedback matrices. Examples are given to illustrate the synthesis procedure using both unity rank and unrestricted rank feedback matrices.     

Estimation of pulse transfer function parameters by quasilinearization

An iterative algorithm for the identification of single-input single-output linear stationary discrete systems is developed using the method of quasilinearization. The resulting procedure is similar to mode 1 of the method of Steiglitz and McBride but has the advantage of the quadratic convergence property of quasilineariza-tion. It is shown that this algorithm becomes mode 1 if the measured plant output is used in the calculations in place of the model output. Consequently, the two methods are extremely compatible and it is a simple matter to combine them in a single program, which generates its own initial estimates has the wide range of convergence of mode 1, and possesses the quadratic convergence property of quasilineariza-tion for final convergence to a solution. The method also permits the estimation of plant initial conditions in those cases where they must be considered. Results of a few numerical applications are discussed.     

Understanding transfer pricing: a case-based reasoning approach

This paper introduces the domain of transfer pricing, followed by an overview of case-based reasoning. Then, case-based reasoning is used to analyse the results of a questionnaire survey in transfer pricing. Finally, some conclusions are drawn and opportunities for future research identified.     

Robust design of a microactuator for HDD head positioning

With rapid growth in areal recording density of hard disk drive (HDD), ultra-high precise head positioning and higher servo bandwidth are required. Dual-stage actuation system using MEMS actuator has been considered as one promising precise positioning solution. The microacatuator we designed has a low in-plane resonant frequency, which limits the servo bandwidth and needs compensation by servo control system. However, this resonant frequency may shift from its mean value during MEMS fabrication process. This shift adds difficulty to effective compensation for servo control. In this paper, a robust solution is proposed to minimize microactuator’s resonant frequency shift so that its first resonance frequency is insensitive to the effects of sources of processing variations and then effective compensation for high servo bandwidth can be achieved.     

Document Analysis by Crosscount Approach

In this paper a new feature called crosscount for document analysis is introduced.The reature crosscount is a function of white line segment with its start on the edge of document images.It reflects not only the contour of image,but also the periodicity of white lines(background)and text lines in the document images.In complex printed-page layouts,there are different blocks such as textual,graphical,tabular,and so on.Of these blocks,textual ones have the most obvious periodicity with their homogeneous white lines arranged regularly.The important property of textual blocks can be extracted by crosscount functions.here the document layouts are classified into three classes on the basis of their physical structures.Then the definition and properties of the crosscount function are described.According to the classification of document layouts,the application of this new feature to different types of document images‘ analysis and understanding is discussed.     

Stability analysis and control synthesis for switched systems: a switched Lyapunov function approach

This paper addresses the problem of stability analysis and control synthesis of switched systems in the discrete-time domain. The approach followed in this paper looks at the existence of a switched quadratic Lyapunov function to check asymptotic stability of the switched system under consideration. Two different linear matrix inequality-based conditions allow to check the existence of such a Lyapunov function. The first one is classical while the second is new and uses a slack variable, which makes it useful for design problems. These two conditions are proved to be equivalent for stability analysis. Investigating the static output feedback control problem, we show that the second condition is, in this case, less conservative. The reduction of the conservatism is illustrated by a numerical evaluation.     

Determination of a transfer function matrix in multivariablesystems

A conceptually simple method is presented for determining the system transfer function matrix of linear multivariable systems described by their state-space equations. The polynomial coefficients of the transfer function matrix are computed directly by using the controllability matrix method and the transformation to canonical (phase-variable) form matrices, even if the transformation matrices are singular. The method does not require matrix inversion and calculation of eigenvalues, and requires only O(n³) operations     

Numerical analysis of single-layered graphene sheets by a mesh-free approach

The paper presents a numerical analysis of the behavior of single-layered graphene sheet (SLGS) using a mesh-free approach based on a nonlocal continuum plate model (NCPM). The adopted NCPM constructed by incorporating the nonlocal elasticity theory into the first order shear deformation elastic plate theory is able to capture small length scale effects. Through the NCPM, the SLGS is modeled as a continuous orthotropic nanoplate. the obtained nonlocal nonlinear partial differential equations completed by boundary conditions are solved numerically by a mesh-free approach associating the asymptotic numerical method with the mesh-free collocation method based on moving least square approximation. The effects of the small-scale parameter and aspect ratio on the nonlinear bending and post-buckling behaviors of SLGS are considered. Good agreement has been established between the obtained results and those of the literature.

     

A more efficient approach for investigation of effect of various HDD components on the shock tolerance

This paper emphasizes the importance of a fast and reliable tool for design parametric studies. Three methods that have been used in studying the shock phenomena in hard disk drives (HDD) are discussed briefly. A more efficient (i.e. faster and reasonably accurate) method is proposed. This new method uses state-space formulation to model the structural components of the HDD and quasi-static concept to model the non-linearity of the air bearing. The structural and the air bearing response, and their coupled interaction are evaluated simultaneously. The computation time and the accuracy of the proposed method are compared against two existing methods: the full finite element model and CML’s HDD dynamics event simulator. The proposed method is found to be much faster with reasonably good accuracy. Experimental shock tests also show that the method is reliable in predicting the shock tolerance of the HDD. A procedure on how to do parametric study with this method is proposed and the effect of overmold and voice coil stiffness on the shock tolerance is studied.