Image classification without segmentation using a hybrid pyramid kernel

Image classification usually requires complicated segmentation to separate foreground objects from the background scene. However, the statistical content of a background scene can actually provide very useful information for classification. In this paper, we propose a new hybrid pyramid kernel which incorporates local features extracted from both dense regular grids and interest points for image classification, without requiring segmentation. Features extracted from dense regular grids can better capture information about the background scene, while interest points detected at corners and edges can better capture information about the salient objects. In our algorithm, these two local features are combined in both the spatial and the feature-space domains, and are organized into pyramid representations. In order to obtain better classification accuracy, we fine-tune the parameters involved in the similarity measure, and we determine discriminative regions by means of relevance feedback. From the experimental results, we observe that our algorithm can achieve a 6.37 % increase in performance as compared to other pyramid-representation-based methods. To evaluate the applicability of the proposed hybrid kernel to large-scale databases, we have performed a cross-dataset experiment and investigated the effect of foreground/background features on each of the kernels. In particular, the proposed hybrid kernel has been proven to satisfy Mercer’s condition and is efficient in measuring the similarity between image features. For instance, the computational complexity of the proposed hybrid kernel is proportional to the number of features.     

The analysis of solids without mesh generation using trimmed patch boundary elements

The tools available to the mechanical engineer—for example, geometric modeling and computer-aided analysis—are individually quite powerful, but they are based on different geometric representations. Hence, they do not always work well together. In this paper an analysis method is presented that operates directly on the geometric modeling representation. Therefore, the time-consuming and error-prone procedure of generating a mesh is skipped. The method is based on boundary integral equations, but unlike previously published methods, the boundary elements aren-sided trimmed patches, the same patches that are used by modern geometric modelers to represent complex solids. The method is made practical by defining shape functions over the trimmed patches in such a way that the number of degrees of freedom can be controlled. This is done by using a concept called virtual nodes. The paper begins by deriving the trimmed patch boundary element. Then its properties are discussed in comparison with existing boundary element and finite element methods, and several examples are given.     

Adaptive mesh refinement for shells with modified Ahmad elements

This paper presents a simple scheme for the generation of a quadrilateral element mesh for shells with arbitrary three-dimensional geometry. The present mesh generation scheme incorporates a normal mesh generator for generating a mesh in the two-dimensional plane and a specific mapping technique which maps the two-dimensional mesh onto the three-dimensional curved surface. As the mapping is a one-to-one mapping between the mesh in the plane and that on the curved surface, the resulting surface discretization is compatible with the local mesh parameters in two dimensions. This scheme is further combined, both with a sophisticated error estimate determined by using the best guess values of bending moments and membrane and transverse shear forces obtained from a previous solution, and an effective mesh refinement strategy established at an element level in order to complete an adaptive analysis for shell structures. Numerical examples are shown to illustrate the principles and procedure of the present adaptive analysis.     

Subdivision elements for large deformation of liquid capsules enclosed by thin shells

This paper presents a numerical method for studying the large deformation of a liquid capsule enclosed by a thin shell in a simple shear flow. An implicit immersed boundary method is employed for calculating the hydrodynamics and fluid–structure interaction effects. A thin-shell model for computing the forces acting on the shell middle surface during the deformation is described within the framework of the Kirchhoff–Love theory of thin shells. This thin-shell model takes full account of finite-deformation kinematics which allows thickness stretching as well as large deflections and bending strains. The interpolation of the reference and deformed surfaces of the shell is accomplished through the use of Loop's subdivision surfaces. The resulting limit surface is C¹-continuous which significantly improves the ability of the method in simulating capsules enclosed by hyperelastic thin shells with different physical properties. The present numerical technique has been validated by several examples including an inflation of a spherical shell and deformations of spherical, oblate spheroidal and biconcave capsules in the shear flow. In addition, different types of motion such as tank-treading, tumbling and transition from tumbling to tank-treading have been studied over a range of shear rates and viscosity ratios.     

Link-state routing without broadcast storming for multichannel mesh networks

A link-state routing protocol tailored for multichannel mesh networks is proposed. One drawback of using multichannel communications is the high overhead involved in broadcast operations: a transmitter should transmit a broadcast packet to all channels that may be occupied by receivers. This makes certain broadcast-intensive mechanisms, such as link-state routing, difficult to implement. The link-state routing protocol proposed in this paper is tailored for multichannel mesh networks by minimizing the broadcast overheads. This is achieved by a special set of nodes, called cluster-heads. We have implemented our protocol on a multichannel mesh network test bed and compared its performance with an AODV-like reactive routing protocol, also tailored for multichannel mesh networks. The measurements show that the proposed link-state routing protocol provides transient communications with comparable or better performance. Ways to improve the performance of the proposed routing with infrastructure access is also discussed.     

Automatic triangular mesh generation in flat plates for finite elements

The paper describes a method of generating triangular finite elements for a two dimensional region bounded by rectangles. The work is divided into two parts: Part 1 describes the method dealing with the problem of generating graded triangular meshes for two dimensional regions. The mesh could be uniform or graded in four directions at prescribed regions where stress concentrations are expected to appear. Five typical groups of triangular elements are introduced and their properties and recurring formulae referring mainly to nodal and triangular numbering are given in detail. These properties are used later for developing the computer programme. Part 2 describes the process required for developing the computer programme in FORTRAN. The programme is divided into three sections (a) preparation of data (b) running of the programme and (c) printing of results.The method is demonstrated by an illustrative example of a simply supported thin beam loaded in its plane.     

New method for graded mesh generation of all hexahedral finite elements

Mapping method is widely applied by most of commercial mesh generators because of its efficiency, mesh quality. One of the obstacles to apply the mapping method and generate a graded all hexahedral mesh of high quality in an arbitrarily three-dimensional domain is the generation of hexahedral parent elements on a super-element that allows for gradations in three co-ordinate directions. This paper presents a pattern module’s method to generate the graded mesh of all hexahedral elements in a cube and thus improves the mapping method. The method requires few calculations.     

A two-dimensional mesh generator for variable order triangular and rectangular elements

An improved method is presented for generating variable order elements by superelement generation. This method is simple to apply and requires less execution time in comparison with other variable order mesh generators. Depending on geometrical complexity and material variation, the superelements are manually determined to be refined into high or low order elements. Different mesh generation subroutines are employed to generate elements of different orders. The refined elements of different orders are finally patched to form a hybrid mesh. A FORTRAN program is given to generate finite element meshes of linear and quadratic triangular and rectangular elements automatically.     

Automatic mesh refinement and data structure for multigrid finite elements techniques

A general algorithm for locally refining any conforming triangulation to generate a new conforming one is presented. The proposed algorithm ensures that all angles in subsequent refined triangulations are greater than, or equal to, half the smallest angle in the original triangulation, the shape regularity of all triangles is maintained and the transition between small and large triangles is smooth. The generated triangulations are nested, so it is possible to implement the approach with adaptive and/or multigrid techniques. A complete algorithm for solving two-dimensional elliptic boundary value problems adaptively by multigrid is presented. The development and implementation of the main parts of this algorithm; automatic mesh generator, a posteriori error estimator, refinement strategy and the multigrid solver are presented in some detail. An appropriate data structure is developed to meet the excess data required for the generation process also to keep track of different grid levels. By the aid of this data structure, it becomes easy to design simple algorithms to store only the non-zero elements of stiffness matrices for different grids and to design a very simple multigrid transfer operator. Numerical examples are presented to show the generated grid sequence for two different boundary value problems.     

Feature enhancement modules applied to a feature pyramid network for object detection

Pattern Analysis and Applications - A feature pyramid network (FPN) improves the ability of an object detection model to detect multiscale targets. However, the simple upsampling used in an FPN is...     

Algorithm for converting a forest of quadtrees to a binary array

An algorithm for reconstructing a binary array of size N sx N from its forest of quadtree representation is presented. The algorithm traverses each tree of the forest in preorder and maps each ‘black’ node into the spatial domain. The time complexity in mapping is O(log N × B_n + B_p), where B_n is the number of black nodes in the forest and B_p is the number of black pixels in the N × N array. The algorithm has been implemented on an Apple II.     

8- and 12-node plane hybrid stress-function elements immune to severely distorted mesh containing elements with concave shapes

By simply improving the first version of hybrid stress element method proposed by Pian, several 8- and 12-node plane quadrilateral elements, which are immune to severely distorted mesh containing elements with concave shapes, are successfully developed. Firstly, instead of the stresses, the stress function ? is regarded as the functional variable and introduced into the complementary energy functional. Then, the fundamental analytical solutions (in global Cartesian coordinates) of ? are taken as the trial functions for 2D finite element models, and meanwhile, the corresponding unknown stress-function constants are introduced. Thus, the resulting stress fields must be more reasonable because both the equilibrium and the compatibility relations can be satisfied. Thirdly, by using the principle of minimum complementary energy, these unknown stress-function constants can be expressed in terms of the displacements along element boundaries, which can be interpolated directly by the element nodal displacements. Finally, the complementary energy functional can be rewritten in terms of element nodal displacement vector, and thus, the element stiffness matrix of such hybrid stress-function (HS-F) element is obtained. This technique establishes a universal frame for developing reasonable hybrid stress elements based on the principle of minimum complementary energy. And the first hybrid stress element proposed by Pian is just a special case within this frame. Following above procedure, two 8-node and two 12-node quadrilateral plane elements are constructed by employing different fundamental analytical solutions of Airy stress function. Numerical results show that, the 8-node and 12-node models can produce the exact solutions for pure bending and linear bending problems, respectively, even the element shape degenerates into triangle and concave quadrangle. Furthermore, these elements do not possess any spurious zero energy mode and rotational frame dependence.     

A simple technique for converting from a pascal shop to a C shop

We discuss a simple technique that we used for quickly converting a group of Pascal programmers into a group of C programmers. Our basic approach was to use the literal definition facility of C to make C programs ‘resemble’ Pascal programs. The approach is of interest to programmers because of its utter simplicity, and to management level personnel, familiar with Pascal but not with C, because it allows them to engage in productive code reading.     

From scores to face templates: a model-based approach

Regeneration of templates from match scores has security and privacy implications related to any biometric authentication system. We propose a novel paradigm to reconstruct face templates from match scores using a linear approach. It proceeds by first modeling the behavior of the given face recognition algorithm by an affine transformation. The goal of the modeling is to approximate the distances computed by a face recognition algorithm between two faces by distances between points, representing these faces, in an affine space. Given this space, templates from an independent image set (break-in) are matched only once with the enrolled template of the targeted subject and match scores are recorded. These scores are then used to embed the targeted subject in the approximating affine (non-orthogonal) space. Given the coordinates of the targeted subject in the affine space, the original template of the targeted subject is reconstructed using the inverse of the affine transformation. We demonstrate our ideas using three, fundamentally different, face recognition algorithms: Principal Component Analysis (PCA) with Mahalanobis cosine distance measure, Bayesian intra-extrapersonal classifier (BIC), and a feature-based commercial algorithm. To demonstrate the independence of the break-in set with the gallery set, we select face templates from two different databases: Face Recognition Grand Challenge (FRGC) and Facial Recognition Technology (FERET) Database (FERET). With an operational point set at 1 percent False Acceptance Rate (FAR) and 99 percent True Acceptance Rate (TAR) for 1,196 enrollments (FERET gallery), we show that at most 600 attempts (score computations) are required to achieve a 73 percent chance of breaking in as a randomly chosen target subject for the commercial face recognition system. With similar operational set up, we achieve a 72 percent and 100 percent chance of breaking in for the Bayesian and PCA based face recognition systems, respectively. With three different levels of score quantization, we achieve 69 percent, 68 percent and 49 percent probability of break-in, indicating the robustness of our proposed scheme to score quantization. We also show that the proposed reconstruction scheme has 47 percent more probability of breaking in as a randomly chosen target subject for the commercial system as compared to a hill climbing approach with the same number of attempts. Given that the proposed template reconstruction method uses distinct face templates to reconstruct faces, this work exposes a more severe form of vulnerability than a hill climbing kind of attack where incrementally different versions of the same face are used. Also, the ability of the proposed approach to reconstruct actual face templates of the users increases privacy concerns in biometric systems.     

A mesh optimization algorithm to decrease the maximum interpolation error of linear triangular finite elements

We present a mesh optimization algorithm for adaptively improving the finite element interpolation of a function of interest. The algorithm minimizes an objective function by swapping edges and moving nodes. Numerical experiments are performed on model problems. The results illustrate that the mesh optimization algorithm can reduce the W ^1,∞ semi-norm of the interpolation error. For these examples, the L ², L ^∞, and H ¹ norms decreased also.     

Bounded irregular pyramid: a new structure for color image segmentation

This paper presents a new segmentation technique for color images. It relies on building an irregular pyramid into a regular one, presenting only nodes associated to homogeneous color regions. Hence, the size of the irregular pyramid is bounded. Segmentation is performed by rearranging the set of links among pyramid nodes. Unlike other hierarchical methods based on relinking procedures, our algorithm does not operate in an iterative way and it preserves region connectivity.     

SEON: a pyramid of ontologies for software evolution and its applications

The Semantic Web provides a standardized, well-established framework to define and work with ontologies. It is especially apt for machine processing. However, researchers in the field of software evolution have not really taken advantage of that so far. In this paper, we address the potential of representing software evolution knowledge with ontologies and Semantic Web technology, such as Linked Data and automated reasoning. We present Seon, a pyramid of ontologies for software evolution, which describes stakeholders, their activities, artifacts they create, and the relations among all of them. We show the use of evolution-specific ontologies for establishing a shared taxonomy of software analysis services, for defining extensible meta-models, for explicitly describing relationships among artifacts, and for linking data such as code structures, issues (change requests), bugs, and basically any changes made to a system over time. For validation, we discuss three different approaches, which are backed by Seon and enable semantically enriched software evolution analysis. These techniques have been fully implemented as tools and cover software analysis with web services, a natural language query interface for developers, and large-scale software visualization.