PROGRAM

The 6th WSEAS International Conference on
SIGNAL, SPEECH AND IMAGE PROCESSING

(SSIP '06)

Lisbon, Portugal, September 22-24, 2006

Friday, September 22, 2006

PLENARY LECTURE 1

Video Systems and Robot Arms

Professor Vincenzo Niola

Departement of Mechanical Engineering for Energetics

University of Naples Federico II

Via Claudio n. 21, 80125 Naples

Italy

E-mail: vincenzo.niola@unina.it

Abstract: Video applications represent an useful tool for many robotic applications. Among others, very interesting can be considered: the robot cinematic calibration and the trajectories recording.

First of all it is important to consider that, by a suitable cameras calibration technique, it is possible to record three dimensional objects and trajectories by means of a couple of television cameras.

By means of perspective transformation it is possible to associate a point in the geometric space to a point in a plane. In homogeneous coordinates the perspective transformation matrix has non-zero elements in the fourth row. An expression of perspective transformation is proposed with the scope  to introduce the perspective concepts for the application in robotic field.

By means of studies on a camera vision model, an algorithm for stereoscopic vision system has been obtained.

This algorithm will be used to apply vision model to robotic applications, mainly for robot’s mechanical calibration and three-dimensional trajectories recording, but also for general vision systems in robotic applications.

The proposed algorithm uses the fourth row of the Denavit and Hartemberg transformation matrix that, for kinematics’ purposes, usually contains three zeros and a scale factor, so it is useful to start from the perspective transform matrix.

A camera can be modelled as a thin lens and an image plane with CCD sensors. The objects located in the Cartesian space emit rays of light that are refracted from the lens on the image plane. Each CCD sensor emit an electric signal that is proportional to the intensity of the ray of light on it; the image is made up by a number of pixels, each one of them records the information coming from the sensor that corresponds to that pixel.

In order to indicate the position of a point of an image it is possible to define a frame u,v (see fig.6) which axes are contained in the image plane. To a given point in the space (which position is given by its Cartesian coordinates) it is possible to associate a point in the image plane (two coordinates) by means of the telecamera. So, the expression “model of the camera” means the transform that associates a point in the Cartesian space to a point in the image space.

The proposed techniques can be also used for the robot cinematic calibration. The procedure can be summarized in two main steps:

I.       positioning and orientation error of the end-effector in a given number in the work space:

II.      developing of a mathematic technique to predict and offset the errors.

The cinematic calibration techniques generally doesn’t not consist in the direct measurement of the geometric parameters of the robot arm but needs the possibility to measure the end-effector position with a very high accuracy.

So, the proposed calibration technique can be applied to existing industrial robots and doesn’t require to set up a complex device, as it is based on the employment of a vision system and uses a couple of telecameras.

Another application of vision systems in robotics is the trajectory recording; this is essential to study robot arm dynamical behaviour has been obtained by means of two digital television camera linked to a PC.

A vision algorithm is proposed by means of which it is rather easy to record trajectories of a point belonging to a robot arm in the three dimensional space.

The rig, that has been developed, allows us to obtain the velocity vector of each point of the manipulator by means of which it is possible:

-        to control the motion giving the instantaneous joint positions and velocities;

-        to measure the motions between link and servomotor in presence of non-rigid transmissions;

-        to identify the robot arm dynamical parameters.

PLENARY LECTURE 2

Scale Free Networks – A Challenge in Modeling Complexity

Professor Radu Dobrescu

"POLITEHNICA" University of Bucharest

Splaiul Independentei no.313

Faculty of Control and Computers

E-mail: radud@isis.pub.ro

Abstract: The Lecture proposes a model that relieves the characteristics of several complex systems having a similar scale free network architecture. The properties of this kind of networks are compared with those of other methods which are specific for studying complex systems: nonlinear dynamics and statistical methods. We place particular emphasis on scale free network theory and its importance in augmenting the framework for the quantitative study of complex systems, by discussing three important applications: Internet topology and traffic characteristics, epidemics broadcast and cellular communication system in biological networks. Finally the new ways in modeling complex systems with scale-free networks are discussed.

PLENARY LECTURE 3

Univariance Optimization in High Dimensional Model Representation over Uniformly Data Filled Hypergrid

Professor Metin Demiralp

Informatics Institute

Istanbul Technical University, Turkey

E-mail: demiralp@be.itu.edu.tr

Abstract: Recent fifteen years brought a new powerful tool which is now called High Dimensional Model Representation to multivariate analysis. It is a divide–and–conquer type algorithm and finds its roots in the works of Sobol, Rabitz’s group, and most recently Demiralp’s group. It is based on an expansion in ascending multivariance such that the components of the expansion start with a constant followed by univariate components each of which depends on a different independent variable. The next terms are bivariate functions followed by the trivariate functions and so on. HDMR contains a finite number of components (2N if the number of the independent variables is N). However this number may become impractically large when N tends to grow higher values like hundreds, thousands. In these circumstances and generally for the practical point of view the univariate truncation of HDMR is desired to be an approximation for the multivariate function.

The dominancy of univariance may not be encountered in certain multivariate functions. These cases urge us to increase this dominancy by optimizing certain flexibilities. Since HDMR contains a weight function which can be somehow arbitrarily chosen, the choice becomes important since it affects the dominancy of constant and univariate components of HDMR.

HDMR’s weight function can be chosen as a continuous function or a generalized function like the product of certain linear combinations of Dirac’s delta function. The latter one becomes the only alternative when the multivariate function under consideration is given not analytically but a finite set of values on a hypergrid whose all nodes are accompained by the corresponding values of the multivariate function under consideration.

Since there are flexibilities in the coefficients of the linear combination of the delta functions they can be optimized to get maximum univariance in HDMR.

Lecture will be held at phenomenological level although sufficient instructions will also be given for numerical implementations.

PLENARY LECTURE 4

Mixed Discretization-Optimization Methods for Optimal Control of Nonlinear Parabolic Systems

Professor Ion Chryssoverghi

Department of Mathematics, School of Applied Mathematics and Physics

National Technical University of Athens

Zografou Campus, 15780 Athens

GREECE

E-mail: ichris@central.ntua.gr

Abstract: An optimal control problem is considered, for systems governed by a parabolic partial differential equation, jointly nonlinear in the state and control variables, with control and state constraints. Since no convexity assumptions are made on the data, this problem may have no classical solutions, and thus it is also formulated in the relaxed form. The classical and relaxed problems are discretized by using a finite element method in space and an implicit theta-scheme in time, while the controls are approximated by blockwise constant classical or relaxed controls. Various necessary/sufficient conditions for optimality are given for the control problems, in the continuous and discrete cases. Results are then obtained on the behavior in the limit of discrete optimality, and of discrete admissibility and extremality. Next, we propose a conditional descent method, applied to the discrete relaxed problem, and a penalized gradient projection method, applied to the discrete classical problem, and also progressively refining versions of these methods that reduce computing time and memory. The behavior in the limit of sequences constructed by these methods is examined. Finally, several numerical examples are given.

PLENARY LECTURE 5

From the Magic Square to the Optimization of Networks of AGVs and from MIP to an Improved GRASP like Optimization Algorithm and from this one to an Improved Evolutionary Algorithm

Professor Jose Barahona da Fonseca

Department of Electrical Engineering and Computer Science

Faculty of Sciences and Technology

New University of Lisbon

2829-516 Monte de Caparica, Portugal

E-mail: jbfo@fct.unl.pt

Abstract: In a previous work we presented an algorithm inspired in the Strong Artificial Intelligence and in the minimax optimization that imitates the human being in the solution of the magic square and we showed that in most cases its performance was much better than the human’s performance and even better than the performance of the best algorithms to solve the magic square, in terms of number of changes.

In this paper we adapt and transform this algorithm to solve the optimization of an AGVs network problem, using as a first test case 9 workstations in fixed positions and 9 operations to be executed, and the optimization problem is translated in the search of which of the 9! possible manners to distribute 9 operations by the 9 workstations that minimizes the total production time for a given plan of production.

As a final validation test, using random search, in 1000 runs it never reached the optimal solution at the end of 100000 iterations.

Finally we considered the more general case where the number of workstations is greater than the number of operations, and so there are some workstations that make the same operation, and we will have a layout with repetitions and multiple trajectories that implement the same product. This turns the problem more complex since when a product has operations that are executed by various workstations we must search all the possible combinations and find the average distance over all possible trajectories associated to a product. Furthermore the generation of all ‘permutations with repetitions’ is more complex and in the literature there are no published algorithm to generate this type of combinatorial entities. The Mixed Integer Programming approach proves to be impractical even for a simple test case of two products defined as sequences of four operations since the implementation of the division of the total distance over all trajectories that implement a product by their number turns the MIP model very big and combinatorial explosive. Using the BDMLP Solver with the GAMS software we only did obtain a sub-optimal solution that corresponds to a production time of 752s (the optimal being 690s) after 5 hours of computation in a 3.6GHz clock Pentium IV with 2G RAM and after exhausted the memory. Next using the CPLEX Solver we already obtain the optimal solution after 5.6 hours of computation. Again our algorithm adapted to layouts with repetitions presented very good results for this simple test case of 9 machines, 4 operations and 2 products. Finally we adapt and improve the OmeGA algorithm [1] and we apply it to our test cases and we got much better runtimes and almost always the optimal solution.

[1] D. Knjazew, OmeGA: A Competent Genetic Algorithm for Solving Permutation and Scheduling Problems, Kluwer Academic Publishers, 2002.

Saturday, September 23, 2006

SESSION: Image Processing: Analysis and Advanced Technology

Chair: Kriengkrai Porkaew, Alexei Zakharov

SESSION: Advanced Signal processing Applications

Chair: Alberto Pérez, Giorgio Quaglia

Sunday, September 24, 2006

SESSION: Advanced Image Analysis and Applications

Chair: Imran Touqir, Villegas and Raúl

SESSION: Biomedical Signal Processing and Applications

Chair: Maria Eugenia Torres, Rajasvaran Logeswaran

SESSION: Signal Processing Analysis and Advanced Applications

Chair: Roman M. Vitenberg, D. Berkani