Browsing by Author "de Figueiredo, Rui J. P."

Now showing 1 - 20 of 20
  • Thumbnail Image
    Item
    3-D segmentation and volume estimation of radiologic images by a novel, feature driven, region growing technique
    (1992) Agris, Jacob Martin; de Figueiredo, Rui J. P.
    Magnetic Resonance (MR) imaging is a 3-D, multi-slice, radiological technique that acquires multiple intensities corresponding to each voxel. The transverse relaxation time, T$\sb1$, and the axial relaxation time, T$\sb2$, are two commonly obtained intensities that tend to be orthogonal. Automated segmentation of 3-D regions is very difficult because some borders may be delineated only in T$\sb1$ images, while others are delineated only in T$\sb2$ images. Classical segmentation techniques based on either global histogram segmentation or local edge detection often fail due to the non-unique and random nature of MR intensities. A 3-D, neighborhood based, segmentation method was developed based on both spatial and intensity criteria. The spatial criterion requires that only voxels connected by an edge or face to a voxel known to be in the region be considered for inclusion. Therefore, the region "grows" outward from an initial voxel. An intensity criterion that tries to balance local and global properties must also be satisfied. It determines the vector distance between the intensity of the voxel in question and a characteristic intensity for the neighboring voxels known to be in the region. Voxel intensities within a 95% confidence interval of the characteristic intensity are considered part of the region. The kernel size used to determine the characteristic intensity determines the balance between global and local properties. The segmentation terminates when no additional voxels satisfy both spatial and error criteria. Some regions, such as the brain compartments, are highly convoluted, resulting in a large number of border voxels containing a mixture of adjoining tissues. A sub-voxel estimate of the fractional composition is necessary for accurate quantification. A least-squares estimator was derived for the fractional composition of each voxel. Additionally, a maximum likelihood estimator was derived to globally estimate the fraction for all mixture voxels. This estimator is a minimum variance estimator in contrast to the least-squares estimator. The estimation methods in conjunction with the 3-D, neighborhood based, segmentation method resulted in an automated, highly accurate, quantification technique shown to be successful even for the brain compartments. Widespread applicability of these methods was further demonstrated by segmentation of kidneys in CT images.
  • Thumbnail Image
    Item
    A theory of photometric stereo for a general class of reflectance maps
    (1990) Tagare, Hemant D.; de Figueiredo, Rui J. P.
    Photometric stereo is an image processing technique for 2$1\over 2$ dimensional surface reconstruction from local shading. The classical theory of photometric stereo has been developed only for surfaces that reflect in a Lambertian plus specular manner. However, there is plenty of experimental evidence that most real-world surfaces are not Lambertian plus specular. This thesis develops the theory of photometric stereo for non-Lambertian surfaces. First, based on the physics of reflection and scattering, a general class of reflectance maps is proposed. This class is shown to model real world data more accurately than the Lambertian model. Then, the normalized photometric stereo equation using these reflectance maps is analyzed and conditions for a globally unique solution for the equation are obtained. Furthermore, the un-normalized photometric stereo equation is studied and conditions for getting a globally unique solution using only three light sources are identified. The problem of jointly estimating the reflectance map and the surface normal is proposed and shown to be ill-posed. A regularized solution to the problem is demonstrated. Finally, it is shown that extra light sources are needed to obtain a complete reconstruction of the surface, and the number of new light sources needed to achieve this is identified.
  • Thumbnail Image
    Item
    Adaptive robot path planning with obstacle avoidance in a dynamic environment
    (1991) Pham, Trung Tat; de Figueiredo, Rui J. P.
    The problem of path planning for a robotic system is considered, under the conditions in which both the target to be reached and the obstacles with which collision must be avoided are moving in a way not known in advance. To address this situation, an adaptive path planning scheme is proposed: A path is designed for a short time interval I between consecutive data points based on the data gathered at the beginning of I. In addition to this adaptive path planning scheme, a key contribution of this research has been the introduction, in the obstacle avoidance problem, of a new artificial potential energy function. This function, induced by an obstacle, depends not only on the robot's position but also on its velocity. Its use permits avoidance of false alarm and trapping caused by obstacles. Both the cases of a mobile robot system and a manipulator system are considered in detail. For a mobile robot, the system is modeled as a linear time-invariant system. The path planning problem is formulated in the form of two optimization problems: (i) Find the path as if there are no obstacles by minimizing the error between the state of the robot and that of the destination, and (ii) Find the path with obstacle avoidance by minimizing the sum of the deviation from the path computed in (i) and the artificial potential energy function resulting from the obstacles. For a manipulator, the problem is formulated and solved in a similar way. Solutions, analytical examples, and computer simulations are presented for both the mobile robot and manipulator cases.
  • Thumbnail Image
    Item
    Control methods for rigid and flexible kinematically redundant robot manipulators
    (1991) Nguyen, Luong An; de Figueiredo, Rui J. P.; Walker, Ian D.
    In the coming era of the Space Station Freedom, many robotic manipulators will be working simultaneously on various parts of the space station structure. Since each link of a manipulator is a moving body relative to other bodies, they form a tree structure of interconnected bodies. Undoutedly, many of these manipulators will possess certain degree of flexibility, at links as well as at joints. To describe accurately the dynamics of such a system in a generic way, is certainly not a trivial task. However, if we approximate each flexible link as a long slender beam and assume that flexible joints behave like torsional springs, a scalar set of equations of motion can be derived explicitly for use in real time simulation or control applications. Kinematic redundancy of manipulators has been used in control algorithms to avoid singularities, evade obstacles, minimize joint torques, manipulator kinetic energy, end effector contact forces, etc ... All of these approaches have been associated with rigid manipulators where there are no unpredictable flexible motions. When dealing with flexible manipulators, the flexibility of the system will cause undesired inaccuracy in end effector motion. However, if these manipulators are kinematically redundant, we show in this thesis that their kinematic redundancy can be used to compensate for the end effector motion inaccuracy and in many cases help damp out the vibrations. Based on our newly developed dynamic model of a system of multiple space-based flexible manipulators, control algorithms are designed to regulate the flexibility while maintaining precise tracking of the end effector trajectory. These control algorithms can either utilize a manipulator kinematic redundancy to control its flexibility or borrow other arms' motion to accomplish the same task provided that the arms are interconnected. Kinematic redundancy is also useful in optimizing the robustness of controllers. Given a manipulator's characteristics, parameter uncertainties, desired trajectories and controller design, it can be shown that under certain conditions, the tracking errors in end effector space are L$\sb\infty$ bounded. If the manipulator is kinematically redundant, we show how its redundancy can be used to minimize these tracking error bounds and in some cases, stabilize an unstable system.
  • Thumbnail Image
    Item
  • Thumbnail Image
    Item
  • Thumbnail Image
    Item
    GENERALIZED REPLACEMENT IN ITERATIVE RESTORATION ALGORITHMS
    (1983) BRZUSTOWICZ, THOMAS JUDE; de Figueiredo, Rui J. P.
    A new formulation of an iterative restoration algorithm is presented in this dissertation for the solution of noisy, non-linear restoration problems. The constraints and distortions are assumed to be known and are modeled as general (not necessarily linear) projections. The constraints are divided into two fundamentally different types, system constraints and signal constraints, of which only the latter are susceptible to noise. Imposing the system constraints and imposing the signal constraints are performed as two separate steps. Comparison of the signals produced by each step enables one to determine whether or not the generated sequence of signals is converging. Without the separation of steps certain types of errors would be undetectable. The problem caused by additive noise is overcome by modifying the replacement step. This simple change improves the quality of the reconstructed or restored signal and the convergence properties of the algorithm. The replacement step is changed from the direct substitution of given signal values to the operation of projecting onto a set surrounding these given values. Projection onto a surrounding hypersphere, called partial replacement, is the simplest case. Projection onto a surrounding hyperellipsoid, called windowed replacement, is a more complicated non-linear case that can take into account specific properties of the noise to more effectively restore the original signal. Both of these cases are developed and examined.
  • Thumbnail Image
    Item
    Global asymptotic stability of a class of nonlinear systems and networks
    (1970) Ho, Chen-yao; de Figueiredo, Rui J. P.
    Stability results for a class of nonlinear systems which is composed of several nonlinear subsystems are obtained by means of Liapunov's direct method. The desired Liapunov function is constructed by a linear combination of the Liapunov functions for the subsystems. The stability condition is expressed in terms of the positive definiteness of some matrix. Stability results for a class of .nonlinear systems and networks which are described by vector Lurie type system equations are also obtained through an extension of Popov's theorem. Under the assumption that, with the nonlinear elements deleted, the system is asymptotically stable and nonlinear characteristic of each element is constrained to a sector, the steps in the proof of Popov's theorem are followed. The stability condition can be achieved by requiring some Hermitian matrix to be positive definite. The results are extended to the case where time delays are involved. Finally, a system of nonlinear networks interconnected by lossless transmission lines is considered. Here both time delay and loading effects are introduced. The stability results are derived in the same manner as above. Examples are included to illustrate the results.
  • Thumbnail Image
    Item
    Hierarchical robot path planning using a distributed blackboard
    (1990) Ciscon, Lawrence A.; de Figueiredo, Rui J. P.
    Hierarchical path planning techniques for Intelligent Autonomous Mobile Robots (IAMRs) attempt to subdivide the overall task of path planning into several layers. This subdivision is made more difficult by the physical robot's computational architecture. The most important problem with classic hierarchies is their inability to cope with a diversity of environments and obstacles. In this research, we present a hierarchical path planning paradigm which uses a distributed blackboard architecture as its foundation. We subdivide the path planning problem into three levels: a global path planner, a local planner, and a set of constraint analyzers. By creating a uniform distributed blackboard that spans many processors in a network, we have formed the basis of a uniform operating environment for experimenting with layered path planning, as well as other mobile robot issues. And by dividing the planning into three layers instead of two, the system is able to handle much more complicated and diverse environments.
  • Thumbnail Image
    Item
    Image enhancement of a convex polyhedral object by optimal illumination
    (1989) Gateau, Aline Dominique; de Figueiredo, Rui J. P.
    It is possible to enhance the image of an object without any image processing by optimal illumination. This thesis first reviews some basic results in radiometry and image formation. Then it proposes a method for optimally illuminating a convex polyhedral object made of k Lambertian faces by a set of directional light beams according to an edge detection criterion. The illumination of the scene is considered to be optimal when the edges of the corresponding image have the best chance of being detected simultaneously. This can be expressed as a maximin optimization problem on a set of values characterizing the edges. The solution to this problem is derived in the case of a convex prism lighted by a single directional light beam and the results of an experiment corresponding to this case are given.
  • Thumbnail Image
    Item
    Machine recognition of occluded two-dimensional objects in clutter
    (1991) Giritlioglu, Bogac; de Figueiredo, Rui J. P.
    This thesis presents a tree search based approach for recognizing occluded 2-D objects in clutter. The planar shapes are represented in terms of their contour line segments. Each is allowed three degrees of freedom (x, y translation and rotation) and a variable scaling factor. It is shown that incorrect pairings of line segments may be eliminated efficiently by using ordering of line segments, local, and global constraints. It is also shown by simulation and by mathematical bounds that the number of hypotheses consistent with these constraints is small.
  • Thumbnail Image
    Item
    Microwave and optical sensor fusion for the shape extraction of three-dimensional space objects
    (1988) Shaw, Scott Warren; de Figueiredo, Rui J. P.
    Two sensors that have been proposed for use on a space robot are cameras and radar. Considered individually, neither of these sensors provides enough information for a computer to derive a good surface description of a remote object. Their combination, however, can produce a complete surface model. The lack of atmosphere in space presents special problems for optical image sensors. Frequently, edges are lost in shadow and surface details are obscured by diffraction effects caused by specularly reflected light. An alternate sensor for space robotic applications is microwave radar. The polarized radar cross-section (RCS) is a simple, well-understood, microwave measurement that contains limited information about a scattering object's surface shape. These two data sets are fused through an error minimization procedure. First, an incomplete surface model is derived from the camera image. Next, the unknown characteristics of the surface are represented by some parameter. Finally, the correct value for this parameter is computed by iteratively generating theoretical predictions to the RCS and comparing them to the observed value. A theoretical RCS may be computed from the surface model in several ways. One such RCS prediction technique is the method of moments. The method of moments can be applied to an unknown surface only if some shape information is available from an independent source. Here, the camera image provides the necessary information. When the method of moments is used to predict the RCS, the error minimization algorithm will converge in most cases. By combining the microwave and optical information in this way, the shapes of some three-dimensional objects have been accurately recovered. Simulations and experiments were performed on plates, ellipsoids, and an arbitrary curved object. Simulations show that error in the recovered shapes is very small when the RCS measurement error is not too large. Experiments prove that the RCS can be measured within this tolerance. In general, this investigation has shown the usefulness of sensor fusion applied to the shape reconstruction problem in space. Furthermore, a specific framework has been developed and proved effective for integrating the two types of sensors that are typically found on space vehicles.
  • Thumbnail Image
    Item
    Nonlinear system identification
    (1969) Netravali, Arun Narayan; de Figueiredo, Rui J. P.
    Stochastic approximation methods for the identification of parameters of nonlinear systems without dynamics have been widely discussed in the literature. In this work, two classes of discrete-time nonlinear dynamical systems driven by independent noise are considered. The measurements are assumed to be linear scalar and are made in the presence of independent noise. The systems under consideration are identified by the estimation of the parameters appearing in the evolution operator. These parameters are assumed to be constant during the identification time and they are estimated by means of stochastic approximation algorithms. A computer algorithm based on the above method is used to identify the parameters for some typical examples.
  • Thumbnail Image
    Item
    On asymptotic stability properties of direct control systems with a feedback characteristic lying partly outside the Hurwitz sector
    (1968) Dutertre, Jacques Andre; de Figueiredo, Rui J. P.
    In his original work,Popov developped a simple frequency domain criterion by which global asymptotic stability of a nonlinear control system is guaranteed for all nonlinearities lying inside a certain sector called the Popov sector. Following this result, particular attention has been devoted in the scientific literature to the cases in which the Popov sector differs from the Hurwitz sector, this with a view to establishing additional conditions on the nonlinearity that may permit the so-called Aizerman conjecture to be verified. In the present work, second order nonlinear control problems, with finite Popov sector, are investigated with respect to asymptotic stability of the origin for nonlinear feedback characteristics lying partly outside the Popov-Hurwitz sector, which seems to be a new viewpoint. A new kind of stability ,near asymptotic stability is introduced and studied for second order control systems with a piecewise linear feedback characteristic and is extended to systems with a general feedback characteristic of the type mentioned above. A theorem is given which permits to predict near asymptotic stability for such second order systems. All the studies carried out in this work are obtained through the use of a state space approach and their validity is illustrated both by analog and digital computer solutions.
  • Thumbnail Image
    Item
    On the dynamic range of forced systems via Liapunov functions
    (1967) De Vries, Michel Marie; de Figueiredo, Rui J. P.
    The forced linear and nonlinear systems considered are assumed to be asymptotically stable at large when the input is absent. Given prescribed bounds on the state variables and the output, the dynamic range of the system is defined as the maximum allowable range of the magnitude (norm) of the input that will keep the system variables within those prescribed bounds. In this work, the Liapunov functions for the unforced system are used to obtain an estimate of its dynamic range. The method developed is based on a theorem of Malkin. An estimate of the dynamic range of linear time invariant lumped-parameter systems is obtained by means of a quadratic form Liapunov function. It is shown that the best estimate results from a quadratic form generated by a diagonal matrix having all its elements equal (the common value of those elements being an arbitrary positive number). For the class of nonlinear control systems, using a Lurie-Postnikov Liapunov function, the best estimate is obtained by choosing the same Liapunov function as the one which determines the weakest requirements for absolute stability of the unforced system. A computer algorithm based on the above method is described and used in the solution of an example.
  • Thumbnail Image
    Item
    Optimum linear and non-linear transformations in feature extraction
    (1969) Caprihan, Arvind; de Figueiredo, Rui J. P.
    In pattern recognition one tries to classify a pattern based on a certain number of observed variables. The observed variables are more often than not redundant. The method of reducing the number of observed variables, or features, is called "Feature Extraction." The solution of the above problem requires the aid of a distance measure. Here a distance measure is obtained geometrically and is compared with the Bhattacharya distance and the Divergence for the n-dimensional Gaussian random variable. Then a method is given to optimally select the features of gaussian random process. A special case of this result gives the optimum linear transformation to reduce the dimension of an n-dimensional Gaussian random vector. The distance measure derived geometrically was used to obtain an optimum non-linear transformation to reduce the number of observed variables. An example was solved on the computer to give a certain amount of comparison among the different methods.
  • Thumbnail Image
    Item
    Programming using a constraint satisfaction system for robotics
    (1992) Raja, Basalat Ali; de Figueiredo, Rui J. P.
    Programming robots is a difficult task, especially for autonomous robots that must operate in an open, unstructured environment. A programmer building such a system should have a number of sophisticated tools at his/her disposal. In this thesis, we provide the basic framework for the development of a programming language that can be used to program an autonomous robot. The programming language is based on a constraint system. We describe our implementation of the constraint system, as well as our research done into developing a language based on this system.
  • Thumbnail Image
    Item
    Some new techniques for the design of recursive digital filters
    (1970) Kao, Charles Tzu-Tai; de Figueiredo, Rui J. P.
    Two new digital filter design techniques are proposed for obtaining the transfer function, in the z-domain, of a recursive digital filter which is to simulate the performance of a given continuous filter. The first one which we refer to as "frequency method" is obtained by minimizing the expected value of the squared error at the sampling instants when the input signal is band-limited. The resultant digital filters will have the best frequency responses in the Nyquist frequency range. The second one which we call the "spline method" relies on the approximation of the convolution integrand associated with the continuous filter by a simple cubic interpolating spline. The best approximation properties of the spline interpolation and the approximation error are studied in the context of the problem under consideration. Some numerical results obtained on a Burroughs 5500 digital computer are presented which illustrate the frequency responses of the resultant digital filters.
  • Thumbnail Image
    Item
    Spline-based methods for acquisition of signal and image models
    (1989) Kaul, Jeetendra; de Figueiredo, Rui J. P.
    In this thesis spline functions have been used to do data fitting for modeled signals. Optimum splines are obtained based on least squares error for the specific signal and image models. In the case of one-dimensional signal the model is obtained by using the Maximum Entropy Method (MEM). After obtaining the best estimates for the model parameters, the best type of spline is generated based on the source model chosen. For the two-dimensional case a linear least squares technique is used for the estimation of model parameters and then an optimal interpolating spline is used for image reconstruction.
  • Thumbnail Image
    Item
    The design of a laser range imaging device
    (1992) Denney, Bradley Scott; de Figueiredo, Rui J. P.
    In this thesis, a methodology and techniques for the design of a laser range imaging device (LRID) for robotic applications are presented. Furthermore, a prototype incorporating the concepts presented has been successfully tested in the laboratory. In this LRID, the range is determined by the shift in phase in the return of a modulated laser beam illuminating an object. The device is designed using a low power and visible (for laboratory safety reasons) diode laser and an APD detector. The device is also equipped with a fully programmable X-Y galvonometric scanning unit. The scene, which can be dark as well as brightly lit, is scanned in either a vector or raster mode and both range and reflected intensity data are collected. In addition to typical range data applications, the range and intensity images can be combined to provide useful data sets to aid in object recognition and scene understanding.