leff@smu.UUCP (Laurence Leff) (05/31/88)
Below is a listing of Spring 1988 technical reports from Department of Electrical Engineering, Tulane University. Each paragraph lists: author(s), title of paper, technical report number 88-xxx, date, and finally the abstract. Technical reports are free. If you would like a copy, you can request them by writing to: Lauren Oddo Department of Electrical Engineering Tulane University New Orleans, LA 70118-5674 Listing of reports: =================== Ansari, A. and A. B. Martinez A Bayesian Approach to Optimal Quantization for Distributed Detection 88-009 05-19-88 The problem of optimal quantization by a network of distributed sensors is considered. It is shown that the quantization rule at the sensors and the fusion rule at the central processor are coupled. To minimize the cost, a multidimensional search for the parameters is needed. Suboptimal systems are obtained for either a fixed sensor performance or a given fusion rule. Computational complexity reduction is achieved for a restricted case of identical quantization rule at each sensor. Ansari, A. and A. B. Martinez Locally Optimum Quantization for Distributed Detection 88-008 04-22-88 A distributed network of sensors is considered where each sensor quantizes its observation optimally. It is shown that if the global decision test statistic is obtained as the sum of the statistics of each sensor, then the efficacy of the network is maximized by maximizing the efficacy of each sensor. The quantizer minimizing the mean square error between the locally optimum nonlinearity and the quantizer output can, therefore, be used at each sensor. The result is an extension of locally optimum quantization to the multisensor case. Barad, Herb An Object-Oriented Prototype of a Pyramid Architecture for Computer Vision 88-004 03-02-88 This paper describes a working software prototype of a pyramid architecture to investigate its use and effectiveness in computer vision. The results demonstrate that the pyramid architecture is an effective architecture for a wide range of computer vision tasks. An object-oriented methodology for modeling the individual processors and ports is used. The method of modeling and constructing the prototype is efficient and flexible. The prototype is used as a testbed for simulations of computer vision tasks. The segmentation results are presented. Implications of these results are discussed as well as the advantages of using this method of modeling to examine alternative architectures. Barad, Herb Rapid Prototyping of Massively Parallel Architectures 88-010 05-30-88 This paper describes a working software prototype of a pyramid architecture, known as the SCOOP pyramid. The purpose of the prototype is to analyze the effectiveness of the architecture for computer vision work. The SCOOP pyramid was built using Smalltalk. Each functional element of the entire architecture (i.e., processors, internal elements of processors, ports connecting processors, etc.) was modeled using different classes. The SCOOP environment is flexible enough to handle architectures of different topologies and those created from different types of hardware (e.g. optical). The construction of a prototype encourages the fine-tuning of the architecture design. The prototype is used as a test bed for simulations of computer vision tasks and the results of some of these simulations are presented. The SCOOP environment is a paradigm in which one can rapidly prototype architectures of arbitrary topology, complexity, and composition. The pyramid architecture is shown by simulation to be an effective architecture for a wide range of computer vision tasks from low level pixel-oriented operations to segmentation to high level symbolic operations. A description of the environment and results of some benchmarks are presented. Bruce, R. N., R. L. Drake, C. J. Sperry, and H. A. Thompson Nondestructive Inspection of Cable Stays by Electronic Means 88-006 04-20-88 The research reported here concerns a technique for nondestructive inspection of cable stays for a cable-stayed bridge at Luling, Louisiana. The method uses acoustic pulse transmission through the steel cables. The results of laboratory tests on stay samples and on test cables are described. These results establish the apparent feasibility of the method. Field measurements at the bridge are described. These measurements indicate that the cables tested are free of major faults. Certain difficulties have been encountered in field tests. These are described. Recommendations are made to improve the practicality and the applicability of the proposed method. Lee, M. S., B. S. Bourgeois, S. T. Hsieh, and A. B. Martinez A Laser Sensing Scheme for Detection of Underwater Acoustic Signals 88-005 04-15-88 A technique has been developed to detect underwater acoustic signals by utilizing the resultant surface perturbations. These perturbations result in an amplitude modulation effect on a laser beam reflected by the water surface. A theory is proposed to explain this phenomenon. A detection system was developed based on this theory. Laboratory investigations were conducted to evaluate the characteristics of the system. Preliminary results are included and indicate that detection is possible based on this technique. Martinez, Andrew B., Richard T. Joy, Maria K. Kalcic, Greg Terrie, and Stephen P. Haimbach Adaptive Estimation of Water Depth Using Multispectral Remote Sensing 88-007 04-22-88 An adaptive procedure for the estimation of water depth from passive multispectral scanner data is presented. While many authors have proposed nonadaptive, model-based estimators, most are computationally intensive and require accurate estimates of model parameters (directly or through regression) and bottom classification. By using an adaptive estimator based on the LMS algorithm, computational overhead is greatly reduced. Parameter estimation is unnecessary due to the inherent robustness of the algorithm to changes in ocean environment. This results in significant improvements in performance. Examples are given illustrating these points, and comparisons are made of methods based on adaptive estimation and on regression. Trade-offs between rate or convergence and residual error are discussed. -- Herb Barad Electrical Engineering Dept., Tulane Univ. INTERNET: barad@tulane.edu USENET: barad@tulane.uucp