Thesis (Ph.D.)--Hong Kong University of Science and Technology, 2006. / Includes bibliographical references (leaves 143-165). Also available in electronic version.

University of Macau / Faculty of Science and Technology / Department of Computer and Information Science

M.Ing. / A machine vision system must, by definition, be intelligent, adaptable and reliable to satisfY the objectives of a system that is highly interactive with its dynamic environment and therefore prone to outside error factors. A machine vision system is described that utilizes a 2D captured web cam image for the purpose of intelligent object recognition, gripping and handling. The system is designed to be generic in its application and adaptable to various gripper configurations and handling configurations. This is achieved by using highly adaptable and intelligent recognition algorithms the gathers as much information as possible from a 2D colour web cam image. Numerous error-checking abilities are also built into the system to account for possible anomalies in the working environment. The entire system is designed around four separate but tightly integrated systems, namely the Recognition, Gripping and Handling structures and the Component Database which acts as the backbone of the system. The Recognition system provides all the input data that is then used for the Gripping and Handling systems. This integrated system functions as a single unit but a hierarchical structure has been used so that each of the systems can function as a stand-alone unit. The recognition system is generic in its ability to provide information such as recognized object identification, position and other orientation information that could be used by another handling system or gripper configuration. The Gripping system is based on a single custom designed gripper that provides basic gripping functionality. It is powered by a single motor and is highly functional with respect to the large range of object sizes that it can grip. The Handling Sub-system controls gripper positioning and motion. The Handling System incorporates control of the robot and the execution of both predetermined and online adaptable handling algorithms based on component data. It receives data from the Component database. The database allows the transparent ability to add and remove objects for recognition as well as other basic abilities. Experimental verification of the system is performed using a fully integrated and automated program and hardware control system developed for this purpose. The integration of the proposed system into a flexible and reconfigurable manufacturing system is explained.

Please read the abstract in the section 00front of this document / Thesis (PhD (Electronic Engineering))--University of Pretoria, 2007. / Electrical, Electronic and Computer Engineering / unrestricted

This dissertation investigates the cost-complexity-performance relationship between two automatic language identification systems. The first is a state-of-the-art archi¬tecture, trained on about three hours of phonetically hand-labelled telephone speech obtained from the recognised OGLTS corpus. The second system, introduced by our¬selves, is a simpler design with a smaller, less complex parameter space. It is a vector quantisation-based approach which bears some resemblance to a system suggested by Sugiyama. Though trained on the same data, it has no need for any labels and is therefore less costly. A number of experiments are performed to find quasi-optimal parameters for the two systems. In further experiments the systems are evaluated and compared on a set of ten two-language tasks, spanning five languages. The more com¬plex system is shown to have a substantial performance advantage over the simpler design - 81% versus 65% on 40 seconds of speech. However, both results are well under reported state-of-the-art performance of 94% and would suggest that our systems can benefit from additional attention to implementation detail and optimisation of various parameters. Given the above, our suggested architecture may potentially provide an adequate solution where the high development cost associated with state-of-the-art technology and the necessary training corpora are prohibitive. / Dissertation (M Eng (Computer Engineering))--University of Pretoria, 2006. / Electrical, Electronic and Computer Engineering / unrestricted

This dissertation introduces an analytic approach to the problem of circuit network pattern recognition. The approach involves a two-stage methodology which is described in detail. Initially, a theoretical process is used to generate a bench mark, or yardstick, to which descriptions of form can be oriented; secondly, a link-by-link examination of circuit network structure is undertaken to determine each link's relationship to the bench mark. The graphs composing the bench mark have a continuous structural distribution, ranging from completely connected to rudimentary graphs. A comprehensive description of these graphs is presented, and several properties of the bench mark are examined and compared to those of known families of circuit graphs. It is argued that the bench mark introduced here is more flexible than other yardsticks because its generative process creates a continuous spectrum of graph structures. The internal link-by-link approach also allows comparisons to be made within the link structure of the empirical network and thus an intra-network examination of the network's pattern is possible. Such comparison aids in revealing trends within the link structure of a given empirical circuit network. Finally, an illustration of the application of the proposed approach is presented. Three road networks, all located in western Canada, are chosen as examples of empirical planar networks and the major airline network of Canada is used as a non-planar example. / Arts, Faculty of / Geography, Department of / Graduate

M.Ing. / Please refer to full text to view abstract

Advances in storage technology have long been driving the need for new data mining techniques. Not only are typical data sets becoming larger, but the diversity of available attributes is increasing in many problem domains. In biological applications for example, a single protein may have associated sequence-, text-, graph-, continuous and item data. Correspondingly, there is growing need for techniques to find patterns in such complex data. Many techniques exist for mapping specific types of data to vector space representations, such as the bag-of-words model for text [58] or embedding in vector spaces of graphs [94, 91]. However, there are few techniques that recognize the resulting vector space representations as units that may be combined and further processed. This research aims to mine important vector-item patterns hidden across multiple and diverse data sources. We consider sets of related continuous attributes as vector data and search for patterns that relate a vector attribute to one or more items. The presence of an item set defines a subset of vectors that may or may not show unexpected density fluctuations. Two types of vector-item pattern mining algorithms have been developed, namely histogram-based vector-item pattern mining algorithms and point distribution vector-item pattern mining algorithms. In histogram-based vector-item pattern mining algorithms, a vector-item pattern is significant or important if its density histogram significantly differs from what is expected for a random subset of transactions, using χ² goodness-of-fit test or effect size analysis. For point distribution vector-item pattern mining algorithms, a vector-item pattern is significant if its probability density function (PDF) has a big KullbackLeibler divergence from random subsamples. We have applied the vector-item pattern mining algorithms to several application areas, and by comparing with other state-of-art algorithms we justify the effectiveness and efficiency of the algorithms.

No description available.

Note: