Nikolaos G. Bourbakis (Advisor), Soon M. Chung (Committee Member), Yong Pei (Committee Member), Arnab K. Shaw (Committee Member)
Doctor of Philosophy (PhD)
The recognition of single objects is an old research field with many techniques and robust results. The probabilistic recognition of incomplete objects, however, remains an active field with challenging issues associated to shadows, illumination and other visual characteristics. With object incompleteness, we mean missing parts of a known object and not low-resolution images of that object. The employment of various single machine-learning methodologies for accurate classification of the incomplete objects did not provide a robust answer to the challenging problem. In this dissertation, we present a suite of high-level, model-based computer vision techniques encompassing both geometric and machine learning approaches to generate probabilistic matches of objects with varying degrees and forms of non-deformed incompleteness. The recognition of incomplete objects requires the formulation of a database of six sided views (e.g., model) of an object from which an identification can be made. The images are preprocessed (K-means segmentation, and region growing code to generate fully defined region and segment image information) from which local and global geometric and characteristic properties are generated in a process known as the Local-Global (L-G) Graph method. The characteristic properties are then stored into a database for processing against sample images featuring various types of missing features. The sample images are then characterized in the same manner. After this, a suite of methodologies is employed to match a sample against an exemplar image in a multithreaded manner. The approaches, which work with the multi-view model database characteristics in a parallel (e.g, multithreaded manner) determine probabilistically by application of weighted outcomes the application of various matching routines. These routines include treating segment border regions as chain codes which are then processed using various string matching algorithms, the matching by center of moments from global graph construction, the matching of chain code starting segment location, the differences in angles in the center of moments between the model and sample images to find the most similar graphs (e.g., image), and the use of Delaunay triangulations of the center of moments formed during global graph construction. The ability to find a most probable match is extensible in the future to adding additional detection methods with the appropriate weight adjustments. To enhance the detection of incomplete objects, separate investigations have been made into rotating the exemplars in standard increments and by object extraction of segment border regions’ chain codes and subsequent synthesis of objects from the multi-view database. This approach is novel and potentially extensible to compositing across multi-view segmented regions at the borders between views by either human aided input of border relations or a systematic, automated evaluation of common border objects between the views of an exemplar. The first results are promising and trigger again the field of recognition of incomplete objects in a different way that was not studied before.
Department or Program
Department of Computer Science and Engineering
Year Degree Awarded
Copyright 2019, all rights reserved. My ETD will be available under the "Fair Use" terms of copyright law.