Date of Award
Level of Access Assigned by Author
Master of Science (MS)
Spatial Information Science and Engineering
Second Committee Member
Third Committee Member
From being an expensive toy in the hands of governmental agencies, computers have evolved a long way from the huge vacuum tube-based machines to today's small but more than thousand times powerful personal computers. Computers have long been investigated as the foundation for an artificial vision system. The computer vision discipline has seen a rapid development over the past few decades from rudimentary motion detection systems to complex modekbased object motion analyzing algorithms. Our work is one such improvement over previous algorithms developed for the purpose of object motion analysis in video feeds. Our work is based on the principle of multi-dimensional object signatures. Object signatures are constructed from individual attributes extracted through video processing. While past work has proceeded on similar lines, the lack of a comprehensive object definition model severely restricts the application of such algorithms to controlled situations. In conditions with varying external factors, such algorithms perform less efficiently due to inherent assumptions of constancy of attribute values. Our approach assumes a variable environment where the attribute values recorded of an object are deemed prone to variability. The variations in the accuracy in object attribute values has been addressed by incorporating weights for each attribute that vary according to local conditions at a sensor location. This ensures that attribute values with higher accuracy can be accorded more credibility in the object matching process. Variations in attribute values (such as surface color of the object) were also addressed by means of applying error corrections such as shadow elimination from the detected object profile. Experiments were conducted to verify our hypothesis. The results established the validity of our approach as higher matching accuracy was obtained with our multi-dimensional approach than with a single-attribute based comparison.
Srinivasan, Sabeshan, "Object Tracking in Distributed Video Networks Using Multi-Dimentional Signatures" (2006). Electronic Theses and Dissertations. 564.