Abstract
This dissertation develops a methodology to identify an object consisting of moveable subparts and to determine the position and orientation of those subparts. For the purpose of this research a simple object is defined as a single part capable of undergoing a rigid transformation. A complex object is an entity comprised of more than one simple object, and is defined in either two or three dimensional space. In order to accomplish this goal, three tasks are performed. First, a representation for complex objects is developed which is unique in that only one database instance is required. The complex object is described by a relational graph. In this graph, each subpart of the object is a node of the graph, and the relationship between adjacent nodes is given by labeled edges. The labels describe the allowable movement between adjacent nodes. Each subpart of the complex object is represented by a point pattern where each point corresponds to an extreme point of the object. Second, algorithms are developed for localized matching of the subparts of the complex object to determine their position and orientation. These algorithms are applicable in n-dimensional space and their results are invariant to rotation, translation, noise, missing points and extra points. Finally, global matching algorithms are developed to permanently assign the localized matchings. An algorithm is presented which makes global assignments based on the feasibility of adjacent subparts. This method conserves time in that it utilizes non-numerical methods to check the feasibility of local matchings.
Griffin, Paul Marshal (1988). A point pattern matching methodology for complex objects. Texas A&M University. Texas A&M University. Libraries. Available electronically from
https : / /hdl .handle .net /1969 .1 /DISSERTATIONS -991960.