Optimization of vision system pose for three-dimensional object motion estimation

dc.contributor.advisorBasir, O.A.
dc.contributor.authorTsui, Patrick P. C.
dc.degree.departmentSchool of Engineeringen_US
dc.degree.grantorUniversity of Guelphen_US
dc.degree.nameMaster of Scienceen_US
dc.description.abstractIn this thesis, a control paradigm for vision system pose planning is proposed for achieving reliable object motion estimates. The parameters of the vision system are controlled in the motion estimation process to adapt to the dynamic object motion behavior. A Kalman filter is employed as the motion estimator. The motion estimation uncertainties are developed as a function of the vision system parameters. The task of controlling the vision system is formulated as an optimization problem of finding the vision system poses that minimize the motion estimation uncertainties. A novel coordinate system transformation is developed to facilitate the optimization process. As a result of the transformation, the vision system poses are expressed in terms of the image plane poses. Therefore, the inverse kinematics of the vision system is needed for determining the vision system poses from the image plane poses. A novel approach for deriving closed-form solutions of this inverse kinematics problem is presented. The problem of object occlusion is formulated as an optimization constraint. A hybrid technique involving a genetic algorithm, simulated annealing and gradient-based search is proposed to search for optimal image plane poses.en_US
dc.publisherUniversity of Guelphen_US
dc.rights.licenseAll items in the Atrium are protected by copyright with all rights reserved unless otherwise indicated.
dc.subjectcontrol paradigmen_US
dc.subjectvision system pose planningen_US
dc.subjectobject motion estimatesen_US
dc.titleOptimization of vision system pose for three-dimensional object motion estimationen_US


Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
8.96 MB
Adobe Portable Document Format