Face perception--Computer simulation Human face recognition (Computer science) Optical pattern recognition Computer Graphics
3D object structure estimation is important in object tracking, object recognition, video-conferencing, and scene analysis. This dissertation considers the problem of structure estimation from images with applications in vehicle tracking and face recognition. First, I present a new approach to object tracking by the convex hull vertices of the segmented object over the sequence of frames. The structure of the convex vertices and the 3D motion of the moving object are obtained through the tracking process. The second part of this thesis concentrates on the face structure estimation and recognition. A 3D face structure is extracted from a few image views of a human face taken at prior unknown poses by appropriately morphing a generic 3D face. A cubic explicit polynomial in 3D is used to morph a generic face into the specific face structure. The choice of the morphing function (a cubic polynomial) was dictated by the nature of the differences in shape between the face surfaces. The estimation of a 3D face structure as well as the view poses is achieved through the use of a distance map metric. The use of this metric avoids either resorting to the formidable task of establishing feature point correspondences in the image views, or even more severely, relying on the extremely view-sensitive image intensity (texture). Automatic Face Recognition (AFR) is one of the most successful applications of image analysis and understanding. However, recognition of faces under varying pose remains one challenging issue in AFR. This work introduces an approach to the task of identifying a person in the database from an arbitrary image. The estimation of the 3D face structure allows the creation of a 3D face database, from which the feature-based recognition is achieved after pose estimation. It also allows the synthesis of a virtual face to be matched with the test image in intensity-based face identification. Finally a mechanism that fuses the feature-based face classifier with intensity-based classifier is also presented for achieving classification rates that surpass those obtained by each classifier on its own.
Metrics
37 File views/ downloads
18 Record Views
Details
Title
3D structure estimation from images with applications in object tracking and recognition
Creators
Chongzhen Zhang
Contributors
Fernand S. Cohen (Advisor) - Drexel University, Drexel University (1970-)
Awarding Institution
Drexel University
Degree Awarded
Doctor of Philosophy (Ph.D.)
Publisher
Drexel University; Philadelphia, Pennsylvania
Number of pages
xi, 130 pages
Resource Type
Dissertation
Language
English
Academic Unit
College of Engineering (1970-2026); Drexel University