خط مشی دسترسیدرباره ماپشتیبانی آنلاین
ثبت نامثبت نام
راهنماراهنما
فارسی
ورودورود
صفحه اصلیصفحه اصلی
جستجوی مدارک
تمام متن
منابع دیجیتالی
رکورد قبلیرکورد بعدی
Document Type:Latin Dissertation
Language of Document:English
Record Number:53977
Doc. No:TL23931
Call number:‭1434983‬
Main Entry:Mohammad Tayabur Rahman
Title & Author:Three dimensional model based recognition using multidimensional Hausdorff distance and geometrical invariantsMohammad Tayabur Rahman
College:University of South Alabama
Date:2006
Degree:M.S.
student score:2006
Page No:43-43 p.
Abstract:Several model based pattern recognition techniques have been proposed in the literature which primarily utilizes Hausdorff distance and geometrical invariants in two dimensional (2D) planes [1--3]. However, as the depth information is usually missed in a single 2D image, these techniques cannot accurately recognize an object. In this research, three dimensional (3D) matching algorithms using multi-dimensional Hausdorff distance and also geometrical invariants have been proposed. In the first algorithm a line based matching technique is used. For matching line sets, multidimensional Hausdorff distance minimization technique is used. A 3D matching algorithm using geometrical invariants for finding the relation between the 3D model and the stereo image pair is also proposed. By constructing a 3D invariant space, a 3D model can be represented as a set of points in the invariant space. While matching with the 2D image a set of invariant light rays in 3D coordinates is drawn, where each ray passes through a 3D invariant model point. If sufficient number of rays intersects the model in the 3D invariant space, then we infer that the model is present in the image. In the proposed methods, as the matching is performed using stereo image pairs, it is found to be more reliable and accurate compared to existing techniques.
Subject:Applied sciences; Electrical engineering; 0544:Electrical engineering
Added Entry:M. S. Alam
Added Entry:University of South Alabama