Document Details

Document Type : Thesis 
Document Title :
Image Processing for Robust Object Detection and physical classification
معالجة الصور من أجل استخراج وتحديد الأهداف وتصنيف الخواص الفيزيائية
 
Subject : Faculty of Engineering 
Document Language : Arabic 
Abstract : In our new approach we succeed in achieving the desired result, to detect all edge information without losing the most significant detail information (junction). For localization of these information our significant initial step in our new approach is the necessary derivatives computation of images with two new invented 3x3 or 5x5 kernels (X, Y), which result gradient magnitude of a smoothed image by median filter. We precede the preliminary performance of detector to thin binary gradient magnitude by filtering it with two structure matrices horizontal and vertical direction. If more than one pixel selected as an edge pixel, then the structure matrix in horizontal direction will filtered out the neighbor pixel along the line of gradient magnitude, in this way in vertical direction. This process thins the broad ridges of gradient magnitude into ridges that are only one pixel width. The goal of this work is to improve the detection performance of corner features detectors with different transform condition of the same image. In this research a new proposed version of standard Harris detector developed, which emphasize all the issues and result a very powerful local features from the formulation of proposed detector. The improved version of Harris detector gives significantly better results than the other detectors in the presence of image rotation. In all cases the results of the improved version of the Harris detector are 50% to 80% better or equivalent to some of the other detectors. 
Supervisor : Dr.Amr Almaddah 
Thesis Type : Master Thesis 
Publishing Year : 1439 AH
2017 AD 
Added Date : Thursday, December 21, 2017 

Researchers

Researcher Name (Arabic)Researcher Name (English)Researcher TypeDr GradeEmail
توصيف أحمدAhmad, Tauseef ResearcherMaster 

Files

File NameTypeDescription
 42958.pdf pdf 

Back To Researches Page