1. UKM Learning and Research Repository
2. PHD Thesis / Tesis PHD
3. Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Title:	Moving object detection using moment invariant motion modeling and edge based segmentation from aerial images
Authors:	A.F.M. Saifuddin Saif (P67408)
Supervisor:	Zainal Rasyid Mahayuddin, Dr.
Keywords:	Moving object detection Image processing Edge based segmentation Aerial images Dissertations, Academic -- Malaysia
Issue Date:	7-Jan-2016
Description:	Moving object detection is an image processing procedure for extracting moving objects which have a relatively apparent movement to background in image sequences based on edge, corner and color features. Moving object detection using aerial images is still an unsolved issue due to lack of motion model for moving object in the existing methods. Besides, current research works on moving object detection depend on segmentation or frame difference separately. Frame difference can detect motion pixel but cannot obtain complete object while segmentation can extract a more complete object but due to inability to differentiate moving regions from the basic static region background, segmentation alone could not grant the expected detection performance. In addition, most of the previous research works focus on detection rate rather than focusing on computationally less complex moving object detection which mainly depends on feature extraction. This research, therefore, aims to propose a new motion model based on moving object detection incorporated with a new segmentation method. Later, established motion model and segmentation method are combined to propose a new feature extraction algorithm to improve the computation time and detection rate. Three objectives were established in order to achieve the aim. Firstly, motion model called as Advanced Moment based Motion Unification (AMMU) is proposed where moment is used as principal component rather than using other feature like edge and corner. Secondly, a new segmentation method called as Segmentation Using Edge Difference (SUED) is proposed which calculates color difference on all neighboring pixels instead of 3x3 matrix multiplication operations upon all pixels. Finally, Moment based Fast Feature Extraction Algorithm (MFEA) is proposed to decrease computation time and increase detection rate for overall detection performance. Proposed AMMU achieved outstanding detection rate of 76.42% with remarkable minimal false alarm rate at 34.01%. Secondly, SUED achieved extended detection rate of 74.93% which is higher than previous research works based on applying frame difference and segmentation together or independently. Finally, MFEA achieved supreme detection rate of 81.19% with least false alarm rate at 29.91% where required computation time is 154.73 ms. In this research, three methods have been proposed: i) new motion model called as AMMU; ii) new segmentation method called SUED; iii) new feature extraction algorithm called MFEA. Experimental results reveal that proposed three methods are validated with computation time and false alarm rate reducing dramatically and increasing detection rate effectively.,Certification of Master's/Doctoral Thesis" is not available
Pages:	201
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.