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

Title:	Ear recognition using stochastic clustering method and Iterative Closest Point algorithm
Authors:	Khamiss Masaoud Salem Ahmed (P61090)
Supervisor:	Khairuddin Omar, Prof. Dr.
Keywords:	Ear recognition Stochastic clustering Iterative Closest Point Dissertations, Academic -- Malaysia
Issue Date:	2-Jan-2016
Description:	The ear biometric is a perfect source of data for passive identification. The four important characteristics of ear biometrics: universality, uniqueness, permanence and collectability make it a very potential biometric trait for the identification of persons. Ear biometric has been deployed to improve the security level and hence reducing the complication for personal identification. There are some problems closely related to ear Identification techniques need to obtain the best performance and highest matching rate. Thus, this study aims: (i) To propose an ear detection method which is feature independent and environment oriented in order to produce optimum results in minimum possible of time, (ii) To reduce unnecessary calculations through using Integral Image (II) method and proposing a novel approach for feature point calculation by combining Stochastic Clustering Method (SCM) with Iterative Closest Point (ICP) algorithm to increase matching accuracy rate, (iii) To enhance ear identification performance based on classification via incorporating different characteristics of ear features, (iv) To evaluate the result of the proposed method in different datasets for validation. The proposed II with SCM and ICP for Ear Recognition (ISIER) model aims to minimize to overall computation complexity by eliminating unnecessary or duplicated calculations through this proposed method such as but not limited to segmentation, feature extraction and classification. The methodology of this research consists of four stages: (i) Preprocessing for Ear extractions are procedures based on Skin segmentation to remove the non-skin regions from the input image that followed by ear extraction based on individual components, (ii) Feature extraction comes with ISIER Model that include a sets of local features for ear image extraction, and combined SCM with ICP which enable a better combining of ear properties in order to have a successful identify human ear, (iii) Ear classification comes with Back-propagation Classifier to measure the ear similarity for the recognition resolutions to eliminate irrelevant ear images in comparison, (iv) Ear matching to clarify the issues identified from the data and method to achieve the research objectives. This experimental investigation is involving the time effect and acceptance rate on ear biometrics. For the determination of recognition, the study involves with an ear image and implemented using a JAVA program. The standard datasets that have been used are two different databases which are Indian Institute of Technology Kanpur (IIT Kanpur) database and benchmark database from the University of Science and Technology Beijing (USTB) consisting at least of 500 images from each database and compared with both West Pomeranian University of Technology Ear Database (WPUTED) and Mathematical Analysis of Images (AMI) database for image acquisition and define the performance of ear recognition. The average accuracy recognition rate of 98.37% is achieved for a significance improvement of the proposed ISIER Model. The detection acceptance rate is 97.18% for ICP, 96.72% for SCM, and 96.44% for PCA respectively. Finally, by integrating the predicted Ear recognition rate, the average time for ear recognition can be managed more efficiently.,Certification of Master's/Doctoral Thesis" is not available
Pages:	161
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Science and Technology / Fakulti Sains dan Teknologi

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