feature selection using dynamic binary particle swarm...

Feature Selection using Dynamic Binary Particle

Swarm Optimization for Arabian Horse

Identification System

Presented by: Samar Ibrahem Youssef

Annual Conference of the Institute of Studies and Statistical Research 2017

Workshop in Intelligent Systems and Applications

MSc Student , Information Technology Department,

Faculty of Computers and Information,

Cairo University

Scientific Research Group in Egypt

27/12/2017

Outline

27/12/2017

Introduction.

Problem Statement.

Proposed Methodology:

Phase I : Pre-processing.

Phase II : Segmentation.

Phase III : Feature Extraction.

Phase IV : Feature Selection.

Conclusions & Future Work.

2 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Outline

27/12/2017

Introduction.

Problem Statement.

Proposed Methodology:

Phase I : Pre-processing.

Phase II : Segmentation.

Phase III : Feature Extraction.

Phase IV : Feature Selection.

Conclusions & Future Work.

3 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Introduction

Need for Arabian Horse Identification

Biosecurity.

Retrieval after theft.

Fairness in competition.

Medical record management.

27/12/2017 4 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Outline

27/12/2017

Introduction.

Problem Statement.

Proposed Methodology:

Phase I : Pre-processing.

Phase II : Segmentation.

Phase III : Feature Extraction.

Phase IV : Feature Selection.

Conclusions & Future Work.

5 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Problem Statement

27/12/2017

Tattooing

Ear Tagging

Traditional Identification Methods

Branding

6 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Problem Statement

Using Biometric identifiers (contactless methods) for identification

and getting rid of harmful identification methods .

Muzzle Print IRIS

27/12/2017 7 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Problem Statement

27/12/2017 8 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Outline

27/12/2017

Introduction.

Problem Statement.

Proposed Methodology:

Phase I : Pre-processing.

Phase II : Segmentation.

Phase III : Feature Extraction.

Phase IV : Feature Selection.

Conclusions & Future Work.

9 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Phase I: Pre-processing

Pre-Processing

Image Resizing

Convert to grayscale

Contrast Stretching

Noise Removal

Input Image Processed Image

27/12/2017 10 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Outline

27/12/2017

Introduction.

Problem Statement.

Proposed Methodology:

Phase I : Pre-processing.

Phase II : Segmentation.

Phase III : Feature Extraction.

Phase IV : Feature Selection.

Conclusions & Future Work.

11 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Phase II : Segmentation

27/12/2017

Pre-Processing

Image Resizing

Convert to grayscale

Contrast Stretching

Noise Removal

Segmentation

Otsu-IFOA

Opening & Masking

12 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Phase II: Segmentation(continued)

Otsu-IFOA Segmentation

Binary Mask Thresholded Image Segmented Area

27/12/2017

Processed Image

13 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Outline

27/12/2017

Introduction.

Problem Statement.

Proposed Methodology:

Phase I : Pre-processing.

Phase II : Segmentation.

Phase III : Feature Extraction.

Phase IV : Feature Selection.

Conclusions & Future Work.

14 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Phase III: Feature Extraction

Pre-Processing

Image Resizing

Convert to grayscale

Contrast Stretching

Noise Removal

Segmentation

Otsu-FOA

Opening & Masking

Feature Extraction

Gabor Filtering

TDCT

27/12/2017 15 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Phase III :Feature Extraction

27/12/2017

(Gabor Filtering)

16 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Phase III :Feature Extraction

(Gabor Filtering)

27/12/2017 17 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Phase III: Feature Extraction(TDCT)

Pre-Processing

Image Resizing

Convert to grayscale

Contrast Stretching

Noise Removal

Segmentation

Otsu-FOA

Opening & Masking

Feature Extraction

Gabor Filtering

TDCT

27/12/2017 18 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Phase IV: Feature Selection

27/12/2017

Pre-Processing

Image Resizing

Convert to grayscale

Contrast Stretching

Noise Removal

Segmentation

Otsu-FOA

Opening & Masking

Feature Extraction

Gabor Filter

TDCT

Feature Selection

DBPSO

19 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

The Flowchart of DBPSO

27/12/2017 20 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

DBPSO Results

We have 1800 extracted

features .

265 selected features using

DBPSO.

27/12/2017 21 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Outline

27/12/2017

Introduction.

Problem Statement.

Proposed Methodology:

Phase I : Pre-processing.

Phase II : Segmentation.

Phase III : Feature Extraction.

Phase IV : Feature Selection.

Conclusions & Future Work.

22 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017

Conclusions & Future work

Automatic Periocular region segmentation using

Otsu-IFOA.

Feature Extraction(Gabor Filter + TDCT).

Feature Selection using DBPSO.

Selected Features can be used for Arabian Horse

Identification System.

27/12/2017 23 / 23 Annual Conference of the Institute of Studies and Statistical Research 2017