Identifikasi Muzzle Pattern pada Sapi Menggunakan Convolutional Neural Network (CNN) dan Random Forest dengan Scale Invariant Feature Transform (SIFT)

Ilmi, Muhammad Rofiqul (2021) Identifikasi Muzzle Pattern pada Sapi Menggunakan Convolutional Neural Network (CNN) dan Random Forest dengan Scale Invariant Feature Transform (SIFT). Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

Text 06211740000087_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2023. Download (3MB) | Request a copy

Abstract

Identitas individu ternak penting untuk mengetahui rekam data individu ternak. Sapi perah merupakan ternak yang memiliki kebermanfaatan untuk dilakukan identifikasi. Kebermanfaatan tersebut seperti pencatatan kepemilikan, pencatatan vaksinasi, pencatatan inseminasi buatan, dan sebagainya. Alat identifikasi umum digunakan yaitu ear tag, neck tag, dan sebagainya. Kelemahan metode tersebut dapat diduplikasi, manipulasi, hilang atau rusak. Identifikasi biometrik lebih konsisten dan unik tiap individu dan tidak menambah suatu benda atau mengubah bentuk tubuh hewan. Corak moncong (muzzle pattern) salah satu biometrik yang dapat digunakan untuk identifikasi sapi. Identifikasi membutuhkan metode yang cepat dan akurat, serta metode perekaman data yang mudah dan cepat. Dalam tugas akhir ini, dilakukan perbandingan metode SIFT dengan Random Forest dan CNN dalam identifikasi muzzle pattern sapi perah. Penelitian ini menggunakan citra foto dan citra tinta. Variabel respon yang digunakan adalah individu sapi sebanyak 10 ekor. Kinerja klasifikasi terbaik pada citra foto dan tinta adalah Random Forest dengan SIFT. Nilai rata-rata kebaikan klasifikasi model tersebut pada data testing yaitu untuk akurasi sebesar 96,86%, presisi 97,52%, recall 96,83%, dan Fscore 96,57% pada citra foto dan nilai akurasi 98,57%, presisi 99%, recall 99% dan Fscore 98,67% pada citra tinta.
=====================================================================================================
Identification is important in keeping track of cattle’s individual records. The management of dairy cattle can benefit greatly from individual identification. Some benefits include recording ownership status, vaccination, and artificial insemination history. Commonly used techniques to identify individual cattle may include but not limited to branding/tagging the cattle on its neck, ear, or other parts of its body. The weakness of this tagging method is that the tag may be duplicated, manipulated, destroyed, or simply faded over time. In comparison to this manual technique, biometric identification is more consistent and unique, and does not give any modification to the cattle's physical body. One characteristic that can be used for biometric identification is muzzle pattern. Biometric identification system requires an accurate and fast method of acquiring and classifying data. This study carries out comparison of SIFT with Random Forest and CNN methods in the identification of individual dairy cattle. This study uses digital photo and ink-on-paper images. The response variable used is ten individuals of dairy cattle. This study concludes that combination of SIFT and Random Forest has best classification performance both on digital and ink-on-paper image data. The average value of the classification performance of the model in the data test is for accuracy 96,86%, precision 97,52%, recall 96,83%, Fscore 96,57% in digital photo, and values of accuracy 98,57%, precision 99%, recall 99%, Fscore 98,67% in ink-on-paper images.

Item Type:	Thesis (Undergraduate)
Uncontrolled Keywords:	Convolutional Neural Network, Kinerja Klasifikasi, Muzzle Pattern, Random Forest, Scale Invariant Feature Transform, Classification Performance
Subjects:	Q Science > Q Science (General) > Q337.5 Pattern recognition systems Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Rofiqul Ilmi
Date Deposited:	01 Sep 2021 03:02
Last Modified:	01 Sep 2021 03:02
URI:	http://repository.its.ac.id/id/eprint/91321

Actions (login required)

View Item