Image Matching Comparison Using the K-Nearest Neighbor (KNN) Method and Support Vector Machine (SVM)


View Publication Info
Field Value
Title Image Matching Comparison Using the K-Nearest Neighbor (KNN) Method and Support Vector Machine (SVM)
A Komparasi Image Matching Menggunakan Metode K-Nearest Neightbor (KNN) dan Support Vector Machine (SVM)
Creator Umar, Rusydi
Riadi, Imam
Faroek, Dewi Astria
Description Image matching is the process of finding digital images that have a degree of similarity. matching images using the classification method. In measuring image matching, the images used are original logo images and manipulated logo images. Comparison of classification algorithms from the two methods namely K-Nearest Neighbor (KNN) and Support Vector Machine with Sequential Minimal Optimization (SMO) optimization used to calculate matches based on accuracy values. The K-Nearest Neighbor (KNN) classification method is based on proximity or K calculations while the Support Vector Machine (SVM) classification method measures the distance between the hyperplane and the nearest data. Image match values are measured by Precision, Recall, F1-Score, and Accuracy. The image matching steps start from the preparation of data processing, extraction of HSV color features and shapes, then the classification stage. Digital images are used as many as 10 images consisting of one original logo and 9 manipulated logos. In the classification testing stage, using the WEKA application by applying the 10-fold cross-validation method. From the results of tests conducted that the closest k-neighbor (KNN) classification method is 80% and has a k = 0.889 which is quite good in measuring proximity, while the SVM classification method is 70%. The results of this image matching comparison can be concluded that the K-Nearest Neighbor classification method works better than SVM for image matching.
Pencocokan gambar adalah proses menemukan gambar digital yang memiliki tingkat kesamaan. mencocokkan gambar menggunakan metode klasifikasi. Dalam mengukur pencocokan gambar, gambar yang digunakan adalah gambar logo asli dan gambar logo hasil manipulasi. Perbandingan algoritma klasifikasi dari dua metode yaitu K-Nearest Neighbor (KNN) dan Support Vector Machine dengan optimasi Sequential Minimal Optimization (SMO) yang digunakan untuk menghitung kecocokan berdasarkan nilai akurasi. Metode klasifikasi K-Nearest Neighbor (KNN) didasarkan pada kedekatan atau perhitungan K sedangkan metode klasifikasi Support Vector Machine (SVM) mengukur jarak antara hyperplane dan data terdekat. Nilai kecocokan gambar diukur dengan Precision, Recall, F1-Score, dan Accuracy. Langkah-langkah pencocokan gambar mulai dari persiapan pemrosesan data, ekstraksi fitur dan bentuk warna HSV, kemudian tahap klasifikasi. Gambar digital digunakan sebanyak 10 gambar yang terdiri dari satu logo asli dan 9 logo yang dimanipulasi. Pada tahap pengujian klasifikasi, menggunakan aplikasi WEKA dengan menerapkan metode validasi silang 10 kali lipat. Dari hasil tes yang dilakukan bahwa metode klasifikasi k-neighbor (KNN) terdekat adalah 80% dan memiliki k = 0,889 yang cukup baik dalam mengukur kedekatan, sedangkan metode klasifikasi SVM adalah 70%. Hasil perbandingan pencocokan gambar ini dapat disimpulkan bahwa metode klasifikasi K-Nearest Neighbor bekerja lebih baik daripada SVM untuk pencocokan gambar.
Publisher Politeknik Negeri Batam
Date 2020-10-26
Type info:eu-repo/semantics/article
Format application/pdf
Source Journal of Applied Informatics and Computing; Vol 4 No 2 (2020): Desember 2020; 124-131
Language eng
Rights Copyright (c) 2020 Dewi Astria Faroek

Contact Us

The PKP Index is an initiative of the Public Knowledge Project.

For PKP Publishing Services please use the PKP|PS contact form.

For support with PKP software we encourage users to consult our wiki for documentation and search our support forums.

For any other correspondence feel free to contact us using the PKP contact form.

Find Us


Copyright © 2015-2018 Simon Fraser University Library