Pengenalan Ras Kucing Menggunakan Metode Ekstraksi Ciri Pyramid Histogram Of Oriented Gradients (p-hog)

Authors

  • Atika Ayunda Murwanti Telkom University
  • Kurniawan Nur Ramadhani Telkom University
  • Prasti Eko Yunanto Telkom University

Abstract

Abstrak Kucing merupakan hewan dengan macam ras yang cukup banyak. Dari banyak ras kucing, beberapa memiliki kemiripan ciri visual yang cukup tinggi seperti ras Siamese, Ragdoll, dan Siamese. Untuk mempermudah pengenalan kedua ras tersebut, akan dilakukan pengenalan menggunakan metode ekstraksi ciri Pyramid Histogram of Oriented Gradients (P-HOG) berdasarkan 599 Oxford IIIT Pet Dataset berupa citra ras kucing Siamese, Ragdoll, dan Birman yang telah dioptimasi. Citra akan melewati tahap pra-proses untuk mempermudah proses ekstraksi ciri dengan P-HOG. Tahap praproses meliputi pencarian ROI menggunakan segmentasi grabcut dan proses grayscale. Pengujian yang dilakukan menghasilkan akurasi tertinggi sebesar 69.2% menggunakan metode P-HOG + SVM dengan kernel linear dan parameter P-HOG level 3 bin 10. Kata kunci : Pengenalan ras kucing, histogram of oriented gradients, P-HOG, computer vision Abstract Cats are animals with quite a lot of breeds. Of the many cat breeds, some have quite high visual similarities such as the Siamese, Ragdoll, and Siamese breeds. To facilitate the introduction of the two races, a recognition introduction will be made using the Pyramid Histogram of Oriented Gradients (P-HOG) feature extraction method based on optimized 599 Oxford IIIT Pet Datasets in the form of images of Siamese, Ragdoll, and Siamese cats. The image will go through the preprocessing stage to simplify the feature extraction process with P-HOG. The preprocessing stage includes ROI search using grabcut segmentation, and grayscale process. Tests carried out resulted in the best accuracy of 69.2% using the P-HOG + SVM method with linear kernel and P-HOG parameters on level 3 bin 10. Keywords: cat breeds recognition, histogram of oriented gradients, P-HOG, computer vision

Downloads

Published

2020-12-01

Issue

Section

Program Studi S1 Informatika