Claim Missing Document
Check
Articles

Found 1 Documents
Search
Journal : Jurnal Teknoinfo

EKSTRAKSI CITRA MENGGUNAKAN METODE LAPLACIAN DAN SVM (SUPPORT VECTOR MACHINE) UNTUK IDENTIFIKASI JENIS TANAMAN PAKU BERDASARKAN CITRA SPORA Sufiatul Maryana; Herfina Herfina; Arie Qurania; Herlinda Herlinda
Jurnal Teknoinfo Vol 17, No 2 (2023): Vol 17, No 2 (2023) : JULI
Publisher : Universitas Teknokrat Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33365/jti.v17i2.2579

Abstract

Spores are a breeding tool for ferns (Pteridophyta) which are generally found under the surface of each leaf. Research on the characteristics of Pteridophyta spores is generally done by observing their size and shape to see the type of spore. The method that will be carried out in this study requires a lot of understanding, experience, accuracy and time to achieve high accuracy in determining the type. Based on these reasons, it is necessary to develop another technique with modeling to help identify the type of Pteridophyta in the spore. This paper discusses image extraction for fern identification based on spore images using the Laplacian method and SVM. Laplacian method is used to extract features from spore images, while SVM is used to classify ferns. This method requires a lot of understanding, experience, accuracy and time to achieve high accuracy in determining the type. Based on these reasons, it is necessary to develop another technique with modeling to help identify the type of Pteridophyta in the spore. The purpose of this research is to make Support Vector Machine (SVM) modeling to identify fern species based on spore images in Eigen space. The data used are spore images with four classes, each of which has 24 data per class with a total of 96 images. The method used in this research is Laplacian method for feature extraction and SVM for data classification using Gaussian RBF and Polynomial kernel function experiments. The result of this research is the accuracy of each extraction that has been tried. The best result lies in the RBF kernel function with 70% feature extraction and 10 parameters with an accuracy of 98.96%.