Klasifikasi Jenis Kayu Menggunakan Support Vector Machine Berdasarkan Ciri Tekstur Local Binary Pattern

Abstract

metode LBP ini adalah mean, standar deviasi, skewness, energi, dan entropi. Data citra jenis kayu yang digunakan dalam penelitian ini adalah jenis kayu bayur, cempaka, damar, meranti, dan merbau. Citra kayu tersebut diambil secara manual menggunakan kamera digital dengan jarak pengambilan 20 cm. Hasil akurasi klasifikasi terhadap citra jenis kayu bayur, cempaka, damar, meranti, dan merbau dalam penelitian ini dengan jarak ketetanggaan R=1 adalah sebesar 91,3% berada pada parameter sigma 0,3. Sedangkan hasil error terkecil hasil klasifikasi adalah sebesar 8,7%. Abstract Indonesia is a country rich in natural resources, one of which is timber. Wood and wood products are the leading export commodities. Given the many types of wood that have almost the same texture, understanding is needed to recognize them. Currently, the identification of wood species is generally still done by humans visually. The ability to identify wood species must be done repeatedly and requires a long training process. The limited ability of humans to identify unskilled wood species visually sometimes affects the desired results for the industrial world. Currently, digital image processing technology has been widely used to classify wood types based on their texture. In this study, the local binary pattern (LBP) method was used to classify wood species. This method will produce texture features that will be used as input in the training and testing process using a support vector machine (SVM). Texture features used in this LBP method are the mean, standard deviation, skewness, energy, and entropy. Image data of wood species used in this study were bayur, cempaka, damar, meranti and merbau. The wood image was taken manually using a digital camera with a distance of 20 cm. The results of the classification accuracy of the image types of bayur wood, cempaka, damar, meranti, and merbau in this study with a neighbouring distance of R = 1 were 91.3% in the sigma parameter of 0.3. Meanwhile, the smallest error in the classification results was 8.7%.