Optimalisasi Teknik Image Enhancement untuk Klasifikasi Varietas Apel Menggunakan SVM dan CNN

Abstract

One of the largest export commodities in Indonesia is fruit commodities, one of which is apples. Apples have many varieties that differ in shape, color and size, which can cause identification and highlighting of apples to have limitations by requiring manual inspection from experts. This manual inspection is influenced by the expert's ability and experience in assessing the texture, color pattern, smell and characteristics of apples. In addition, the large diversity of apple varieties does not guarantee the completeness and ease of access related to information and data on apple varieties. The availability of this information is very important in supporting increased fruit production and determining superior apple varieties. So, a system is made that can classify apple varieties such as ana apples, manalagi apples, fuji apples, red delicious apples and rome beauty apples automatically. The apple variety classification methods used are SVM and CNN. The accuracy result of the SVM method is 94% based on texture feature parameters. While the CNN accuracy result is 100% Using learning rate 0.001 and epoh 20.