Perbandingan Akurasi Metode Convolutional Neural Network (CNN) dan Sobel untuk Klasifikasi Buah Rambutan melalui Pengolahan Citra

Abstract

Abstract
The automatic determination of rambutan fruit ripeness is a crucial step in enhancing efficiency in the agricultural sector.
This study compares two image processing methods, namely Sobel and Convolutional Neural Network (CNN), for the
classification of rambutan fruit ripeness. The Sobel method was used for edge detection with an accuracy of 75.32%,
while CNN was applied to recognize complex visual patterns, achieving an accuracy of 98.09%. The dataset used
consisted of 1,416 rambutan images categorized into four ripeness stages: unripe, semi-ripe, ripe, and rotten. The CNN
model training process was conducted over several epochs, resulting in a training accuracy of 99% and a validation
accuracy of 100%. The results of this study indicate that CNN outperforms Sobel in handling more complex image
classification tasks, with a significant difference in accuracy. These findings provide a valuable contribution to the
development of automatic classification systems to support quality improvement of fruits in the agricultural industr