Klasifikasi Tanah Berdasarkan Jenis Tanaman Menggunakan Convolutional Neural Network Di Pusat Pengembangan Sumber Daya Manusia Regional Makassar Indonesia

Abstract

Soil is an important element in plant growth and food production. Soil fertility is a key factor in determining agricultural potential. Today, AI and image processing technologies have presented new opportunities for more in-depth soil analysis. This study aims to develop a fertile soil classification system using the Convolutional Neural Networks method based on image analysis. This research began by collecting datasets in the form of images of soil samples of various types and conditions. This dataset is annotated and used as training data to train Convolutional Neural Networks models. This training process allows the model to identify  visual features that differentiate fertile and infertile soils. This method is implemented in an AI system that is capable of detecting soil fertility through photographs. The results showed that the CNN model succeeded in classifying argosol (fertile) and lateritic (infertile) soils with high accuracy, reaching 99-100%. The developed system also allows the use of a webcam to detect soil fertility directly. The information displayed is accuracy information, webcam display, description of the name of the land, the percentage of trust, and the meaning of the related land.