Prediksi Pertumbuhan Tanaman Tomat Berdasarkan Pengambilan Citra oleh Robot Menggunakan Gradient Boosting

Abstract

In conventional farming methods, farmers must be directly involved in monitoring the growth of plants in the field. Any negligence in the monitoring process can hinder the growth of the plants themselves. Soil moisture levels and acidity (pH) are two crucial factors that can affect plant growth. One way to address these issues is by using a plant growth prediction system. This study aims to predict the growth of tomato plant images and tomato fruit images based on soil moisture and soil pH. In the process, the author is assisted by a plant growth monitoring robot equipped with a camera and sensors to periodically capture images of the plants and fruits, as well as soil pH and moisture sensor data. The first stage involves image pre-processing, which includes cropping, gamma correction, and HSV color space. The next stage is feature extraction using length features extraction to measure the area of the plant images and fruit images. The obtained image analysis results will be combined with the soil moisture and pH sensor dataset, which will then be used as input for a Gradient Boosting-based prediction model. This model allows for the prediction of plant growth during the growth period and can be monitored through a desktop application. The accuracy of plant growth prediction in this study is as follows: (a) the model's accuracy for the area of tomato plant images is 88.04%, and (b) the model's accuracy for the area of tomato fruit images is 92%.