Features Extraction of Capsicum Frutescens (C.F) NDVI Values using Image Processing

Abstract

There is yet an application for monitoring plant condition using the Normalized Difference Vegetation Index (NDVI) method for Capsicum Frutescens (C.F) or chili. This study was carried out in three phases, where the first and second phases are to create NDVI images and recognize and extract features from NDVI images. The last stage is to assess the efficiency of Neural Network (N.N.), Naïve Bayes (N.B.), and Logistic Regression (L.R.) models on the classification of chili plant health. The images of the chili plant will be captured using two types of cameras, which can be differentiated by whether or not they have an infrared filter. The images were collected to create datasets, and the NDVI images' features were extracted. The 120 NDVI images of the chili plant were divided into training and test datasets, with 70.0% training and 30.0% test. The extracted data was used to test the classification accuracy of classifiers on datasets. Finally, the N.N. model was found to have the highest classification accuracy, with 96.4 % on the training dataset and 88.9 % on the test dataset. The state of the chili plant can be predicted based on feature extraction from NDVI images by the end of the study.