Image-based ANN Modeling for Nitrogen and Chlorophyll Assessment in Red Betel Leaves using Relief F-selected Texture and Colour Features

Accurate and timely detection of chlorophyll and nitrogen content in plants is essential for optimizing nutrient management and crop health monitoring. This study aimed to develop an effective image-based artificial neural network (ANN) model to estimate chlorophyll and nitrogen (N) levels in red betel (Piper crocatum) leaves. Leaf images were captured using standardized digital imaging setups. Color Red Green Blue (RGB) and texture (entropy, homogeneity) features were extracted from these images and analyzed using a carefully optimized ANN model with robust validation techniques to ensure prediction accuracy and generalizability. Results showed that the developed ANN model achieved high accuracy (R² = 0.99), with color intensity and texture parameters strongly correlated to measured chlorophyll and N concentrations. A paired-sample t-test showed no significant difference between ANN-predicted and laboratory Kjeldahl N values (P = 0.12), while SPAD-502 readings differed significantly from the laboratory reference (P < 0.01). Validation errors for chlorophyll and N prediction remained below 4%. These results suggest that an ANN combined with machine vision can provide accurate, real-time, and non-destructive monitoring of plant nutrient status.