PLANT NUTRIENT DEFICIENCY DETECTION FROM LEAF IMAGES USING AI/ML - DRIVEN ENHANCED CHANNEL BOOSTED CONVOLUTIONAL NEURAL NETWORK

Abstract

An early prediction and detection of nutrient deficiency empower farmers to appropriately categorise and apply essential nutrient supplements on time. This work presents a novel methodology built on Transfer Learning (TL) with a Convolutional Neural Network (CNN) to offer enhanced accuracy in the early detection of nutrient deficiency using leaf patterns and colour through an Enhanced Channel Boosted - Convolutional Neural Network (CB-CNN). Leaf features are extracted using Oriented FAST and Rotated BRIEF (ORB) before processing by the proposed CB-CNN. The present work precisely forecasts the type of nutrient deficiency from the leaf images, leaf pattern and leaf shape. It is observed that experimental results show 99.37% prediction accuracy over conventional neural network models. Additionally, there is considerable improvement in other performance metrics, viz., precision, specificity, sensitivity and F-score. The proposed methodology beats its existing counterparts by magnitudes ranging from 1.17% to 10.27%. It is thus clinched that the proposed model outperforms existing neural network models with the highest precision and accuracy.