Non-Destructive Estimation of Leaf Carbon Content in Cash Crops Using Deep Learning-Based Image Analysis

Accurate estimation of plant carbon content is essential for understanding carbon sequestration, crop productivity, and climate change mitigation. Conventional laboratory-based methods are destructive, time-consuming, and unsuitable for largescale deployment. This study proposes a non-destructive, image-based approach using deep learning for estimating leaf carbon content in economically important crops. A convolutional neural network based on EfficientNetB4 is employed to learn complex visual patterns from RGB leaf images of mango (Mangifera indica) and potato (Solanum tuberosum). The problem is formulated as a regression task, and the model is trained using laboratory-measured carbon values as ground truth. Experimental results demonstrate that the proposed model achieves strong predictive performance with low error metrics, validating the feasibility of image-based carbon estimation. The approach provides a scalable and cost-effective solution for precision agriculture and environmental monitoring.