Hybrid Deep Learning Algorithms for Predicting Nutrient Deficiencies in Paddy Crops using CNN and Super Resolution Generative Adversarial Neural Networks

Abstract

In the realm of agriculture, predicting and addressing nutrient deficiencies in paddy crops is pivotal for sustaining crop yield and ensuring global food security. Farmers face challenges due to limited high-resolution images, impacting the effectiveness of models in deficiency detection. This research introduces a hybrid approach, merging Super-Resolution Generative Adversarial Networks (SRGANs) and Convolutional Neural Networks (CNNs), to elevate image resolution and enhance nutrient deficiency detection efficiency in paddy crops. SRGANs generate synthetic nutrient-deficient crop images, augmenting the training dataset and refining model generalization. These images, with increased detail, complement image analysis techniques for precise deficiency identification. High-resolution outputs from SRGANs serve as improved inputs for CNNs, facilitating accurate classification and localization of deficiencies based on color, texture, and morphology patterns. Synthetic images enable the hybrid model to learn comprehensive nutrient deficiency representations; enhancing detection accuracy. Extensive experiments on a large-scale dataset with varying deficiency levels validate the efficacy of the hybrid approach in real-world scenarios. SRGANs and CNNs, trained and fine-tuned on this dataset, exhibit improved image quality, as measured by metrics like Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM). CNN models demonstrate heightened accuracy in detecting nutrition deficiencies in high-resolution images, showcasing the potential of this hybrid solution for robust nutrient deficiency prediction in paddy crops.