Enhancing Agricultural Yield Forecasting with Deep Convolutional Generative Adversarial Networks and Satellite Data

Ensuring food security amidst growing global population and environmental changes is imperative. This research introduces a pioneering approach that integrates cutting-edge deep learning techniques. Deep Convolutional Generative Adversarial Networks (DCGANs) and Convolutional Neural Networks (CNNs)...

Full description

Saved in:
Bibliographic Details
Published in:International journal of advanced computer science & applications 2024, Vol.15 (2)
Main Authors: Anuradha, D., Kuchipudi, Ramu, Ashreetha, B, Ramesh, Janjhyam Venkata Naga, Rami, Ayadi
Format: Article
Language:English
Subjects:
Agricultural production
Artificial neural networks
Crop growth
Data augmentation
Datasets
Food security
Forecasting
Generative adversarial networks
Image enhancement
Image resolution
Machine learning
Mathematical models
Neural networks
Satellite imagery
Synthetic data
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Ensuring food security amidst growing global population and environmental changes is imperative. This research introduces a pioneering approach that integrates cutting-edge deep learning techniques. Deep Convolutional Generative Adversarial Networks (DCGANs) and Convolutional Neural Networks (CNNs) with high-resolution satellite imagery to optimize agricultural yield prediction. The model leverages DCGANs to generate synthetic satellite images resembling real agricultural settings, enriching the dataset for training a CNN-based yield estimation model alongside actual satellite data. DCGANs facilitate data augmentation, enhancing the model's generalization across diverse environmental and seasonal scenarios. Extensive experiments with multi-temporal and multi-spectral satellite image datasets validate the proposed method's effectiveness. Trained CNN adeptly discerns intricate patterns related to crop growth phases, health, and yield potential. Leveraging Python software, the study confirms that integrating DCGANs significantly enhances agricultural production forecasting compared to conventional CNN-based approaches. Against established optimization methods like RCNN, YOLOv3, Deep CNN, and Two Stage Neural Networks, the proposed DCGAN-CNN fusion achieves 98.6% accuracy, a 3.62% improvement. Synthetic images augment model resilience by exposing it to varied situations and enhancing adaptability to diverse geographic regions and climatic shifts. Moreover, the research delves into CNN model interpretability, elucidating learnt features and their correlation with yield-related factors. This paradigm promises to advance agricultural output projections, advocate sustainable farming, and aid policymakers in addressing global food security amidst evolving environmental challenges.
ISSN:2158-107X
2156-5570
DOI:10.14569/IJACSA.2024.0150269