Design of an Automated System for Finger Millet Disease Detection and Prediction Using Multimodal Fusion and Dynamic Graph Attention Networks

International Journal of Electrical and Electronics Engineering

© 2025 by SSRG - IJEEE Journal

Volume 12 Issue 4

Year of Publication : 2025

Authors : Shailendra Tiwari, Anita Gehlot, Himani Maheshwari, P. Venkata Ramana Rao, Shailesh Mishra, Rajesh Singh, Sachin Kumar

10.14445/23488379/IJEEE-V12I4P120

How to Cite?

Shailendra Tiwari, Anita Gehlot, Himani Maheshwari, P. Venkata Ramana Rao, Shailesh Mishra, Rajesh Singh, Sachin Kumar, "Design of an Automated System for Finger Millet Disease Detection and Prediction Using Multimodal Fusion and Dynamic Graph Attention Networks," SSRG International Journal of Electrical and Electronics Engineering, vol. 12,  no. 4, pp. 261-267, 2025. Crossref, https://doi.org/10.14445/23488379/IJEEE-V12I4P120

Abstract:

There is a critical need to detect and predict diseases in Finger Millet due to crop yield and quality losses. Traditional methods mostly fail to provide appropriate and timely detection as they are driven by single data sources and can't adapt to the spatial and temporal dynamics of development in complex ways. Towards this, we develop a unified framework for disease detection in Finger Millet leaves that combines 1) multimodal fusion, 2) dynamic graph neural networks, and 3) temporal sequence modeling state-of-the-art techniques. Our overall framework is driven by three main models: the Multimodal Fusion Network with Attention, the Dynamic Graph Attention Network, and the Temporal Fusion Transformer. The MFNA model considers multiple data types, including RGB and multispectral images, which are fed into the model with IoT sensor data. CNN is utilised for feature extraction from images, and fully connected layers are applied to sensor data samples. Afterwards, it applies an attention mechanism to automatically weigh the importance of features from each modality and then applies a fusion layer to integrate these features for robust disease detection. DGAT builds a dynamic graph wherein nodes represent the different parts of the Finger Millet leaf, hence encoding the attributes pertaining to color, texture, and health status. It is inbuilt with self attention mechanisms within the graph that can adjust the importance of nodes and edges by considering factors such as spatial spreads of the disease with temporal updates for evolving patterns of the diseases. The TFT model generates temporal attention to process time-series data from IoT sensors and sequential image data handling long-term dependencies. The recurrent layers, either LSTM or GRU, deal with short-term dependencies, and the outputs are combined using a fusion module for disease progression and severity forecasting. Our framework integrates these models to give a complete solution that encapsulates spatial intricacies, robust feature extraction, and temporal dynamics of disease progression. This approach greatly improves accuracy and robustness in disease detection and prediction, thus allowing timely interventions in crop management. The proposed work will go so far as to revolutionize agriculture technology by rendering precise spatial identification, robust detection, and accurate forecasting of the crop, hence improving health and increasing productivity.

Keywords:

Disease detection, Multimodal fusion, Dynamic graph neural networks, Temporal sequence modeling, Finger millets.

References:

[1] R. Lokeswari, and R. Mahendran, “Effect of Plasma Bubbling on Textural and Engineering Properties of Ready-to-Eat Pearl Millet Flakes and Puffs,” IEEE Transactions on Plasma Science, vol. 50, no. 6, pp. 1423-1429, 2022.
[CrossRef] [Google Scholar] [Publisher Link]
[2] Adduru U.G. Sankararao et al., “Water Stress Detection in Pearl Millet Canopy with Selected Wavebands Using UAV Based Hyperspectral Imaging and Machine Learning,” IEEE Sensors Applications Symposium, Sundsvall, Sweden, pp. 1-6, 2022.
[CrossRef] [Google Scholar] [Publisher Link]
[3] Yingying Ning et al., “The Traceability of Millet Based on Blockchain Smart Contracts in Agricultural Supply Chain,” 2nd International Conference on Artificial Intelligence and Blockchain Technology, Zibo, China, pp. 65-70, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[4] Ibrahima Diack et al., “Combining UAV and Sentinel-2 Imagery for Estimating Millet FCover in a Heterogeneous Agricultural Landscape of Senegal,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 17, pp. 7305-7322, 2024.
[CrossRef] [Google Scholar] [Publisher Link]
[5] M. Birundadevi et al., “A Machine Learning Strategy for Reducing Childhood Obesity Using Millet,” 9th International Conference on Smart Structures and Systems, Chennai, India, pp. 1-6, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[6] B. Rabi Prasad et al., “Physicochemical Characterisation of Lignocellulosic Biomass for the Identification of Potential Candidacy towards Alternative Renewable Energy,” International Conference on Power, Instrumentation, Energy and Control, Aligarh, India, pp. 1-5, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[7] Balaji Tedla et al., “A Deep Learning-Based System for Classification and Quality Evaluation of Seeds,” 7th International Conference on Intelligent Computing and Control Systems, Madurai, India, pp. 117-121, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[8] Gattu Priyanka, P. Rajalakshmi, and Jana Kholova, “Two-Dimensional Histogram based on Relative Entropy Thresholding for Crop Segmentation Using UAV Images,” IEEE International Geoscience and Remote Sensing Symposium, Pasadena, CA, USA, pp. 3518-3521, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[9] Lucas Gruss et al., “Moving Horizon Estimation of Xenon in Pressurized Water Nuclear Reactors Using Variable-Step Integration,” European Control Conference, Bucharest, Romania, pp. 1-6, 2023. [CrossRef] [Google Scholar] [Publisher Link]
[10] Ba Mamadou Seydou, Philippe Bernard Himbane, and Lat Grand Ndiaye, “Determination and Comparison of Combustion Kinetics Parameters of Peanut Shells, Cashew Nut Shells, Palm Nut Shells, and Millet Stem,” IEEE Multi-Conference on Natural and Engineering Sciences for Sahel's Sustainable Development, Ouagadougou, Burkina Faso, pp. 1-7, 2022.
[CrossRef] [Google Scholar] [Publisher Link]
[11] J. Madhuri, and M. Indiramma, “Hybrid Filter and Wrapper Methods based Feature Selection for Crop Recommendation,” International Conference on Electronic Systems and Intelligent Computing, Chennai, India, pp. 247-252, 2022.
[CrossRef] [Google Scholar] [Publisher Link]
[12] Kyle Ngo et al., “Automated Weed Detection System for Bokchoy Using Computer Vision,” IEEE 14th International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment, and Management, Boracay Island, Philippines, pp. 1-6, 2022.
[CrossRef] [Google Scholar] [Publisher Link]
[13] Ramesh Babu N. et al., “Development and Study of Acoustic Properties of Desmostachya Bippinata Reinforced Composite,” IEEE 2nd Mysore Sub Section International Conference, Mysuru, India, pp. 1-5, 2022.
[CrossRef] [Google Scholar] [Publisher Link]
[14] Ali Ahmad et al., “Turning Smartphone Camera into a Fungal Infection Detector for Chickpea Seed Germination,” International Conference on Multimedia Computing, Networking and Applications, Valencia, Spain, pp. 27-32, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[15] Tingxi Yu et al., “Predicting Phenotypes from High-Dimensional Genomes Using Gradient Boosting Decision Trees,” IEEE Access, vol. 10, pp. 48126-48140, 2022.
[CrossRef] [Google Scholar] [Publisher Link]