Detecting Rice Disease Early with Advanced Imaging and AI Technology

Jim Crocker
22nd May, 2024

Detecting Rice Disease Early with Advanced Imaging and AI Technology

Hyperspectral analysis reveals that rice seedlings infected with Fusarium fujikuroi develop a distinct spectral signature of increased reflectance compared to healthy plants (a–c), corresponding to disease-induced chlorosis and reduced chlorophyll content.

Image adapted from: Chen et al. / CC BY (Source)

Key Findings

  • Researchers from Zhejiang University developed a new method to detect Bakanae disease in rice using hyperspectral imaging and deep learning
  • The deep learning model, RBD-VGG, achieved 92.2% accuracy on the 21st day and 79.4% accuracy on the 9th day of infection
  • This method allows for early and accurate detection of the disease, enabling timely intervention and better disease management
Bakanae disease, caused by the fungal pathogen Fusarium fujikuroi, is a significant threat to rice production worldwide. The disease manifests through various symptoms such as elongated and weak stems, slender and yellow leaves, and in severe cases, necrosis of seedlings. Traditional methods for monitoring this disease are labor-intensive and time-consuming, limiting their widespread application. Researchers from Zhejiang University have addressed this challenge by developing a novel approach that combines hyperspectral imaging technology with deep learning algorithms for the in situ detection of rice seedlings infected with Bakanae disease[1]. The study aimed to create an efficient and non-destructive surveillance methodology, enhancing early disease detection and, consequently, disease prevention and control. Hyperspectral imaging captures a broad spectrum of light, providing detailed information about the physiological and biochemical state of the plants. This imaging was performed on the 9th, 15th, and 21st days after infection to monitor the progression of the disease. The researchers developed a deep learning model called Rice Bakanae Disease-Visual Geometry Group (RBD-VGG), which leverages the hyperspectral data to identify infected rice seedlings. The model achieved an impressive accuracy of 92.2% on the 21st day of infection and 79.4% as early as the 9th day. These results demonstrate the model's potential for early and accurate disease detection, which is crucial for timely intervention and management. The study's findings align with previous research on hyperspectral imaging technology for plant disease detection. For instance, a study on wheat seedlings exposed to various herbicide stress levels utilized a similar approach, achieving high accuracy in early stress detection[2]. This underscores the versatility and effectiveness of hyperspectral imaging combined with advanced algorithms in agricultural applications. In addition to the imaging technology, the study also focused on extracting universal characteristic wavelengths. These wavelengths can be used with portable spectral equipment, making field surveillance more feasible and practical. This aspect of the research is particularly relevant for large-scale agricultural operations where rapid and reliable disease monitoring is essential. Bakanae disease has been a persistent problem in rice cultivation, with significant economic impacts due to crop losses. Traditional breeding for disease resistance, as explored in other studies, has shown promise. For example, research on quantitative trait loci (QTLs) governing resistance to Bakanae disease in rice identified specific genetic markers that could be used in breeding programs[3]. However, breeding resistant varieties is a long-term solution and may not address immediate outbreaks. Biological control methods have also been investigated, such as the use of Bacillus oryzicola YC7007, which has shown effectiveness in reducing Bakanae disease severity through induced systemic resistance and antibiotic production[4]. While these methods offer alternative strategies for disease management, they require integration with advanced monitoring techniques for optimal results. The hyperspectral imaging and deep learning approach developed by Zhejiang University researchers provides a complementary tool to these existing strategies. By enabling early and accurate detection of Bakanae disease, this method can facilitate timely interventions, reducing the spread and impact of the disease. In conclusion, the combination of hyperspectral imaging technology and deep learning algorithms represents a significant advancement in agricultural disease monitoring. The RBD-VGG model developed in this study offers a promising solution for the early detection and management of Bakanae disease in rice, contributing to more sustainable and efficient agricultural practices. This research not only builds on previous findings but also sets the stage for future innovations in plant disease surveillance and control.

AgricultureBiotechPlant Science

References

Main Study

1) Early surveillance of rice bakanae disease using deep learning and hyperspectral imaging

Published 21st May, 2024

https://doi.org/10.1007/s42994-024-00169-1

Related Studies

2) Hyperspectral imaging with shallow convolutional neural networks (SCNN) predicts the early herbicide stress in wheat cultivars.

https://doi.org/10.1016/j.jhazmat.2021.126706

3) Mapping quantitative trait loci responsible for resistance to Bakanae disease in rice.

https://doi.org/10.1186/s12284-016-0117-2

4) Biological Control of Rice Bakanae by an Endophytic Bacillus oryzicola YC7007.

https://doi.org/10.5423/PPJ.OA.10.2015.0218


Related Articles

An unhandled error has occurred. Reload 🗙