This study is led by Associate Professor Yufei Liu (College of Biosystems Engineering and Food Science, Zhejiang University, Hangzhou, China). The authors developed a fast and non-destructive early diagnosis method for rice bakanae diseases based on hyperspectral data. 

Bakanae_Disease_sporulation_JIRCAS

Source: JIRCAS Library

A rice stem affected by the bakanae disease at the sporulation stage.

The authors first obtained phenotypic information of infected rice seedlings. The morphological differences caused by fungal infection lead to abnormal plant type proportions in rice seedlings, resulting in thin and weak stems that are prone to breakage, thereby affecting further normal growth.

After being infected, the seedling height, leaf length, internodes, and leaf angle of rice seedlings will increase. At the same time, the accumulation of metabolites from the pathogenic fungus Fusarium Fujikuroi severely damages the chlorophyll structure, causing a decrease in chlorophyll content, a significant decrease in stomatal conductance of seedling leaves, and a decrease in transpiration rate and photosynthetic rate.

Earlier monitoring

In this study, non-destructive detection was performed on rice plants infected with early stage of bakanae disease using hyperspectral imaging technology and an improved lightweight deep learning model, named the RBD-VGG model. The average detection accuracy of seedlings infected for 21 days reached 92.2%. In addition, the model can detect the infection status of seedlings after 9 days of infection, with an average accuracy of 79.4%.

By comparison, the detection accuracy of the RBD-VGG model is higher than that of the SVM and CNN models, indicating superiority. Compared to existing research, this method has achieved earlier monitoring of bakanae disease and significantly improved the accuracy of disease identification.

Meanwhile, 20 universally applicable feature bands were proposed based on CARS and SPA algorithms in the study, which can achieve high-precision classification results. In the future, the research can be used for the development of low-cost portable spectrometers, potentially combining with drones for large-scale crop field monitoring.

Topics