Use of UAV images to assess narrow brown leaf spot severity in rice

Abstract

Abstract: Unmanned aerial vehicle (UAV) remote sensing is a potential tool to reduce crop yield losses caused by numerous diseases through near real-time detection and monitoring on disease progression. However, limited research has been conducted to effectively integrate this technology into current crop management systems for disease control. In this study, the feasibility of assessing the severity of narrow brown leaf spot (NBLS) in rice based on UAV remote sensing platform was explored. RGB and NIR images were obtained using Sentera Multispectral Double 4K sensor attached to DJI INSPIRE 2 drone flying at two altitudes (10 m and 15 m). Ground-truth data on disease severity were collected through visual assessment of field plots with different levels of disease severity. Five out of 21 vegetation indices have a coefficient of determination ( R 2 ) value greater than 0.8 based on unitary linear regression. The index with the highest R 2 is Excess Green minus Excess Red (ExGR). The results of unitary regression analysis demonstrated more suitability of using RGB images for rice NBLS assessment over NIR images. Further analyses were conducted on disease-infected plot data that were divided into two groups with 2/3 of the plot data as modeling set and the remaining as evaluation set. The ExGR has the highest R 2 value and the lowest RMSE value in both modeling and evaluation sets regardless of drone flight height (10 m or 15 m). The RMSE at 15 m is lower than at 10 m but there was no significant difference of R 2 , thus the 15-m flight height is better than the 10-m height in detecting the levels of disease severity. The comparison of ExGR and HIS-H demonstrated that vegetation index is more suitability for detecting rice NBLS disease with more spectral information. When disease severity data were divided into two score groups (0 to 5 and 6 to 9 for the low and high levels of disease, respectively) or three score groups (0 to 3, 4 to 6 and 7 to 9 for the low, moderate and high levels of disease, respectively), the ExGR was more suitable for the detection of the high levels of disease. These results demonstrated the feasibility of using UAV images as a potential tool to assess the severity of NBLS, an important fungal foliar disease in rice worldwide. Keywords: UAV remote sensing, multispectral sensor, narrow brown leaf spot, disease severity, rice DOI:  10.33440/j.ijpaa.20190202.47.  Citation: Cai N, Zhou X G, Yang Y B, Wang J, Zhang D Y, Hu R J.  Use of UAV images to assess narrow brown leaf spot severity in rice.  Int J Precis Agric Aviat, 2019; 2(2): 38–42.