Evaluation of efficacy of fungicides for control of wheat fusarium head blight based on digital imaging

Authors: ZHANG, DONGYAN - Anhui Agricultural University; WANG, ZHICUN - Anhui Agricultural University; JIN, NING - Anhui Agricultural University; GU, CHUNYAN - Anhui Agricultural University; CHEN, YU - Anhui Agricultural University; Huang, Yanbo

Submitted to: IEEE Access
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/3/2020
Publication Date: 6/11/2020
Citation: Zhang, D., Wang, Z., Jin, N., Gu, C., Chen, Y., Huang, Y. 2020. Evaluation of efficacy of fungicides for control of wheat fusarium head blight based on digital imaging. IEEE Access. 8:109876-109890. https://doi.org/10.1109/ACCESS.2020.3001652.
DOI: https://doi.org/10.1109/ACCESS.2020.3001652

Interpretive Summary: Fusarium head blight (FHB) is one of the important diseases that endanger the healthy growth of wheat. It is difficult to predict the extent of disease in each wheat ear group suffering from FHB and it is time-consuming and labor-intensive to evaluate the effect of pesticide spraying for crop protection. Scientists of Anhui University in China and USDA-ARS Crop Production Systems Research Unit at Stoneville MS have collaboratively proposed a method for evaluating the effects of pesticide spray application based on machine learning image segmentation and image analysis algorithms. The results indicated that the newly developed algorithms can meet the actual needs of evaluating the effect of pesticide spraying on wheat ear groups infected by FHB, which can provide technical support for dynamic monitoring of FHB based on sensor networking in crop fields.

Technical Abstract: Fusarium head blight (FHB) is one of the important diseases that endanger the healthy growth of wheat. It occurs in major wheat production countries in the world, and it has seriously threatened the global grain production and human health. As it is difficult to predict the extent of disease in each wheat ear group suffering from FHB and it is time-consuming and labor-intensive to evaluate the effect of pesticide spraying for crop protection, this study proposes a method for evaluating the effects of pesticide spray application based on machine learning image segmentation and width mutation counting algorithms. In data process and analysis, firstly, images were processed and transformed from red-green-blue space to red-green-green space. Secondly, the transformed color space image data were run through K-means clustering for rough segmentation of wheat-ears from the images; Thirdly, random forest classifier were used with new features for fine segmentation of wheat-ears from the images to further segment disease spots from the wheat-ear segmentation images. Then, the width mutation counting algorithm was developed and used to count wheat ears. Lastly, by grading the disease severity of the wheat ears groups, the effect of pesticide spray application was evaluated. The experimental results show that: 1) the proposed segmentation algorithm can segment wheat ears from a complex field background, with the average segmentation accuracy over 90%. 2) the proposed counting algorithm can effectively solve the problems of wheat ear adhesion and occlusion. The average counting accuracies of all and diseased wheat ears were 95.5% and 94.85% respectively, with the coefficients of determination of 0.9026 and 0.9925, respectively. 3) the newly constructed normalized green-red-green-blue differential vegetation index has a good segmentation effect on wheat ears. 4) the algorithms proposed and developed in this research are robust to different brightness conditions. 5) the average wheat ear disease grading prediction accuracy of the disease severity of the three wheat ears groups reaches 100%. In summary, the results of this research can meet the actual needs of evaluating the effect of pesticide spraying on wheat ears groups infected by FHB, and can provide technical support for dynamic monitoring FHB based on sensor networking in crop fields.