A Novel Method Based on Lesion Expansion to Assess Plant Disease Severity

Feng Qin Haiguang Wang ^* Jiang Qian Hongli Wang

• China Agricultural University, Beijing, China

Severity is a key indicator utilized in plant disease monitoring and pathogen-plant interaction phenotyping. To more accurately and quickly assess the severity of plant diseases for which the lesion area ratio of an investigated plant unit at each severity class in the corresponding severity grading standard is not the actual ratio of the lesion area to the area of the whole investigated plant unit, a plant disease severity assessment method based on lesion expansion was proposed in this study. By taking wheat stripe rust caused by Puccinia striiformis f. sp. tritici as an example, after image segmentation operations of the single diseased wheat leaves of wheat stripe rust, lesion expansion processing was carried out by using nine method combinations with three proposed lesion expansion methods and three proposed lesion expansion coefficient determination methods, and then the severity assessments of the single diseased wheat leaves were conducted. The results showed that the accuracy of severity assessments of the single diseased wheat leaves of each severity class was in the range of 78.00%－100.00%. No matter which method was used to determine the lesion expansion coefficient/coefficients, the performance for severity assessments of the single diseased leaves achieved after lesion expansion by using Lesion expansion method 3 (the lesion expansion method based on image scaling algorithm) outperformed that achieved after lesion expansion by using the other two lesion expansion methods. The performance of the method combination with Lesion expansion method 3 and Lesion-expansion-coefficient determination method 1 with the lesion expansion coefficient of 2.74, achieving an accuracy of 96.16% for severity assessments of all the single diseased wheat leaves, was the optimal among the nine method combinations. The results demonstrated that satisfactory severity assessment results could be achieved by using the proposed method based on lesion expansion. The results indicated that the lesion-expansion-based plant disease severity assessment method is feasible, which can be used to solve the severity assessment problem described above. This study provided a new idea and method for the accurate severity assessments of plant diseases, and provided supports for the automatic and intelligent assessments of plant disease severity.