IGDM: Image-Based Grading System of Downy Mildew in Cucumber Using Digital Image Processing and Unsupervised Learning

Downy mildew, caused by Pseudoperonospora cubensis, causes severe losses in crops, especially cucurbits, by destroying them and making them unfit for consumption. Plant pathologists visually declare the disease’s grade after analysing the detection, gradation, and severity. The gradation leads to inconsistency due to its imprecise declaration. However, the specific range of disease distribution is required for an automated system. Unsupervised learning can help to adopt the total number of grades according to the various stages of the image’s pixel distribution. The statistical analysis depends on healthy or unhealthy pixels of a leaf image that signifies the range of gradation scale. The statistically proven gradation scale depends on the percentage of the affected area of the leaf by the total area of the leaf. The percentage of the damaged leaf area divided by the leaf’s total area determines the grade of the leaf image. The grading method was also confirmed by a number of plant pathologists. This computational method is more reliable and valuable for recognizing the gradation of an affected leaf. The proposed technology has been tested over 1500 leaves with different disease grades and is found to be 89.86% accurate. The instant capture of images and calculation creates a tremendous boon for the farmers and agriculturists in knowing the status of the field and taking action on such an account.