Development of Mobile Phone Application for Characterization of Sugarcane Reaction to Red Rot Disease in Breeding Program

Abstract

The selection of resistant sugarcane varieties depends on testing the cuttings' reaction to the causal fungus, followed by visually scoring the severity of symptoms, which may yield inaccurate results. The visual scoring method may yield inaccurate results. This research addresses this limitation by introducing the Redrot Application for Android phones, version 5 and above. The application utilizes image processing algorithms to assess the reaction of 200 sugarcane varieties to red rot disease. Disease reactions were characterized based on visual assessments by experts, considering three parts: the distance of fungus move through nodes, the size of the red area, and the appearance of red color. The collected data were used to calculate symptom level scores and group resistance reaction characteristics. The program reads and analyzes image results using Python and the OpenCV2 library, incorporating traditional feature extraction, geometry feature extraction, and image segmentation for size and wound detection, including processes such as contouring and morphological operations. Color analysis is performed using Histogram analysis. The confusion matrix was employed to evaluate the prediction results, revealing an application accuracy of 92.50 percent, a recall value of 0.65, a precision value of 0.74, and a balanced F1 score of 0.69. The findings indicate that the application enhances accuracy and efficiency, reducing errors associated with human visual result interpretation. The program processes and displays reaction characteristics promptly upon capturing or loading images for each experimental sugarcane cane