Study of Detecting Corn Plant Leaf Disease with Fuzzy C-Means and RNN

Abstract

Corn is a top commodity after rice at supporting food self-sufficiency in Indonesia. However, because of corn leaf spot disease caused by plant pests the quality and quantity of corn is greatly decreasing. The problem with detecting spot-on corn leaf is required high accuracy and the plantation is huge. Therefore, this research study to determine corn leaf spot disease using Fuzzy C-Means and RNN methods. The research process in this study is first preprocessing to CIE-L*A*B* color space, next step is doing corn plant leaf disease detection with Fuzzy C-Means and RNN methods, the third step is to reconstruct the image from Fuzzy C-Means and RNN method result in grayscale level, and the last step is to evaluate the Fuzzy C-Means and RNN algorithm. In this paper only Fuzzy C-Means segmentation and training the RNN model are implemented. The result of the experiment is first from training RNN model with 80% training data and 20% testing data. The data trained for 20 epochs with 38 minutes and 1.8 seconds in total execution time and resulting with 0.9403 for accuracy and 0.9572 for validation accuracy. Next is the Fuzzy C-Means segmentation result, the Fuzzy C-Means execution time is 94 minutes and 16 seconds. For future research the RNN can be trained with much higher epoch and for Fuzzy C-Means can be combined with classification algorithm. We hope that this study can contribute to detecting leaf spot disease for corn plant at faster rate.

Keywords—Fuzzy C-Means, Image Segmentation, RNN, Corn Disease Detection