Stalk Rots Diseases of Corn Classification using Morphology Closing and Convolutional Neural Network

• Published in: 2022 IEEE 8th Information Technology International Seminar (ITIS) pp. 640 - 644
• Main Authors: Setiawan, Wahyudi, Helmi, M. Alauddin, Husni
• Format: Conference Proceeding
• Language: English
• Published: IEEE 19-10-2022
• Subjects: Convolutional Neural Network; Convolutional neural networks; Corn Diseases; Image Classification; Information technology; Morphology; Morphology Closing; Optimization; Seminars; Stalk Rots; Stochastic processes; Training
• DOI: 10.1109/ITIS57155.2022.10010055

Corn diseases need to be a concern. The disorders can attack all parts of the corn, including leaves, stalk rots, and cob. In this article, the focus is on discussing conditions in stalk rots. Data was obtained from corn plantations in Bangkalan, Madura. The system built is an image classification of three classes (healthy, anthracnose, and gibberella). The data consists of 335 images for each class, so the total number of images is 1005. The system comprises preprocessing (HSV conversion, thresholding, morphology closing, masking) and classification with Convolutional Neural Network (CNN). The CNN layer arrangement has six convolutional layers, three pooling layers, a fully connected layer, and a softmax for the classification layer. The training process is carried out using the epochs 30, batch-size 10, and learning rate 0.0001 with Stochastic Gradient Optimization (SGD). The experiment using 5-fold cross-validation and split data training vs. testing is 80%:20%. The result shows an average accuracy of 93.13%. The best model is obtained in the second fold with an accuracy of 96.51%.