Determination of Common Maize (Zea mays) Disease Detection using Gray-Level Segmentation and Edge-Detection Technique

Abstract

Maize disease has been one of the common problems for farmers in the Philippines as reported by the Bureau of Plant Industry. The usual process done by farmers is they would need to submit a photo of the possible disease they want to check and wait for the Bureau of Plant Industry to validate what kind of disease it is. This usually takes time and the disease would worsen before the validation would be done. The proposed study by the researcher is to determine the status of the Maize if it is healthy or is infected by a common maize disease which are Gray Leaf Spot, Leaf Rust, and Northern Leaf Blight. The study used an image processing technique which is the Gray-Level Segmentation and Edge-Detection Technique for image pre-processing which is processed by TensorFlow and Keras under a python module to train and create the model using Convolutional Neural Network. Using the open-source dataset provided by PlantVillage, a neural network model for the common maize disease stated by the Bureau of Plant Industry has been created. The study used a Raspberry Pi 3B to classify the status of the Maize in question due to the portability of the device. Using the combined image processing technique, the overall accuracy for the detection rate of the system prosed has achieved 92.50% and having the precision rate with 92.50%.