2021 年 59 巻 2 号 p. 77-85
This study introduces an image processing method capable of performing real-time detection of two common diseases, leaf blast (LB) disease and bacterial blight (BB) disease, in the paddy fields of the Vietnamese Mekong Delta (VMD). The input images were recorded with an RGB camera. The discrimination of the diseases on rice leaves was obtained by an image processing method based on the extraction of texture and color features from disease lesions, in conjunction with either the Gaussian Naïve Bayes classifier or the K-Nearest Neighbors (KNN) algorithm, to classify the disease into various categories. Both methods perform real-time detection of LB and BB disease in the early stages of development with uncontrolled light conditions in rice fields. Our results show that Gaussian Naïve Bayes is simple but effective, with a shorter processing time and higher detection accuracy than KNN.