Abstract
Harvest drop in rice because of leaf blast is a vital issue in the country???s food stock and social life where rice is the primary source of food. Epidemics can cause leaf blasts due to weather conditions or environmental transformation. Therefore, early detection of leaf blast is needed to take precautions action to save the harvest. This research presents a new approach for rice leaf blast detection. It seizes colour distribution and shapes to determine the damaging leaf. Two main features: colour and shape, are key points to measure the similarity of an image by comparing the image query and database. The image extraction uses histogram colour throughout the pre-processing phase. The approach will take the dominant colour of leaf. Since this green colour dominated the leaf, the green will be converted from RGB to the HSV domain with 256 range. The shape feature extraction based on morphology closing will calculate the images??? area, diameter, and perimeter. The process is continued by resizing the image and convert into a grayscale mode to apply canny edge detection. The experiment uses 267 images dataset and 74 testing data consisting of 2 categories: blast disease leaf and healthy leaf. The trial results achieve an 85.71% accuracy rate to detect blast disease by colour feature, 71.42% by shape feature, and 85.71% by combined colour -shape features.