Abstract:
This research presents a method to detect the rice field region in images by using digital image processing techniques and artificial neural network. The proposed method was applied to terrestrial images taken in Thailand. In the algorithm, each image was divided into segments. The features of each segment were extracted and fed to the artificial neural network. The features include texture, color, and position. Artificial neural network was then used to classify each segment to determine whether a segment was a region of a rice field or not. A heuristic function was then applied to the classified segments of the image to determine a heuristic value that represented how much these segments were connected. 600 images were processed to find the heuristic values. An appropriate threshold was then determined from heuristic value. Other 9,158 images were tested with the proposed method. The result showed an average of 96.19% accuracy for rice detection and an average of 96.58% accuracy for non-rice detection.