A novel study on ant-based clustering for paddy rice image classification

Abstract

Paddy rice is the major crop of food in Taiwan. There are three main contributions of rice cultivation on Taiwan: regional eco-friendly of environments, adjustment of floods, and refreshing the air. The estimation of paddy rice crop area is important since this information is related to the national food policy, yearly crop yields calculation, and post-disaster reimbursement. In the past, a large amount of human power and time-consuming works are performed through field exploration and rice pattern revising. A more economic manner to estimate paddy rice area is desired. Accordingly, this study aims to design a paddy rice classifier by which the evaluation of paddy rice area in a remote sensing image can be rationally performed. In this study, the ant-based clustering optimization (ACO) algorithm is used to present image classification on paddy rice mapping. The advantage of this algorithm is that it can effectively cluster data into groups through unsupervised learning. The ACO implementation can be organized to the following steps: (a) initialize the number of ants, (b) compute the pheromone trail matrix of ants, (c) update the pheromone trail matrix, (d) select the best few ants that have the optimal object function value, and (e) repeat iterations until termination criteria are satisfied. In the meantime, self-organizing map (SOM) clustering was used as a parallel study for comparison. A classifier for paddy rice pattern discernment was built as well as paddy rice thematic maps are drawn. The result shows that ACO clustering algorithm has a higher accuracy than that of SOM.