A Novel ACO Based Land Cover Classification Approach Using Optical and SAR Data

Abstract

 Abstract—The land cover classification methods based on statistical theory using remote sensing data have great achievements for the last several decades, but they have exposed some weaknesses in dealing with multi-source and multi-dimensional data. Ant colony optimization (ACO), as an excellent intelligent algorithm, has been applied on many research fields for solving optimization issues. How to solve the optimization problem of sampling data is the key step in process of land cover classification with multi-source and multi-dimensional data, so ACO algorithm has many potential advantages in the field of remote sensing data processing. In this paper, an intelligent method is developed for classifying the land cover types combining Landsat TM data with Envisat ASAR data on the basis of ACO algorithm. For identifying the classification precision of land cover with ACO algorithm, we respectively compare it with the results by Maximum Likelihood Classification (MLC) and decision tree C4.5. The comparing results show that ACO algorithm can well take advantages of optical and radar data to improve classification precision and select the best features subsets to construct simpler rules.