Review on Land Use Classification

Abstract

Land use classification covers a wide range of applications from general land cover determination to specific crop type detection. Many approaches have been developed to infer land usage from satellite and aerial images. Our multispectral satellite image understanding system starts with land use classification. Therefore, it is reasonable to analyze the existing literature on this problem first. To do so, we investigate trends in land use classification between years 1967 and 2002 by reviewing the related literature. We consider the seminal work of Fu et al. 1969 to be the beginning of automated land use classification. We did not attempt to cover the whole literature; however, we tried to explore a significant and influential portion of it. Specifically, we focused on feature extraction methods using passive sensors and excluded work on classifiers, neural networks, and fuzzy logic. To investigate the trends in solving this problem, we reviewed over 90 influential papers published in refereed journals. To identify trends, we grouped papers based on their major contribution. We grouped these papers according to their major contribution. One paper could belong to many groups, but we chose to assign each paper to only one “best” group. We summarize key papers in each group and tabulate each study by the type of image used, geographical location considered, and the average performance obtained. To clarify image types used in the literature, we include a section summarizing remote sensing satellites and airborne equipment.