Machine learning for satellite image classification: A comprehensive review

Classification of satellite images is the most important technique used in remote sensing for the extraction and analysis of satellite information, which consists of grouping the values of the pixels of the image into significant categories. Nowadays, large quantities of medium to high-resolution satellite images are acquired daily. However, the classification of satellite images is mandatory for many fields such as urban planning, military, agriculture, and environmental monitoring. Many researchers are discussing this area, but the sufficient degree is optimal has not yet been reached. The classification of satellite images requires the selection of an appropriate classification method according to the requirements. Existing satellite image classification methods are classified into three main categories according to the features they use: manual feature-based methods, unsupervised feature learning methods, and supervised feature learning methods. Each of these three methods has its own advantages and disadvantages. The objective of this paper is to present an evaluation of publicly available satellite datasets and on the study of different methods used for satellite image classification as well as a brief overview of previous studies proposed in this field.