Remote sensing image scene classification is a fundamental task in remote sensing image interpretation, which has important application value in land-use investigation, geologic disaster monitoring, geospatial intelligence acquisition and so on. This paper systematically summarizes and analyzes representative methods for remote sensing image scene classification. Firstly, the commonly used datasets for remote sensing image scene classification are sorted out, and the challenges and difficulties brought by the characteristics of remote sensing images to scene recognition are elaborated. Secondly, some typical remote sensing image scene classification methods are summarized, which includes hand-crafted features and deep learning. And the optimization and improvement for remote sensing image scene classification task are analyzed. Then, the performance of some mainstream methods is compared. Finally, the unsolved problems and the next research directions of remote sensing image scene classification are summarized and explored. And the research prospects of high-precision fine-grained scene classification task, high-precision lightweight algorithms, few-shot learning, and large model for remote sensing image scene interpretation are explored with the aim of promoting the remote sensing image scene classification and recognition tasks to achieve more in-depth research and a wide range of applications.