Scene classification of remote sensing images is a hot research field because of its wide range of applications. It aims to quickly and accurately identify and locate many objects of predefined categories in a given image scene. Here we use predefined and custom models to recognize an image scene from a given video stream and image. This article details representative algorithmsfor each step. It then introduces public and specific datasets commonly used for target detection, and analyzes and compares various representative algorithms in this area. Finally, potential target acquisition issues are considered.

object detection and recognition, deeplearning, classification performance