Current Issue Cover

袁翔,程塨,李戈,戴威,尹文昕,冯瑛超,姚西文,黄钟泠,孙显,韩军伟(Northwestern Polytechnical University;西北工业大学;中国科学院空天信息创新研究院)

摘 要
Progress in small object detection for remote sensing images

yuanxiang,chenggong,lige,daiwei,yinwenxin,fengyingchao,yaoxiwen,huangzhongling,sunxian,hanjunwei(Northwestern Polytechnical University)

Remote sensing images comprise a mass of objects with limited sizes owing to the particular shooting views and various altitudes, and precisely detecting these small objects plays a crucial role in developing an intelligent interpretation system for remote sensing images. Focusing on the remote sensing images, this paper conducts a comprehensive survey for deep learning-based Small Object Detection (SOD). Firstly, based on the intrinsic properties of small objects, three major challenges to small object detection in remote sensing images are concluded: the bottleneck of representation, the confusion between objects and backgrounds and the sensitivity of regression branch. What is more, this paper extensively reviews the literature of small object detection for remote sensing images in the deep-learning era, then a comprehensive survey in this field is provided. Specifically, by systematically reviewing corresponding methods of three SOD tasks, i.e., SOD for optical remote sensing images, SOD for Synthetic Aperture Radar (SAR) images and SOD for infrared images, an understandable taxonomy of the reviewed algorithms for each task is given. In addition, we retrospect several publicly available datasets for three SOD tasks. Meanwhile, on top of three representative datasets in corresponding domains, namely SODA-A, AIR-SARShip and NUAA-SIRST, we undertake an in-depth evaluation and comparison of main-stream SOD methods for remote sensing images. Afterwards, we discuss the status in applications of small object detection for remote sensing images. In the end, promising directions for SOD in remote sensing images are provided to enlighten the future research.