Deep learning for occluded and multi‐scale pedestrian detection: A review

• Published in: IET image processing Vol. 15; no. 2; pp. 286 - 301
• Main Authors: Xiao, Yanqiu, Zhou, Kun, Cui, Guangzhen, Jia, Lianhui, Fang, Zhanpeng, Yang, Xianchao, Xia, Qiongpei
• Format: Journal Article
• Language: English
• Published: Wiley 01.02.2021
• Subjects: Computer vision and image processing techniques; Optical, image and video signal processing; Traffic engineering computing; Video signal processing
• ISSN: 1751-9659; 1751-9667
• DOI: 10.1049/ipr2.12042

Pedestrian detection, as a research hotspot in the field of computer vision, is widely used in many fields, such as automatic driving, video surveillance, robots and so on. In recent years, with the rapid development of deep learning, pedestrian detection technology has made unprecedented breakthroughs. However, it fails to saturate pedestrian detection research, and there are still many problems to be solved. This study reviews the current research status of pedestrian detection methods based on deep learning. In the first place, we summarised the research results of two stage and one stage pedestrian detection based on deep learning, also summarised and analysed the improvement methods. Meanwhile, we focused on the occlusion and multi‐scale problems of pedestrian detection and discussed the corresponding solutions. At last, we induced the pedestrian detection datasets and evaluation methods and prospected the development trend of deep learning in pedestrian detection.