Crowd Analysis in Video Surveillance: A Review

Abstract

Crowd behavior investigation in images/videos is an important task applied in areas such as people counting, density estimation, emotion recognition, motion detection, and flow analysis, etc. The researchers devoted an excellent quality of work to deal with public issues such as crowd control, traffic monitoring, urban planning, vehicle counting in real-time; however, humanity did not get much success in handling these issues due to the limited cost of energy and time. For evaluation metrics, we need a year-wise analysis of used datasets, publications methodologies, and their performance, which is expected to yield good predictions and conclusions. Therefore, in this work, we have systematically and comprehensively revisited five year studies that conducted crowd analysis in video using deep learning techniques to make more effective research development and progress. We have got some new future directions from some of the prestigious survey works, which is a novel aspect of this study, that would provide potential and reliable solutions for investigating crowd behaviour in videos.