期刊名称：International Journal of Advances in Engineering and Management
电子版ISSN：2395-5252
出版年度：2020
卷号：2
期号：2
页码：703-714
语种：English
出版社：IJAEM JOURNAL
摘要：One of the major concerns throughout the world in all the places of large gatherings during an event is crowd control. The event can be of any form ranging from gatherings of few hundreds to millions. When large number of people assemble or move in an area which might not be suitable to handle the number it can create congestion which would lead to unexpected events such as stampede, riots or emergencies in an unfavourable situation. It would be extremely time consuming and hectic for manual monitoring. Henceforth an attempt is made to build an automated system that would count the number of people assembled and compare it with the thresholds provided by the security management and also would try to identify physical violence in the region of interest. This system would also help in maintaining law and order during protests, curfew and imposed social distancing. These would alert the security personnel to take necessary actions. The system is built to estimate crowd and detect violence based on the street camera‟s feed which are usually positioned at the strategic places of crowd gatherings. The model is built using Convolutional Neural Network (CNN). All the results which are provided by this system is based on image processing. A real time alert system in practicality has it‟s own challenges and limitations. Therefore, this paper focuses on the model which could help the system perform better.
