Machine Learning Based Crowd Behaviour Analysis and Prediction
Keywords:
Anomalous crowd prediction, Crowd behaviour analysis, Deep learningAbstract
Crowd behaviour has been widely known to have the ability to forecast the events a crowd could create. Crowd management can become extremely efficient if situations such as riots, mob lynching, traffic jams, accidents, stampede, etc. could be predicted beforehand. To this end many researchers have made their contributions in the past and there is still immense work being carried out currently. All the researches worked with different algorithms and techniques to analyze images or videos of crowd scenes for counting the number of people in the crowd, predicting the behaviour of the crowd and classifying an image or video as normal or abnormal crowd event. This paper, hence, is directed towards underlining the some of the major researches in this field, the approaches and algorithms adopted by them and their comparisons. Overall, this paper reviews the past researches and presents a summary of the techniques and strategies employed. At the end of this paper is the future scope of work possible in the field of crowd behaviour analysis, prediction and crowd counting.