Abnormal Behavior Analysis Based on Examination Surveillance Video
Miaomiao Ding, Jiahui Zhao, Fangyu Hu · 2016
It's difficult to review a large number of examination surveillance videos at the same time. To reduce the workload of censors, we propose a method to analyze abnormal behaviors during examinations. This paper mainly includes two parts. Firstly, it employs a two-layer classifier with Histograms of Oriented Gradients feature to detect head-shoulder part of examinees, thus we can report the presence of examinees. Secondly, it exploits a method based on sparse combination algorithm to detect suspicious cheating behaviors. Experiments show that our method can achieve decent performance with high efficiency.