拥挤场景下的暴力行为检测

近年来暴力行为检测受到了越来越多的关注,而在图像研究领域中,这仍然是一个充满挑战的话题。尝试对现有的暴力行为检测器进行改进,提出了一个全新的特征构造方法,称为加速度暴力流(Acceleration Violent Flow,简称AViF),它充分利用光流信息中的加速度信息,可以用来检测拥挤场景下的暴力行为;之后采用了Bagging策略,取得了良好的效果。实验结果表明,在bagging策略下,组合特征取得了高于ViF特征的效果。

Violence Detection in Crowded Scene

Nowadays,more and more attention has been paid to the detection of violence.In the field of image research,this is still a challenging topic.So this paper try to improve the existing violent behavior detector,and proposes a new feature construction method,called Acceleration Violent Flow(AViF),which makes full use of acceleration information from optical flow information and can be used to detect violent behavior in crowded scenes.Then bagging strategy is adopted which achieved good results.Experiment results show that under the Bagging strategy the combination of features has achieved higher results than ViF features.