摘要
针对视频检测算法中选取合适的图像语义特征提取算法的困难和检测计算效率不高的问题,该文提出了一种基于狄利克雷多项式共轭隐马尔科夫模型的视频异常行为检测算法,首先提出了狄利克雷多项式共轭模型用于抽取视频的底层特征的语义特征,接着将该模型与隐马尔可夫模型相结合进行视频异常行为的检测。通过校园人群流动和交通灯车辆流动实验证明,该方法具有较高的计算效率和检测性能。
For the difficulty that selecting the appropriate image semantic feature in extraction algorithm and the problem of low detection efficiency,the approach of abnormal behavior detection in video based on Dirichlet multinomial conjugate hidden Markov model（ DMC-HMM） is proposed. The Dirichlet Multinomial Conjugate（ DMC） model is first proposed to extract semantic features of the underlying characteristics of video, then combining the DMC model and the hidden Markov model（ HMM） model to detect the abnormal behavior in the video. The experiment results of complex crowd flow scenarios show that this algorithm is not only able to dig out the patterns of the crowd flow behavior,but also can detect abnormal behavior in the scene,which also has a higher computational efficiency and detection performance.
出处
《杭州电子科技大学学报（自然科学版）》
2014年第3期21-24,共4页
Journal of Hangzhou Dianzi University：Natural Sciences
关键词
异常行为检测
光流语义特征抽取
狄利克雷多项式共轭模型
狄利克雷多项式共轭隐马尔科夫模型
abnormal behavior detection
optical flow semantic feature extraction
Dirichlet multinomial conjugate model
Dirichlet multinomial conjugate hidden Markov model
