基于多条件随机场模型的异常行为检测

Abnormal Activity Detection based on Multiple CRF Ensemble Model

传统的异常行为检测多数是利用单特征建模,检测的行为较为单一,检测率较低,针对这些问题,提出一种基于多条件随机场模型(MCRF)的异常行为检测方法,MCRF模型具有融合多特征和联系上下文信息的优势。通过Kinect获取3D骨架数据,提取角度、位置、速度三类特征,形成多类特征子集,利用基本的CRF模型对每一类特征子集建模,形成多个CRF单元,然后组合所有的CRF单元,得到MCRF模型,最后利用MCRF模型进行异常行为检测。实验结果表明基于MCRF的异常行为检测方法具有较高的检测率。

Most of traditional abnormal activity detection is modeling by a single feature, so the testing activity is simple, and the detection accuracy rate is relatively low. For these problems, abnormal activity detection based on multiple CRF ensemble model is proposed. The advantage of MCRF model is the ability of combining more features and utilizing adaptive contextual information. There are three features： angle feature, position feature, speed feature by the 3 D skeleton data of Kinect. And several features subsets can be formed through more features extraction. Then CRF model is used for each feature subset and to get CRF units. Finally, all the CRF units are combined to produce MCRF model which is utilized to detect abnormal activity. The experimental results indicate that the detection accuracy rate of this method is better.