基于手部轨迹识别的ATM智能视频监控系统

为实时阻止针对自动取款机的犯罪行为发生，设计一种基于手部轨迹识别的ATM智能视频监控系统。对于采集所得的监控区域内的视频图像，利用混合高斯背景建模方法为视频图像建立背景模型，通过背景剪除法和跟踪算法得到监控区域内的人体信息，分析进入监控区域的人体面积变化情况，由此判断是否有异常行为发生，存在异常则报警，否则采用基于颜色空间的皮肤检测算法和位置约束检测人手部分，利用隐马尔可夫模型对分段的手部运动轨迹分别进行匹配识别，进一步判断是否存在犯罪行为。实验结果表明，该方法对于犯罪行为的识别率能达到88%。

Intelligent Video Surveillance System of Automatic Teller Machine Based on Hand Trajectory Recognition

In order to prevent the criminal behaviors for Automatic Teller Machine(ATM) timely,this paper designs an ATM intelligent video surveillance system based on hand trajectory recognition.A background model is built by Gaussian-mixture background modeling method with the video images collected in the monitored area,and the information of human body can be obtained by background subtraction method and tracking algorithm.By analyzing the area change of human body,it can judge whether there exist abnormal behaviors.If there is,it will give an alarm.Otherwise,it will detect the hand trajectory with the skin detection algorithm based on color space and some location constraints,then match and recognize the movement trajectory of hand by building an Hidden Markov Model(HMM) to determine whether there exist other abnormal behaviours further.Experimental results show that the recognition rate for the criminal behaviors of the method proposed can reach up to 88%.