基于全方位计算机视觉的ATM机智能监控

ATM intelligent surveillance based on ODVS

探讨了利用全方位视觉传感器(ODVS)以及计算机视觉技术来实现对自动取款机(ATM)的视频智能监控,采用ODVS来获取整个ATM机周围的全景视频图像,通过透视算法将全景视频图像展开成若干个重点监控领域;为了提高检索效率、减少存储和传输信息量,提出了通过基于高斯肤色模型的人脸检测算法获取ATM使用者的人脸图像并将时间地点等信息进行合并,得到一幅包含使用ATM时间、地点和人物图像;为了提高ATM的使用安全性,通过基于卡尔曼滤波的人脸跟踪和行为语义规则判定等手段来检测和分析ATM设备周围环境中的窥视行为.与现有的ATM视频监控技术相比,所提出的监控方法具有检测范围广、智能化水平高、存储检索效率高等优点,实现窥视行为检测算法具有较高的鲁棒性和检测精度.

This paper discusses a new intelligently monitoring system for the automatic teller machine （ATM） by using an omni-directional vision sensor （ODVS） and the computer vision technology. It uses ODVS to acquire the omni-directional image and perspective algorithm to unwrap the omni-directional image into several separated key surveillance sections. To improve the search efficiency and minimize the storage and data transmission, this paper presents a face detection algorithm based on Gaussian skin-color model to acquire the ATM users＇ face image and combines it with the time and location information of certain activities, which includes the time and the location and the person who is using ATM. To improve the safety of using ATM, this paper provides some means which based on the Kalman filter of face tracking and behavior semantic rules judgment to detect and analyze the peeping activities in the ATM surrounding area. Comparing with the existing ATM surveillance technology, the surveillance method proposed in this paper has the advantages of detecting widely range, high-level intelligence, high-efficiency storage and high-efficiency searching. The peeping behavior detecting algorithm has high robustness and detection precision.