等间隔矫正算法下稳定实时人脸跟踪监控技术  

Stable Real-Time Face Tracking and Monitoring Technology with Equal Interval Correction Algorithm

作　　者：涂毅晗 周慧 贾杰[2] TU Yi-han;ZHOU Hui;JIA Jie(College of Artificial Intelligence,Jiangxi University of Applied Sciences,Nanchang Jiangxi 330000,China;College of Information,Nanchang Hangkong University,Nanchang Jiangxi 330000,China)

机构地区：[1]江西应用科技学院人工智能学院,江西南昌330000 [2]南昌航空大学信息工程学院,江西南昌330000

出　　处：《计算机仿真》2020年第10期425-428,460,共5页Computer Simulation

基　　金：国家自然基金项目(61263012);江西省教育厅科技项目(151317)。

摘　　要：人脸跟踪监控过程中,光照强弱变换等环境复杂性导致视频图像产生几何畸变,降低跟踪精准度,由此提出等间隔矫正算法下稳定实时人脸跟踪监控技术。首先对采集视频图像序列中光照强度分布不均的图像,利用"参考白"方法做光照补偿,同时通过空域滤波法去除图像噪声,提高图像质量;其次,构建等间隔算法模型,计算畸变图像的中心坐标,实现图像校正;最后将Mean Shift跟踪算法与帧间约束相结合,实现稳定实时的人脸跟踪与监控。仿真结果证明,上述技术耗时较少,跟踪速度快,即使在环境较为复杂的情况下,也能保证对人脸的稳定跟踪监控。In the process of face tracking and monitoring, the complex lighting environment leads to the geometric distortion of video image. Therefore, a stable and real-time face tracking and monitoring technology under the equidistant correction algorithm was put forward. For the images with uneven light intensity distribution in video image sequence, "Reference White" method was adopted for light compensation. Meanwhile, the spatial filtering method was used to remove image noise and improve image quality. Secondly, the model of equidistant algorithm was constructed and the central coordinate of the distorted image was calculated to achieve image correction. Finally, Mean Shift tracking algorithm was combined with inter frame constraint to complete the stable and real-time face tracking and monitoring. Simulation results show that the proposed technology is less time-consuming than traditional method. Meanwhile, it is fast and stable in tracking and monitoring human face, even in complex environment.

关 键 词：等间隔校正 人脸跟踪监控 几何畸变 空域滤波 帧间约束 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]