基于背景和帧间差分法的运动目标提取

在自动人脸识别系统中,人体目标的检测是该系统识别人脸的前期关键步骤之一。为了能快速准确地检测出运动的人体目标,提出了背景差分法和帧间差分法相结合的运动目标提取检测方法。针对视频中的背景因镜头摆动或物体的移入/移出而有所变动,给出了几种提取背景的方法,比较各种方法后决定采用中值法快速地进行背景建模,随后采用自适应背景更新,结合两种差分法检测运动目标。通过实验验证了本文算法的有效性。

Moving object extraction based on background difference and frame difference method

In automatic face recognition system, the human target detection is one of the key steps to recognize face. Several methods of moving target detection are introduced, and the difference method of background and frame is presented, in order to detect human targets fast and precisely. Aiming at the change of background due to the moving of lens or targets, a few methods of background extraction are given. After comparing several methods, the median method is chosen to do background modeling. The adaptive background updating is used to detect moving targets, combining two kinds of difference methods. Finally, the effectiveness of this algorithm is verified by experiments.