改进视觉背景提取算法在室内的研究

Research of Improved Visual Background Extraction Algorithm in Indoor

针对传统视觉背景提取算法(visual background extraction algorithm,ViBe)在室内监控条件下容易产生“鬼影”现象以及阴影难以消除的问题,提出一种针对人物室内活动的改进ViBe算法.针对容易引入“鬼影”现象,采用一种外接矩形像素直方图的方法,利用运动目标区域直方图与邻域背景区域直方图相似度匹配,快速检测“鬼影”现象,设置鬼影点像素值并参与背景模型更新,从而快速消除“鬼影”现象.针对阴影难以消除的问题,采用YCbCr颜色空间与混合高斯阴影模型相结合的方法共同确定阴影区域,从而解决阴影问题.仿真结果表明,相较于传统的ViBe算法,本文算法能够有效的消除“鬼影”现象,解决阴影问题,运动目标检测的准确性也得到了有效提高.

Aiming at the problems that visual background extraction algorithm(ViBe)is prone to“ghost”phenomenon and the shadow is difficult to be eliminated under indoor monitoring conditions,an improved ViBe algorithm for indoor activities of characters is proposed.Aiming at the easy introduction of“ghost”phenomenon,an external rectangular pixel histogram method is adopted to quickly detect the“ghost”phenomenon,set the ghost point pixel value and participate in the background model update,so as to quickly eliminate the“ghost”phenomenon.Aiming at the problem that the shadow is difficult to be eliminated,the method combining YCbCr color space and mixed Gaussian shadow model is adopted to jointly determine the shadow area,so as to solve the shadow problem.The simulation results show that compared with the traditional ViBe algorithm,this algorithm can effectively eliminate the“ghost”phenomenon,solve the shadow problem,and improve the accuracy of moving target detection.