前景目标检测中的虚影消除算法

摘要

Virtual shadow is kind of false foreground which is produced by the partial change of the background. As common foreground detection algorithm is hard to identify the virtual shadow in complex environment, a virtual shadow elimination algorithm based on improved codebook model in YUV color space is proposed. Algorithm starts from the production principle of the virtual shadow, and separately models the foreground and background. And a conversion is designed for converting the foreground to background under the control of the preset parameter. Experimental results show that on three categories of ghost the algorithm can quickly and accurately remove the ghost and run in real time. The proposed algorithm is helpful for the foreground object identification.%在复杂环境下，传统前景目标检测算法无法有效区分背景场景中局部区域动态变化所产生的虚影。为解决该问题，提出一种基于YUV色彩空间和改进码本模型的虚影消除算法。从虚影的形成原理出发，分别对前景和背景进行建模和更新，使得前景、背景可以在预设参数的控制下完成相互转换。实验结果表明，该算法能够准确地消除3种不同的虚影，且能满足实时性需求，为后续的前景目标识别消除干扰。