Abstract: Moving object detection under complicated circumstance is one of the important tasks in the smart surveillance system, during the course of which the chosen threshold is one of the key factors. The traditional moving object detection algorithm based on the fixed threshold can hardly meet the practical requirement in complex environment. Therefore, a moving object detection method based on adaptive dynamic threshold is proposed using Bayesian theory, which can overcome the adverse effect of complex conditions such as illumination by introducing the mean and variance of the foreground image and the mean of the background image. The experimental results show that compared with the traditional fixed threshold method, the proposed method can suppress effectively the influence of the noise and take on robust and stable performance.

Key words: Bayesian theory, object detection, fixed threshold, dynamic threshold

摘要： 复杂背景下精准的移动目标检测是智能监控系统的重要任务之一，而移动目标检测中，阈值的选择是关键因素之一。传统的固定阈值检测算法很难满足光照等复杂环境的实际需要，利用贝叶斯理论，提出了自适应的动态阈值移动目标检测算法，通过引入前景图像的均值和方差，以及背景图像的均值，获得自适应的动态阈值，用于克服光照等复杂条件的不利影响。实验结果显示，同传统的固定阈值检测算法相比，提出的算法可以有效地克服噪声的影响，并且在复杂环境下具有更好的鲁棒性和稳定性。

关键词: 贝叶斯理论, 目标检测, 固定阈值, 动态阈值