Abstract: This paper proposes an object detection algorithm combined with regional-level and pixel-level background subtraction method.It can effectively resolve the problem of uneven distribution of noise in video sequence.A self-adaptive local threshold based on entropy power is used to classify foreground and background.On this basis，pixels in the region adjacent to the foreground and background are detected.This paper also proposes a background updating algorithm which is self-adaptive to changes of illumination.Experimental results demonstrate that the integrity of object detection and noise immunity are better than traditional single threshold background subtraction by using this method，and this algorithm can accurately detect moving targets.

Key words: video detection, entropy power, self-adaptive threshold, background updating

摘要： 提出一种结合区域级和像素级背景差分法的目标检测算法，可以有效解决视频序列中噪声分布不均问题。利用一种基于熵能的局部自适应阈值划分出前景和背景，在此基础上对前景和背景邻接区域像素点进行检测，并提出一种自适应光线变化的背景更新算法。实验结果表明，该算法比传统单阈值背景差分法抗噪能力更强，检测目标轮廓更加完整，能够准确检测出运动目标。

关键词: 视频检测, 熵能, 自适应阈值, 背景更新