中文摘要：

针对背景减法前景目标检测前期存在的“鬼影”，通过统计图像序列中各像素点处像素值的频次，高频次的像素值为背景像素，获得了无运动目标的背景，消除了“鬼影”；对于帧差法检测结果内部易产生“空洞”，在前景判断时增加了新的判决条件，改善了“空洞”问题；对于传统背景减法检测结果存在“拖影”问题，通过背景减法和改进帧差法的检测结果进行逻辑与操作，并将“与”结果作为背景图像更新判决条件。实验结果表明，该算法可以获得精确的背景图像，并且解决了“拖影”、“鬼影”及“空洞”问题，满足实时检测的需求。

英文摘要：

To eliminate ＂ghost＂ before foreground target detection of background subtraction method, no-moving target background is obtained according to count the frequency of pixel values at each pixel point within image sequence. High-frequency pixel is considered as background pixel. Then the ＂ghost＂ is eliminated. To eliminate ＂holes＂ inside of detecting result got by frame differ- ence, a new decision criterion is added when judging foreground. The problem of ＂holes＂ is improved. For the problem of ＂image tail＂ of detected result got by traditional background subtraction, a logic ＂and＂ operation is clone according to the detection results of the background subtraction and improved frame difference, and the results is regarded as the decision criterion of background up- date. Experimental results show that the algorithm can obtain accurate background and solve ＂image tail＂, ＂ghost＂ and ＂holes＂, and it fits for the real-time detection.