一种提升鬼影抑制性能的改进视觉背景提取算法
An Improved Visual Background Extractor for Suppression of Ghosts
杜瑾 1苏雨 2张义飞 1邹坤3
作者信息
- 1. 中国空空导弹研究院,河南 洛阳 471000;空基信息感知与融合全国重点实验室,河南 洛阳 471000
- 2. 西北工业大学,陕西 西安 710072
- 3. 中航光电科技股份有限公司,河南 洛阳 471000
- 折叠
摘要
运动目标检测为视频帧生成指示运动目标像素的二值图,因此,正确判别待检测场景中的运动目标像素和背景像素是运动目标检测算法的主要任务.本文针对传统场景提取算法背景模型初始化导致的"鬼影"问题,提出了时域区间参考模块,该模块通过时域区间内像素的统计值生成背景参考值,在像素判别阶段利用候选前景像素与该参考值的差异进一步确定前景像素.实验结果表明,本文提出的方法对传统视觉背景提取算法性能进行了提升,对"鬼影"有较强的抑制作用,提高了检测准确度,具有较好的检测性能.
Abstract
Moving object detection generates a binary map describing moving object pixels for a frame,so it is the main task of the moving object detection algorithm to correctly discriminate foreground moving pixels from background pixels.To address the"ghosting"problem caused by the initialized background model of the classical visual background extraction algorithm,a time-domain interval reference module is introduced.The module generates a background reference value from the statistics of the pixels in the interval,and then uses the difference between the candidate pixel and the reference value to further determine the foreground pixel.The experimental results show that the algorithm in this paper makes up for the shortcomings of the original visual background extractor,has a strong suppression effect on"ghost",improves the detection accuracy,and has a better performance for detecting moving object.
关键词
运动目标检测/背景减除/鬼影Key words
moving object detection/background subtraction/ghost引用本文复制引用
出版年
2024
NETL
NSTL
维普
万方数据