一种能快速抑制鬼影及静止目标的ViBe改进算法

发布时间：2019-07-17 13:33

【摘要】：ViBe算法存在鬼影和静止目标问题,这些问题给目标检测带来误差,需要快速有效地抑制。文章在原始ViBe算法基础上,先通过比较局部区域的背景模型像素值方差和新来帧该区域的像素值方差的大小来判断该区域存在鬼影还是静止目标,存在则进行抑制,然后采用不同的策略更新鬼影区域和静止目标区域的背景。实验结果中,改进算法仅用15帧就可以完全抑制鬼影,仅用20帧就能完全抑制静止目标,而原始ViBe算法完全抑制鬼影需要108帧且抑制静止目标能力有限。实验结果表明,对于普通的以及背景较为复杂的监控场景,文中改进算法可行、有效。
[Abstract]:Vibe's algorithm has the problem of ghost and stationary object, which can bring the error to the target detection, and it needs to be effectively suppressed. On the basis of the original Vibe algorithm, the existence of ghost or stationary object in the region is judged by comparing the pixel value variance of the background model of the local region and the magnitude of the variance of the pixel value of the new frame. The background of the ghost region and the stationary target region is then updated with a different strategy. In the experimental results, the improved algorithm can completely suppress the ghost image by only 15 frames, can completely suppress the static target with only 20 frames, and the original Vibe algorithm completely suppresses the ghost image to need 108 frames and has limited static target capability. The experimental results show that the improved algorithm is feasible and effective for common and complex monitoring scenarios.
【作者单位】：合肥工业大学计算机与信息学院;合肥工业大学电子科学与应用物理学院;
【基金】：国家自然科学基金资助项目(61371155) 安徽省科技攻关计划资助项目(1301b042023)
【分类号】：TP391.41;TN948.6

【相似文献】