文档介绍：学兔兔
第 35卷第 4期仪器仪表学报 VoL35
2014年 4月 InstrumentChinese Scientific Journal of
EVibe:一种改进的 Vibe运动目标检测算法术
余烨,曹明伟,岳峰
(合肥工业大学计算机与信息学院合肥 230009)
摘要:为了提高运动目标检测结果的精确性,针对经典 Vibe算法中存在的问题进行改进,提出一种加强型 Vibe算法(
EVibe);此算法扩大了样本的取值范围,避免了样本的重复选取;采用隔行更新方式对邻域进行更新,避免了错误分类的扩散;
采用小目标丢弃和空洞填充策略去除了噪声的影响;添加了阴影去除模块,增强了算法对阴影的鲁棒性。通过 3套数据集对
EVibe算法进行测试,实验结果表明,EVibe算法在准确率和识别率上均有所提高,且目标检测结果更加精确。
关键词:目标检测;背景建模;视频监控;阴影去除;随机聚类
中图分类号:TP391 文献标识码:A 国家标准学科分类代码:
EVibe:Animproved moving Vibedetecting algorithmfor objects
Yu Ye,Cao Mingwei,Yue Feng
(School puter&Information,Hefei University of Technology,Hefei 230009,China) Technology,Hefei of puter&Information,Hefei (School University
Abstract:In order to improve the accuracy of the moving object detection,aiming at the problems existing in classic Vibe algorithm,an classic in existing problems the detection,aiming at object moving the of accuracy Abstract:In the improve to order
enhanced Vibe(EVibe)algorithm was EVibe algorithm expands the range of sample values,which avoids the repetitive repetitive the values,which sample avoids Vibe(EVibe)algorithmof enhanced range the was expands EVibe algorithm
selection interlacedupdating manner adoptedis theforneighborhood update eachof pixel,which prohibits thediffu—
sion of error object discarding a