EVibe:一种改进的Vibe运动目标检测算法  

EVibe:An improved Vibe algorithm for detecting moving objects

作　　者：余烨[1] 曹明伟[1] 岳峰[1] 

机构地区：[1]合肥工业大学计算机与信息学院,合肥230009

出　　处：《仪器仪表学报》2014年第4期924-931,共8页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(61070124);安徽省科技攻关项目(12010402108)资助项目

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。In order to improve the accuracy of the moving object detection,aiming at the problems existing in classic Vibe algorithm,an enhanced Vibe （EVibe） algorithm was proposed.The EVibe algorithm expands the range of sample values,which avoids the repetitive selection of the samples.The interlaced updating manner is adopted for the neighborhood update of each pixel,which prohibits the diffusion of error classification.Small object discarding and hole filling strategies are used to remove the effect of noise.Meanwhile,a shadow removing module is added,which improves the robustness of the algorithm when shadows exist.The EVibe algorithm was tested with three datasets,Experiment results show that the EVibe algorithm has improvements both in detection precision and recall rate,and the object detection results are more accurate.

关 键 词：目标检测 背景建模 视频监控 阴影去除 随机聚类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]