基于元认知模型的智能混合高斯背景建模①-计算机系统应用.PDF

计 算 机 系 统 应 用 2013 年 第 22 卷 第 9 期 基于元认知模型的智能混合高斯背景建模① 陈 真, 王 钊 ( 中国石油大学(华东) 电子信息工程系, 青岛 266580) 摘 要: 传统混合高斯背景模型(Gaussian mixture model, GMM)不能快速适应动态场景中背景发生突变的情况. 本文提出一种基于元认知模型的智能混合高斯背景建模方法, 每个输入像素经过元认知监控成分刺激元认知体 验成分以提取成功(或失败) 的意识进行认知, 根据提取的意识及时向元认知知识成分传输新的认知知识或直接提 取元认知知识成分, 并作出决策信息. 该方法能够对背景模型产生认知, 当背景突变为认知过的背景时, 可以快 速适应并能更准确地描述复杂场景中的真实背景. 关键词: 混合高斯模型; 元认知模型; 背景建模; 目标检测; 背景减除 Metacognitive Model-based Intelligent Gaussian Mixture Background Modeling CHEN Zhen, WANG Zhao (Electronic Information Engineering, China University of Petroleum, Qingdao 266580, China) Abstract: Traditional Gaussian Mixture Model (GMM) can’t adapt quickly to the sudden changes in the dynamic scene. A Metacognitive Model-based Intelligent Gaussian Mixture background modeling method is proposed. For each pixel, decision making must pass the three elements which are called as Metacognitive Monitoring (MM) element, Metacognitive Cognitive Knowledge (MCK) element and Metacognitive Experience (ME) element. MM element monitors the modeling scheme, stimulates to get ME element, and extracts cognitive knowledge from MCK element. MCK element is composed of the background models cognized ever. ME element is composed of the experiences from success or failure matching. In the complex scene, the proposed method can intelligently cognize background, quickly adapt to the sudden changes of backgrounds cognized ever and describe real background more accurately. Key words: Gaussian Mixture Model (GMM); Metacognitive model; background modeling; object detection; background subtraction 混合高斯背景模型 (Gaussian mixture model, 突变时仍能快速、准确地检测出前景的问题, 学者们 GMM)[1] [2] 提出了许多改进 GMM 的方法. 将场景中的时间或空 是背景减除法 中的经典模型之一, 适用于背 景场景相对复杂的情形, 被广泛应用于运动目标检测