改进理解诊断跟踪系统的目标跟踪方法  被引量:2

A target tracking algorithm for the improved understanding,diagnosing and tracking system

在线阅读下载全文

作  者:于蕾[1] 夏业儒 杨良洁 YU Lei;XIA Yeru;YANG Liangjie(College of Information and Communication Engineering,Harbin Engineering University,Harbin 150001,China)

机构地区:[1]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001

出  处:《应用科技》2018年第4期76-81,共6页Applied Science and Technology

基  金:国家自然科学基金项目(61771155)

摘  要:针对目标跟踪过程中出现的遮挡、光照变化、背景复杂等问题,使用了理解诊断视觉跟踪系统,即把跟踪器分成5个组成部分的跟踪系统,这5个部分分别是运动模型、特征提取器、观察模型、模型更新器以及总体处理器。结合Haar矩形特征的原理,提出了3种Haar-Like特征,用在特征提取器模块。为了提高跟踪的精准性,引入一个简单且快速的鲁棒性算法来改进系统中的运动模型,该方法利用了视觉跟踪中的上下文关系,建立基于贝叶斯框架的目标以及其周围环境的时空关系,在检测方面使用了快速傅里叶变换方法,提高了算法的鲁棒性,使跟踪更加精准,并且在处理遮挡、光照变化、背景复杂等问题上有着较好的效果。In order to solve the problems of occlusion, illumination changes, and background clutters in target tracking, this paper proposes a new tracking model:the understanding and diagnosing visual tracking system, which is based on a framework by breaking a tracker into five constituent parts, namely, motion model, feature extractor, observation model, model updater, and ensemble post-processor. Based on the principle of Haar-Like features, this paper proposes three new Haar-Like features in the feature extractor. To improve the tracking accuracy, a simple yet fast and robust algorithm is used to improve the movement model of the system. This approach utilizes the context relationship in visual tracking to establish the spatio-temporal relationship based on a Bayesian framework and its surrounding environment, adopts the Fast Fourier Transform in the test to increase robustness of the algorithm, making the tracking more accurate, and having better effect in occlusion handling, lummination change, background complexity, and the like.

关 键 词:运动模型 特征提取器 观察模型 更新模型 总体处理器 Haar-Like 快速傅里叶变换 上下文模型 

分 类 号:TP391.4[自动化与计算机技术—计算机应用技术]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象