摘要
提出了一种基于BFGS拟牛顿公式的运动目标跟踪新方法。将跟踪目标描述为特征空间中的直方图概率密度函数,运用Bhattacharyya距离作为相似性度量准则并构造定位目标函数;通过BFGS拟牛顿方法合理确定搜索方向进行目标的定位;采用逐层区域划分搜索策略,有效的避免了目标丢失的问题。实验结果表明该方法在精确性和实时性方面均能满足运动目标跟踪的实用要求。
A moving object tracking method was proposed based on BFGS quasi-Newton formula. The histogram pdf in the feature space was used to describe the tracking target and the Bhattacharyya distance to measure the similarity. Then the localization target function was constructed. In order to find the proper target localization, the BFGS quasi-Newton method was used to decide the searching direction and object localization. The region division strategy can avoid losing of the target. The experiments indicate that this method can acquire good results, satisfying the demands in both accuracy and practice
出处
《系统仿真学报》
EI
CAS
CSCD
北大核心
2007年第1期13-15,34,共4页
Journal of System Simulation
基金
教育部博士点基金(20050422035)
关键词
运动目标跟踪
目标表达
BFGS拟牛顿公式
区域划分
moving object tracking
target representation
BFGS quasi-Newton formula
region division
