摘要
针对视觉跟踪中粒子滤波算法的建议性分布函数选择问题,提出一种目标轮廓跟踪的高斯厄米特粒子滤波算法(GHPF)。该算法采用B样条曲线描述目标轮廓,建立目标运动模型。利用高斯厄米特滤波器产生建议性分布函数,通过实时融入最新的观测数据来逼近系统状态的后验概率,提高了滤波估计的精度。实验仿真结果验证了所提算法的有效性。
A novel target profile tracking algorithm based on Gauss-Hermite particle filter is proposed for the selection of proposal distribution function in particle filter algorithm of visual tracking. The algorithm describes the target contour by adopting B-spline curve to set up target moving models. The Gauss-Hermite filter is used for generating the proposal distribution function to approach the posteriori probability of system state by fusing the latest observations, and it improves the accuracy of filter estimation. The simulation experiment demon- strates the effectiveness of proposed algorithm.
出处
《测控技术》
CSCD
北大核心
2011年第12期20-24,28,共6页
Measurement & Control Technology
基金
国家自然科学基金资助项目(60774091)
空军装备部基金资助项目(KJ09131)
