摘要
针对复杂背景下视觉目标跟踪问题,提出了一种基于多特征融合和改进建议分布函数的粒子滤波目标跟踪算法。为了解决单一特征跟踪稳定性差的问题,该方法在构造粒子滤波算法观测似然函数的过程中,综合利用颜色、梯度和纹理特征,并给出一种有效的特征权值自适应分配策略。针对传统建议分布函数无法利用观测信息的缺陷,提出了一种基于PSO算法的建议分布函数,有效地抑制了粒子退化现象。实验采用复杂地面环境下的多组图像序列,结果表明该算法的有效性。
Aiming at the visual target tracking problem under the complex background, a particle filter tracking algorithm based on multi-features fusion and improved proposal distribution function is proposed in this paper. To resolve the had stability of single-feature tracking, this algorithm comprehensively makes use of color, gradient and texture features during the construction of observation likelihood functions in particle filter algorithm, and an effective feature weight self-adapting allocation strategy is presented. To effectively restraint the particle degeneration phenomenon, a proposal distribution function based on PSO algorithm is put forward. Multiple groups of complex video sequences are adopted for experiment. The results show that the proposed algorithm is effective.
出处
《光学与光电技术》
2017年第4期72-77,共6页
Optics & Optoelectronic Technology
关键词
目标跟踪
粒子滤波
多特征融合
建议分布
PSO算法
target tracking
particle filter
multi-features fusion
proposal distribution
PSO algorithm