基于Kalman预测重要性建议分布的粒子滤波视觉跟踪算法

标准粒子滤波虽然能够实现简单场景下的目标跟踪,但在复杂场景下其性能较差,粒子权值退化是影响视觉跟踪的一个重要方面,为解决这一问题,从选择准确重要性建议分布函数入手,给出了基于EKF和UKF预测采样的粒子滤波视觉跟踪算法EKF-PF（EKF enhanced particle filtering）和UKF-PF（UKF enhanced particle filtering）,并进行了一定改进,通过仿真实验表明：给出的跟踪算法能够很好地跟踪室内运动目标,并对光照变化,目标姿态变化具有良好的鲁棒性.

Video Object Tracking Based on Particle Filter with Kalman Prediction Important Proposal Distribution

The standard particle filter can achieve target tracking in a simple scene, but its performance is poor in complex scenes. The particle weights degradation is an important aspect of visual tracking. To solve this problem, this article gives particle filter visual tracking algorithm EKF-PF （EKF enhanced parti- cle filtering） and UKF-PF （UKF enhanced Particle Filtering） based on EKF and UKF prediction sampling from the accurately selecting important proposal distribution function. The simulation experiments show that the tracking algorithm proposed in this paper can perfectly track indoor moving target. The light changes and the target attitude changes have a good robustness.