论文部分内容阅读
分层卡尔曼粒子滤波成功应用于目标跟踪,但其只对目标位置进行了优化,忽略了其他仿射参数,导致跟踪中的粒子数目仍然很大。为了实现复杂环境下的快速目标跟踪,提出一种带有自调整策略的分层卡尔曼粒子滤波方法。该方法将目标划分为线性和非线性状态空间,并通过少量粒子的迭代过程在非线性状态空间逐步搜索最优状态。其详细过程如下:首先,利用卡尔曼滤波预测目标位置,结合目标运动信息计算潜在目标区域;然后在该区域内生成一组随机粒子,通过在线姿态估计对粒子状态进行调整,并将观测结果与目标模板进行比较,修正粒子摄动的方向以逼近目标。把该方法应用于大机动目标的视频序列中,并与现有的跟踪方法进行了对比。结果表明,所提方法能够以少量粒子实现准确、稳定的目标跟踪,大大降低了跟踪算法的运算量,提高了跟踪效果。
Hierarchical Kalman particle filtering is successfully applied to target tracking, but it only optimizes the target position, ignoring other affine parameters, resulting in a large number of particles in the tracking. In order to achieve fast target tracking in complex environment, a hierarchical Kalman particle filter with self-tuning strategy is proposed. The method divides the target into linear and non-linear state space, and searches the optimal state step by step through the iteration process of small particles in the nonlinear state space. The detailed process is as follows: Firstly, the Kalman filter is used to predict the target position and the potential target area is calculated according to the target motion information. Then a set of random particles are generated in this area, and the particle state is adjusted by on-line attitude estimation. The target template is compared to correct the direction of particle perturbation to approximate the target. This method is applied to the video sequences of large maneuvering targets and compared with the existing tracking methods. The results show that the proposed method can achieve accurate and stable target tracking with a small number of particles, greatly reducing the computational complexity of the tracking algorithm and improving the tracking performance.