基于卡尔曼滤波和粒子滤波的目标跟踪性能对比

Comparison of Target Tracking Performance Based onKalman Filter and Particle Filter

粒子滤波器是一种也称序列蒙特卡洛采样的近似贝叶斯滤波器,其具有不受目标系统是否线性,噪声系统是否高斯分布等限制的特性,能够对目标系统进行最小方差估计,广泛运用在目标跟踪、自动控制和参数估计等领域。而具有线性递推特性的卡尔曼滤波在线性系统中根据前一时刻的目标状态和当前时刻的观测量,来获得当前时刻状态的最优估计,具有无偏,稳定和最优的特点。本文通过建立在目标系统中的状态转移方程和目标观测方程,运用粒子滤波和卡尔曼滤波对在仿真环境下的汽车的行驶轨迹进行跟踪预测,并在MATLAB软件环境下建模并仿真验证:通过对比两种滤波器的估计误差,表明了卡尔曼滤波在线性高斯系统中的优势。粒子滤波估计误差整体较小但在粒子迭代,重要性采样函数,粒子退化,重采样等步骤上具有更高的算法复杂度。

Particle filter is a kind of approximate Bayesian filter and also known as sequential Monte Carlo sampling,which has the property of not being restricted by whether the target system is linear or not,whether the noise system is Gaussian distributed or not,and is able to perform the minimum variance estimation of the target system,and is widely used in the fields of target tracking,automatic control and parameter estimation.The Kalman filter with linear recursive property is based on the target state at the previous moment and the observed values at the current moment and has the characteristics of unbiased,stable and optimal,is used to obtain the optimal estimate of the current state in a linear system.In this paper,by establishing the state transfer equation and the target observation equation in a linear system,we use particle filter and Kalman filter to predict the tracking trajectory of a car in a simulation environment,and simulate it in MATLAB to verify:by comparing the estimation error,particle filter shows better performance for target tracking.Butthe linear recursive calculation of Kalman filter is Simplerin principle,while the particle filter has higher algorithmic complexity in the steps of particle iteration,importance sampling function,particle degradation,and resampling.