基于MLS的三维扩展目标PMBM跟踪算法

3D extended target PMBM tracking algorithm based on moving least square

针对传统三维扩展目标跟踪算法形状估计精度低的问题,提出了一种基于移动最小二乘的泊松多伯努利混合(Poisson multi-Bernoulli mixture based on the moving least square,MLS-PMBM)滤波跟踪算法。该算法基于MLS模型构建三维扩展目标的形状矩阵,通过PMBM滤波器预测和更新目标的运动状态,利用移动最小二乘算法更新形状矩阵,结合目标质心状态与形状估计完成对三维扩展目标的跟踪。仿真实验与实际点云数据的验证表明,与现有算法相比,本文所提算法在多扩展目标的形状估计方面具有更优的性能,具有较高的泛用性。

Aiming at the problem of low accuracy of shape estimation in traditional 3D extended target tracking algorithms,a of filtering tracking algorithm Poisson multi-Bernoulli mixture based on the moving least square(MLS-PMBM)is proposed.The algorithm constructs the shape matrix of the 3D extended target based on the MLS model,predicted and updated the motion state of the target by PMBM filter.The shape matrix is updated by the moving least squares algorithm,and the tracking of the 3D extended target is completed by combining the centroid state and shape estimation of the target.Simulation experiments and actual point cloud data verification show that compared with the existing algorithms,the proposed algorithm has better performance in the shape estimation of multiple extended objects,and has higher universality.