CAJ | 학술논문

针对粒子滤波中粒子匮乏及样本聚集问题，提出一种基于佳点集的拟蒙特卡洛粒子滤波算法（ GPS-QM-CPF）。该算法利用数论中的佳点集理论和方法来构造出一种新的拟蒙特卡洛序列。由于佳点集序列与随机点列和标准的拟蒙特卡洛序列相比分布更均匀、偏差更小，使得在滤波过程中状态估计的精度和收敛速度都得到提高，同时还能增加粒子有效样本数和降低重采样次数。实验结果表明，提出的算法在非线性系统状态估计精度要优于粒子滤波和标准的拟蒙特卡洛粒子滤波算法，并且在视频目标跟踪的应用中，针对跟踪目标受到遮挡的情况，算法具有更高的跟踪精度，同时跟踪的实时性也得到了一定程度的提高。
침대입자려파중입자궤핍급양본취집문제，제출일충기우가점집적의몽특잡락입자려파산법（ GPS-QM-CPF）。해산법이용수론중적가점집이론화방법래구조출일충신적의몽특잡락서렬。유우가점집서렬여수궤점렬화표준적의몽특잡락서렬상비분포경균균、편차경소，사득재려파과정중상태고계적정도화수렴속도도득도제고，동시환능증가입자유효양본수화강저중채양차수。실험결과표명，제출적산법재비선성계통상태고계정도요우우입자려파화표준적의몽특잡락입자려파산법，병차재시빈목표근종적응용중，침대근종목표수도차당적정황，산법구유경고적근종정도，동시근종적실시성야득도료일정정도적제고。
A quasi-Monte Carlo particle filtering algorithm based on the good point set (GPS-QMCPF) is proposed for solving the problem of particle shortage and sample aggregation .In the proposed algorithm , a new quasi-Monte Carlo sequence is constructed by using the good point set theory in the number theory .Considering that the good point set has a more homogeneous distribution and lower discrepancy than the standard QMC sequence and the random se -quence , GPS-QMCPF can obtain a faster convergence speed in the filtering process and a better accuracy of the state estimation .Furthermore the re-sampling frequency is reduced , which results in a lower computational cost .Experi-mental results show that the proposed algorithm gets a more accurate estimation than the standard QMC filter and par -ticle filter in the system state estimation , as well as with the video target tracking application .The proposed algorithm possesses the advantages of good tracking accuracy and a real -time standard , even in the case of occlusions .