CAJ | 학술논문

对基于有限集统计的多目标跟踪算法进行研究。PHD滤波器针对传统多目标跟踪中计算量较大的问题，通过计算多目标分布后验密度的一阶矩模拟多目标贝叶斯迭代，取得了良好的效果。针对PHD滤波器对目标漏检较为敏感的缺陷，Cardinalized PHD（CPHD）滤波器同时估计得到后验密度分布和后验势分布，是对PHD滤波的泛化。通过实现PHD和CPHD滤波的混合高斯模型，分析比较两者的性能。实验表明，由于引入了势分布，CPHD滤波提高了对目标估计的准确率，总体性能优于PHD滤波算法。
대기우유한집통계적다목표근종산법진행연구。PHD려파기침대전통다목표근종중계산량교대적문제，통과계산다목표분포후험밀도적일계구모의다목표패협사질대，취득료량호적효과。침대PHD려파기대목표루검교위민감적결함，Cardinalized PHD（CPHD）려파기동시고계득도후험밀도분포화후험세분포，시대PHD려파적범화。통과실현PHD화CPHD려파적혼합고사모형，분석비교량자적성능。실험표명，유우인입료세분포，CPHD려파제고료대목표고계적준학솔，총체성능우우PHD려파산법。
Multiple targets track based on finite sets statistic is studied. Probability Hypothesis Density (PHD)filter simulates Multi-target Bayesian recursion via propagating the first moment of multi-target posterior probability,and acquires good results. The cardinalized PHD (CPHD) filter is proposed to overcome the excessive sensitivity in missing detection situation during PHD recursion,which jointly propagates the posterior intensity and the posterior cardinality distribution,and is a generalization for PHD filter. The Gaussian mixture implementation of the PHD and the CPHD filter are presented to analyze and compare their performance. Simulation results show that the CPHD filter is more reliable and accurate due to its propagation of cardinality distribution,thus the overall performance of the CPHD filter is better than the PHD filter.