摘要
随着跟踪环境、跟踪对象和跟踪系统的不断变化、发展 ,目标与量测已很难仅仅以一一对应的关联关系来描述 ,这使得多目标跟踪中数据关联这一核心问题更具挑战 .JesusGarrcia、T .Kirubarajan和Bar Shalom等学者从智能方法或重复使用一对一分配JPDA等方面进行了研究 ,取得一定成效 ,但计算量和性能均未达到理想效果 .本文首先提出更符合实际情况的新的目标与量测相关联的可行性规则 ,给出广义联合事件的一种分割与组合方法 ,利用贝叶斯法则推导出了一种全局次优的广义概率数据关联算法 (GeneralizedProbabilityDataAssociation ,GPDA) .通过本文设计的各种典型环境的仿真计算表明 ,GPDA算法的性能在目标与量测无论是否在一一对应的情况下 ,全面优于JPDA算法 ,且由于新算法的设计技巧 ,使计算量和存储量也大大小于JPDA算法 ,为发展同时具有良好实时和关联性能的多目标跟踪算法给出了新的尝试 .
With the change and development of modern multi-target tracking system, it is very difficult to deal with data association problems simply using the feasible rule based on the hypothesis in which the association of measurements with targets is one-to-one correlated to each other, as is commonly used in JPDA. We have noticed that T. Kirubarajan and Bar-Shalom et al. gave some new results trying to solve the problem. But the performance, especially the computing burden of the algorithm can not be satisfied by most real time systems. In this paper, we put forward a new feasible rule which is more suitable for practical environment of multi-target tracking system. Based on the new feasible rule, we define a new concept of generalized joint event. We present a method to segment the generalized joint event set into two generalized event sub-sets and then a combination method with the two sub-sets is put forward. A Generalized Probability Data Association (GPDA) algorithm is deduced by using Bayesian rule. Additionally, we analyze the performance of GPDA algorithm in various given tracking environments by using Monte Carlo simulation. We compare the computation burden and computing memory with JPDA algorithm. All simulation results show that the performance of GPDA is superior to that of JPDA, and the algorithm has much smaller computation burden than JPDA.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2005年第3期467-472,共6页
Acta Electronica Sinica
基金
国家自然科学基金 (No .60 1 72 0 37)
关键词
多目标跟踪
数据关联
广义联合事件
广义慨率数据关联
Algorithms
Associative processing
Computer simulation
Knowledge based systems
Probability
Radar tracking
Sensor data fusion
