Data association is a key problem in multi-target tracking and has been broadly investigated by researchers. At present, the class of data association methods based on Bayes' rule is the mainstream among many kinds of data association algorithms. One of the best-known Bayesian data association algorithms is Joint Probability Data Association (JPDA) which has shown to be effective in handling clutters and missed detection. Based on JPDA, a class of algorithms has been proposed according to actual problems. For example, the Exact Nearest Neighbor Version of the JPDA (ENNPDA), Coupled Probabilistic Data Association (CPDA) and Joint Integrated Probabilistic Data Association(JIPDA), etc. However, Great computational cost makes JPDA difficult to meet the real-time requirement in multi-target data processing system and its feasible rule of strict one-to-one relation between measurement and target appears to be impropriate in many practical situations. For instance, in target crossing flying or in dense military aircraft formation flying, measurements from multiple aircraft may be received as one measurement by sensor. Another example is the overlapping measurements observed in image sequences. Therefore, some researchers attempt to loose the strict one-to-one rale to multiple-to-multiple rule, such as Jesus Garrcia and T. Kirubarajan. However, their algorithms always increase the computational cost.
展开▼
