Data association is one of the core components in multi-target tracking systems. Its goal is to establish the correspondence between observed data and surviving targets, newly emerged targets, and clutter, i.e., to identify the source of the observed data. The main reason for data association problems is the uncertainty in the multi-target tracking environment and the sensor observation process. Firstly, due to the lack of prior knowledge, the time of target appearance and disappearance is uncertain, and the number of targets is also unknown. Secondly, sensors inevitably generate false alarms and miss detections. For correct detections, they are unordered measurement sets, and it is impossible to know their correspondence with the targets, and they contain a large amount of measurement noise.