TARGET TRACKING METHOD AND APPARATUS BASED ON MEASUREMENT ALLOCATION

摘要

A target tracking method and apparatus based on measurement allocation. The method comprises: determining, on the basis of a state distribution, a presence probability, a detection identifier and a track identifier of each target at a previous moment, a predicted state distribution, a predicted presence probability, a predicted detection identifier and a predicted track identifier of each existing target at a current moment (101); generating a state distribution, a presence probability, a detection identifier and a track identifier of each new target at the current moment, and combining the predicted state distribution, the predicted presence probability, the predicted detection identifier and the predicted track identifier of each existing target at the current moment with the state distribution, the presence probability, the detection identifier and the track identifier of each new target at the current moment to obtain predicted state distributions, predicted presence probabilities, predicted detection identifiers and predicted track identifiers of all targets at the current moment (102); using the Bayes' rule to process the predicted state distribution and the predicted presence probability, obtained by means of combination, of each target at the current moment and all measurements at the current moment to obtain an updated state distribution, an updated presence probability and an updated detection identifier, corresponding to each measurement, of each target at the current moment, and an association probability of each target and each measurement (103); building a two-dimensional allocation problem on the basis of the association probability of each target and each measurement and a clutter density, solving the two-dimensional allocation problem to obtain an allocation result of all the measurements in the targets and clutter, and finally adjusting, according to the allocation result, the updated presence probability and the updated detection identifier (104); determining whether each target is an existing target at the current moment and is undetected (105); if so, respectively taking the predicted state distribution and the predicted detection identifier of the target as a state distribution and a detection identifier of the target at the current moment, and taking a product of the predicted presence probability and a preset attenuation factor of the target as a presence probability of the target at the current moment (106); if not, respectively taking an updated state distribution, an updated presence probability and an updated detection identifier corresponding to an index number of the maximum updated presence probability among all adjusted updated presence probabilities of the targets as a state distribution, a presence probability and a detection identifier of the target at the current moment (107); taking the predicted track identifier of the target as a track identifier of the target at the current moment (108); extracting, from all the targets at the current moment, targets for which the presence probabilities are greater than a first probability threshold value, respectively forming a state distribution set and a track identifier set at the current moment with state distributions and track identifiers of the extracted targets, and taking same as outputs of a filter at the current moment (109); and screening out, from all the targets at the current moment, targets for which presence probabilities are greater than or equal to a second probability threshold value, and taking state distributions, the presence probabilities, detection identifiers and track identifiers of all the screened targets as inputs of the filter for the next recursion (110). Multi-target tracking precision is guaranteed, the calculation quantity is effectively reduced, and the applicability in scenarios where clutter and non-detection are present is quite strong.