An approach to tertiary processing of information from radar stations, radio-technical and optical-electronic reconnaissance stations is proposed, which ensures a reduction in the time for identifying the class of a tracked air object in military complexes due to the adaptive calculation of the membership function threshold to the number and density of location targets, increasing the probability of correct identification of heterogeneous data for subsequent merging of source decisions on the target class. More advanced are the methods of strobe-free identification based on the formation of a decision rule based on the minimum of a certain functional or exceeding a fixed threshold. Bayesian fusion methods require accurate data on a priori probabilities, and also assume a uniform density of distribution of objects, which is not always fulfilled in practice and gives an incorrect result. The approach based on the Dempster-Shafer theory for large sets of hypotheses has high computational complexity without preliminary data selection, and is poorly adapted to their dynamic change. Determination of the reference object is an iterative procedure based on the Yagel-Filev mountain clustering algorithm, a well-known approach. The result of the procedure is an array of reference objects and an array of trace marks that must be distributed among the reference ones, i.e. to group the data and determine the true parameters and features of the VO to which this group will belong. Grouping is carried out by calculating the membership function of each mark to each reference object using the known expression of the Bezdek-Dann fuzzy clustering algorithm. Due to the presence of a technique for preliminary recognition and selection of VO classes in the developed method at the initial stage of data merging, the number of pairwise comparison operations is reduced, according to statistical estimates, by an average of 1.5 times.