Target grouping method based on SA-PFCM + + algorithm

Abstract

The invention discloses a target grouping method based on an SA-PFCM + + algorithm. The method includes: firstly initializing parameters, solving a mean value of the data set; calculating a first weighted Euclidean distance between each sample point and the mean value; after sorting, D2 is used for sampling to select a first initial clustering center; calculating a second weighted Euclidean distance between each sample point and the clustering center; after sorting, D2 is used for sampling to select the next initial clustering center; until the number of the initial clustering centers reachesa set condition; then, corresponding parameters are iteratively updated according to the initial clustering center; and calculating a corresponding Xie-Beni-Sun (XBS for short) index until the iteration frequency value reaches a set threshold value or the clustering member does not change any more, then updating the initial parameter value until a set stop condition is reached, comparing the XBS indexes under different clustering numbers, and outputting the cluster number and the class cluster under the set XBS index. The difficulty of target grouping in situation assessment is effectively reduced, and the decision-making efficiency is improved.

Description

technical field [0001] The invention relates to the technical field of target grouping, in particular to a target grouping method based on the SA-PFCM++ algorithm. Background technique [0002] Modern warfare is mainly based on electronic and information warfare. The main problem faced by the commander's decision-making is no longer the lack of battlefield information, but how to dig out useful information scientifically and efficiently from the massive information. In the actual battlefield environment, there are a large number of combat targets, and various types of targets are intertwined and mixed with ambiguity and uncertainty. These massive and complex information not only increase the complexity of the commander's correct analysis of the battlefield situation, but also increase the complexity of the battlefield situation with time. The increase in the amount of information is more and more, which makes it difficult for the commander to make real-time decision-making ...

AI Technical Summary

Problems solved by technology

The ISODATA algorithm is based on the K-means algorithm by adding merging and splitting operations on the clustering results to achieve dynamic clustering, but it uses the distance between the sample and the cluster center as the basis for clustering, which is suitable for solving the problem of clustering samples in spherical clusters. There are deficiencies in the common linear formation grouping problem. At the same time, the ISODATA algorithm requires many parameters to be initially set. It is difficult to set a more appropriate initial parameter value for complex and changeable battlefield target groups.

The nearest neighbor algorithm achieves grouping by setting a threshold, which is simple and easy to implement. However, as a supervised learning algorithm, it cannot adapt to the battlefield environment. None of the above algorithms can reduce the difficulty of target grouping in situation assessment and reduce decision-making efficiency.

Images