Research on Target Grouping Method Based on Graph Neural Network

With the application of a large number of battlefield observation sensors, the battlefield target data obtained by the command and control system has shown a surge trend. The scale of target information data is large, problems such as high dimensionality and complex structure have brought new challenges to the target grouping technology. The traditional clustering method has been unable to effectively target grouping in high-dimensional battlefield target data. Aiming at the dilemma of current target grouping technology, in order to help commanders get different types of target groups, this paper constructs a target grouping algorithm based on graph neural network, which combines graph convolutional neural network and clustering algorithm, and uses self-supervised mechanism to optimize the model. Through the verification experiment of the target clustering model on 4 high-dimensional data sets, three clustering effect evaluation indicators are established, and compared with the two methods of K-means and AE+k-means, it proves that our method is more efficient and intelligent.