Intelligent Identification Method of LEO Satellite Network Business Based on Grayscale Image

With the continuous growth of task demands and the increasing variety of access services, new challenges are posed to the satellite network's ability for multi-service collaborative transmission and global scheduling of network resources. It is necessary to differentiate the services carried by it in order to optimize resource allocation, meeting the quality assurance requirements of multi-service satellite networks. A satellite network business intelligent recognition method based on grayscale graphs is proposed, which analyzes the connectivity between satellites, establishes a spatio-temporal steady-state graph, determines the dynamic satellite network topology, and constructs a satellite simulation network. Based on this, five types of satellite service data are sampled to form a training dataset, where captured data packets are converted into grayscale graphs as algorithm inputs for training and recognition in satellite dynamic scenarios. The results indicate that with an increase in the number of iterations, the recognition accuracy improves, with a testing recognition accuracy of around 97%. This validates the effectiveness of the proposed algorithm and provides technical support for the development of satellite network space payloads.