Air Target Threat Assessment Based on Graph Convolutional Neural Network

In modern air warfare, battlefield commanders need to make decisions based on the assessment of the threat level of enemy targets. However, excessive threat parameters, complex judgment rules and continuous application of flight formations bring great challenges to threat assessment. To this end, a threat assessment method for air targets based on graph convolutional neural networks is proposed in this paper. In this method, six threat features are selected: speed, distance, altitude, radar cross-section(RCS), jamming signal level (JSL) and closest point of approach (CPA). The topology of flight formation is represented by adjacency matrix, which is used to assess the threat degree of air targets in different flight formation. By extracting the feature of threat coefficient and topology structure, the threat level of target in formation can be evaluated. The test results show that the accuracy of this model can reach 96.75%, which is better than the traditional methods.