A target intention recognition method based on information classification processing and information fusion

Intention recognition of non-cooperative target is an important basis for battlefield command decision-making. Recent advances suggest recognizing target intention from a perspective of data-driven. However, existing data driven models do not consider complementary information between features to enhance their robustness in battlefield environments. To solve the problem, this paper constructs a novel neural network fusion model with information classification processing and information fusion to achieve target intention recognition. The model first designs the cross-classification processing method according to attributes' correlations and variation characteristics. Then, an interactive feature-level fusion method is proposed to model the finegrained correlations between attributes to discover salient features. Finally, a decision-level fusion method based on Dempster-Shafer theory is proposed to fuse the complementary information among attributes. The experimental results show that the recognition accuracy of the proposed model can reach 89.63%, and it can be maintained above 75% under the conditions of severe attribute missing or noise interference. It is demonstrated that the proposed model has higher accuracy and robustness in battlefield incomplete information environments.