A Research on Rapid Assessment of Cross-Domain Perceptual Fidelity for Practical Applications

With the rapid development of information technology, digital simulation technology plays a key role in improving training effects, providing accurate decision support, and promoting innovation and research and development. How to provide a more realistic training experience in simulation is particularly important. Guided by practical applications and based on cross-domain perception representation, this paper proposes a cross-domain perception fidelity evaluation algorithm oriented to application requirements. This method uses traditional methods to extract statistical features from the data and calculate feature similarity. Then, machine learning methods are used to extract structural, content and semantic features, effectively complete appearance and texture and calculate similarity. Finally, in order to ensure the effect in the actual application domain, the fidelity evaluation algorithm is designed based on the domain adaptation algorithm. Experimental results show that this method can well complete the cross-domain perceptual fidelity evaluation of simulation domain data and source domain data.