Domain adaptive target recognition method based on deep knowledge integration

Abstract

The invention belongs to the technical field of target recognition, and particularly relates to a domain adaptive target recognition method based on deep knowledge integration. According to the method, feature-level and decision-level deep knowledge integration is realized. A public mapping matrix and a specific mapping matrix are designed at a feature level to realize knowledge integration, and the robustness of target recognition performance is improved; and the public mapping matrix fully excavates public knowledge of heterogeneous features, and the specific mapping matrix retains specific knowledge of different features. Feature weights are designed at the decision level to quantify the importance degrees of different features, meanwhile, the feature weights are updated through online learning by means of target domain samples, the data distribution difference of different fields is overcome, and field self-adaptive target recognition is achieved. Therefore, the domain adaptive target recognition method based on deep knowledge integration provided by the invention is an intelligent domain adaptive target recognition method.

Description

technical field [0001] The invention belongs to the technical field of target recognition, and in particular relates to a field-adaptive target recognition method based on deep knowledge integration. Background technique [0002] Automatic target recognition technology can identify and classify targets based on sensor data, and plays an important role in military and civilian fields such as battlefield awareness and reconnaissance, terrain exploration, and automatic driving. With the continuous development of technology, target recognition methods with their own advantages have been proposed one after another. How to combine the advantages of different methods to improve the performance of target recognition has become one of the hot spots in the research of target recognition technology. [0003] Literature "Q.Yu, H.Hu, X.Geng, Y.Jiang and J.An,"High-Performance SARAutomatic Target Recognition Under Limited Data Condition Based on a DeepFeature Fusion Network,"in IEEE Acce...

AI Technical Summary

Problems solved by technology

However, this method does not explore the relationship between different features, ignoring that there are often both public knowledge and unique knowledge between different features

Literature "J. Zhang, M. Xing and Y. Xie," FEC: A Feature Fusion Framework for SAR Target Recognition Based on Electromagnetic Scattering Features and Deep CNN Features," in IEEE Transactions on Geoscience and Remote Sensing, vol.59, no.3 ,pp.2174-2187,March2021"proposed a decision-level knowledge integration method, splicing the results of different decision-making levels, and training a new classifier to combine the knowledge of different methods, but this method ignores different features There are differences in the importance of knowledge integration

At the same time, due to changes in the environment or the angle of observation of the target, there are differences in the distribution of data in the source domain and the target domain, and most current knowledge integration methods do not consider the problem of domain adaptation. resulting in changes that degrade the performance of the model in the target domain

Images