Radar target adaptive reverse truncation intelligent identification method

Abstract

The invention discloses a radar target adaptive reverse truncation intelligent identification method, and belongs to the field of radar signal processing. In order to solve the problem of poor one-dimensional range profile multi-class identification precision under a small sample condition, an identification framework of a fuzzy truncated cosine loss basic learner combined with a normalized exponential loss constraint meta learner is constructed, and based on the potential difference between meta training task source domain data and meta test task target domain data, the one-dimensional range profile multi-class identification accuracy is improved. Carrying out convolutional neural network loss function difference design in stages, and fully mining personalized features of small samples of a target domain while extracting general features of source domain data; besides, by considering mutual influence of loss of different tasks in the same meta-test task set, regularized constraint stable updating of meta-learner parameters is achieved, the regularity of the calculation process is high, the method can be suitable for effective migration and fusion of meta-training task learning experiences of different sources, the overall reference efficiency of the meta-training task experiences is improved, and the user experience is improved. The popularization and application values are realized.

Description

1. Technical field [0001] The invention belongs to the field of radar signal processing, in particular to a radar target adaptive reverse truncation intelligent identification method. 2. Background technology [0002] High Resolution Range Profile (HRRP) is the projection of the main scattering point of the target in the direction of radar irradiation, which reflects the relative position information of the target scattering point, and reflects the partial structure information of the target to a certain extent. Therefore, it has broad application prospects in the field of target recognition. For cooperative targets, a large number of target HRRP samples can be obtained, but in practical applications, the targets targeted by the radar are often non-cooperative targets, and it is difficult to obtain a sufficient number of non-cooperative target HRRP data. Therefore, HRRP recognition of radar targets under the condition of small samples has become one of the research hotspots...

AI Technical Summary

Problems solved by technology

[0008] Aiming at the problem of poor HRRP multi-category recognition accuracy of radar targets under the condition of small samples, how to build a suitable meta-learning recognition framework, decouple the characteristics of a single task from the commonality between tasks, and design a suitable basic learner and meta-learner for a single task. The characteristics and the commonality between tasks are transferred through experience, and the loss function design of the recognition model is carried out in the pre-training stage and the fine-tuning stage respectively, so as to realize the effective small-sample recognition of radar targets; for the construction of the basic learner, how to design a model with strong generalization ability Loss function, to avoid the defect that the commonly used cross-entropy loss function tends to cause the model to obtain overconfident classification results and tend to converge to a local minimum, so as to effectively extract the general features of the target data in the source domain, and reserve the necessary versatility for subsequent meta-test tasks experience to be migrated; how to design a regularized parameter update method for the construction of a meta-learner, while introducing the differential learning experience of different meta-training tasks, fully consider the mutual influence of different task losses in the same meta-test task set, and ensure that the parameters The update process effectively integrates different task experiences, makes full use of the different characteristics of different task learning experiences, realizes the regularized constraint transfer of different task learning experiences, and is used to guide new recognition tasks to perform effective small-sample recognition and improve the performance of new tasks. The recognition accuracy of small-sample targets; how to design a suitable loss function for the recognition model construction of meta-testing tasks, avoid the problem of the recognition rate decline of small-sample targets caused by the unbalanced number of samples in different categories, and effectively extract small-sample targets The personality characteristics of the data enhance the matching degree of the meta-test task recognition model to the recognition task

Images