CNNKD-based radar HRRP small sample target identification method

Abstract

The invention discloses a CNNKD-based radar HRRP small sample target identification method. The method comprises: constructing a multi-category HRRP sample set; processing the HRRP sample set to obtain an effective HRRP sample set; constructing an in-library target HRRP training sample set by using the effective HRRP sample set; inputting the target HRRP training sample set in the library into a convolutional neural network for training to obtain a base model; obtaining a loss value according to the output of the base model and the sub-model; performing back propagation by using the loss value, and optimizing the sub-model according to an Adam optimization algorithm to obtain a feature extractor; performing feature extraction on the multi-category non-cooperative target small sample training set through a feature extractor to obtain feature data of the non-cooperative target; and training the feature data of the multiple categories of non-cooperative targets to obtain a classifier. According to the target recognition method, the accuracy of the target recognition model is effectively improved, only the shallow linear classifier needs to be retrained for a new non-cooperative target, a feature extractor does not need to be retrained, and the calculation expenditure is saved.

Description

technical field [0001] The invention belongs to the technical field of radar target recognition, in particular to a CNNKD-based radar HRRP small sample target recognition method. Background technique [0002] In specific fields such as military affairs, it is an extremely important issue to improve the recognition accuracy of radar for non-cooperative (outside library) targets. [0003] The development of radar HRRP (High Resolution Range Profile) non-cooperative target recognition technology is mainly limited by two aspects: First, due to the extremely low observation frequency of non-cooperative targets, the number of labeled samples is seriously insufficient, making non-cooperative Object recognition has become a typical small-sample recognition problem, which is still an open hot and difficult problem in the academic circle; It leads to the model learning problem of imbalanced samples in machine learning. [0004] Most of the traditional target recognition methods use ...

AI Technical Summary

Problems solved by technology

However, this method has the following shortcomings: 1. This method is only effective for the problem of the decline in the accuracy of small-sample target recognition in the database, and does not consider the problem of non-cooperative (outside the database) small-sample target recognition; 2. This method adopts the traditional method of LDA multi-batch calculation method for feature extraction, this method needs to perform feature extraction based on deep knowledge and experience in related fields, resulting in greater uncertainty

Images