Radar target identification method based on adaptive neighborhood preserving projection

Abstract

The invention discloses a radar target identification method based on adaptive neighborhood preserving projection, which is mainly suitable for a coherent radar to classify and identify hovering helicopters and small sea surface targets. The method mainly comprises the following steps: firstly, constructing a variation difference distance, constructing a neighborhood for each data point in a training sample library according to the variation difference distance, and calculating a reconstruction weight matrix; then solving a multi-objective function optimization problem to obtain a projection matrix; after the projection matrix is obtained, carrying out feature extraction on data in the training sample library and the test sample library; and finally, carrying out hovering helicopter and sea surface small target classification identification by adopting a minimum distance classifier. According to the method, internal information of the data is deeply mined and fused into a feature extraction process, low-dimensional features containing rich identification information in the JEM data are extracted, and classification and identification of a hovering helicopter and a small sea surface target by a radar are realized.

Description

technical field [0001] The invention belongs to the technical field of radar target recognition. Background technique [0002] The classification and recognition of hovering helicopters and small targets on the sea is a major problem in radar target recognition. The current solution is to detect whether there are adjustment features for classification and recognition. During the detection process, long-dwell and high-repetition frequency detection of the target is required. . However, with the increase of dwell time and repetition frequency, the dimension of JEM data also increases significantly, becoming the most important factor affecting the target recognition rate. In pattern recognition and machine learning theory, for such problems, data dimensionality reduction algorithm is an effective and practical means, which can map data to low-dimensional space, remove its irrelevant information to seek the essential characteristics of data. [0003] The current mainstream dim...

AI Technical Summary

Problems solved by technology

[0003] The current mainstream dimensionality reduction algorithms such as principal component analysis, kernel principal component analysis, and linear discriminant analysis are limited to information processing in a single scale space, and fail to dig deep into the inherent structural characteristics of the data, resulting in the identification information contained in the extracted features. Fewer, less robust

Images