One-dimensional distance image stable identification method based on Euler kernel principle component analysis (KPCA)

Abstract

The invention discloses a one-dimensional distance image stable identification method based on Euler kernel principle component analysis (KPCA). The method comprises the following steps: first of all, extracting normalization frequency spectrum amplitude features from an actually measured one-dimensional distance image signal sample; then, mapping the normalization frequency spectrum amplitude features to a kernel space by use of an Euler kernel function, calculating a kernel matrix, and obtaining a principle component feature projection matrix through a principle component analysis (PCA) method, and obtaining kernel space feature principle components of the sample; and finally, performing feature identification by use of a support vector machine (SVM). According to the invention, same-dimension kernel space mapping is realized based on the Euler kernel function, the linear separability of one-dimensional distance image data is enhanced, the space dimensions are not increased, and the computational complexity is reduced. The kernel space feature principle components are extracted by use of the PCA method, the feature dimensions are further reduced, noise influences are decreased, and the method has a quite rapid processing speed of large-data-size signals, can maintain quite high identification precision in a noise environment and has obvious advantages compared to a conventional KPCA method.

Description

technical field [0001] The invention relates to a method for identifying a radar one-dimensional signal, in particular to a technology for quickly and accurately identifying a radar one-dimensional range image in an interference environment. Background technique [0002] Radar target recognition is an important research direction of radar signal processing. Radar high-resolution one-dimensional range profile (HRRP) reflects the distribution of target scattering points along the radar line-of-sight direction, which contains rich target structural features, and its inherent advantages are easy acquisition and small data volume. Therefore, target recognition based on HRRP has always been a research hotspot in the field of radar target recognition. Since the linear inseparability of large-angle HRRP data itself is more and more serious in the noise interference environment, some traditional classification and identification schemes such as the maximum correlation criterion clas...

AI Technical Summary

Problems solved by technology

[0005] Purpose of the invention: Aiming at the above-mentioned prior art, a robust recognition method for one-dimensional range images based on Euler kernel principal component analysis is proposed, which can obtain satisfactory recognition accuracy at a relatively small calculation cost in a noisy environment. When the classifier is fixed, the recognition accuracy can be increased by 3-4 percentage points compared with the traditional kernel principal component analysis method

Images