Fisher feature selection method based on feature space distribution

A feature selection method and feature space technology, applied in the field of radar, can solve the problems of poor separability, inability to fully reflect the separability of feature component categories, small separability values, etc., and achieve target identification probability improvement , Improve the probability of target recognition and the effect of improving the identification rate

Active Publication Date: 2020-07-10
XIDIAN UNIV
View PDF4 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the traditional Fisher algorithm only pays attention to the overall Fisher value of the feature, and cannot fully reflect the problem of the separability of the feature component category; Cao Ben et al. Weighting is performed, and the entropy-weighted Fisher decision rate is applied to feature selection, so that it fully reflects the features from both the overall and the details.
This method improves the accuracy of feature selection to a certain extent, but in the process of target identification, the target is only two types of in-base and out-of-base, and this method degenerates into the traditional Fisher criterion
[0004]In the actual scene, there is a pose sensitivity problem on the HRRP of the high-resolution one-dimensional range image, which leads to the non-uniform multi-region aggregation distribution of the extracted target feature space. When the feature space presents a non-uniform multi-region aggregation distribution and the internal separability of each region is good, the feature separability value calculated by the current method will be very small, resulting in the feature with better separability being misjudged as Features with poor separability are removed from the identification feature set, thus affecting the identification probability

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Fisher feature selection method based on feature space distribution
  • Fisher feature selection method based on feature space distribution
  • Fisher feature selection method based on feature space distribution

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0025] Since its birth, radar has focused on detection and has made outstanding contributions in different fields. With the increasingly complex radar detection environment and the continuous development of modern technology, target recognition has become an important technology in the development of modern radar. Radar Automatic Target Recognition (RATR) technology is based on radar echo signals, extracts target features, and realizes automatic determination of target attributes, categories or types. Selecting features with good separability will greatly improve the probability of target recognition . At present, the commonly used feature selection is the Fisher criterion based on the distance within and between classes in the Filter algorithm, which is more intuitive and easy to operate. However, the calculation error of the separability of non-uniform multi-region aggregation distribution features is relatively large, and it is easy to misjudge features with good separabili...

Embodiment 2

[0036] The Fisher feature selection method based on feature space distribution is the same as embodiment 1, and the construction training template library original feature set X described in step 1 comprises the following steps:

[0037] 1a) Acquisition of target data: Collect N sets of high-resolution one-dimensional range images of targets measured by dual-polarization radars, which are co-polarization high-resolution one-dimensional range images S LL , cross-polarized high-resolution one-dimensional range image S RL .

[0038] 1b) m-δ decomposition: perform m-δ decomposition on the obtained high-resolution one-dimensional range image, and obtain the secondary scattering component V of the target respectively d , volume scattering component V v and the surface scattering component V s ; m-δ decomposition can be calculated according to Stokes vector:

[0039]

[0040] where S LL , S RL are the amplitudes of the co-polarization and cross-polarization channels respecti...

Embodiment 3

[0051] Embodiment 3: The Fisher feature selection method based on feature space distribution is the same as that of Embodiment 1-2. In step 2 of the present invention, each feature in the original feature set X is divided into regions, specifically including: generating K m K in a feature subspace region m Computed from the minimum sum of squared errors MSSE, denote the qth subspace region of the mth feature as F m,q =(θ m,q , η m,q ,ψ m,q ), where θ m,q Indicates the number of target features in the qth subspace region of the mth feature, η m,q Represents the target feature set of the qth subspace region of the mth feature, ψ m,q ={1,2} indicates the category of the sample in the subspace region q of the mth feature, 1 is the target in the library, 2 is the target outside the library, i=1,2...θ m,q , q=1,2...K m , the subspace region in the present invention is also the sub-block; the minimum MSSE formula is as follows:

[0052]

[0053] in, j=1,2...K m , K m T...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a Fisher feature selection method based on feature space distribution, solves the problem of accurate calculation of separable values of non-uniform multi-region aggregated distribution features of a feature space, and selects features with higher separability for radar target recognition. The method comprises the following steps: performing spatial region division on eachfeature in an original feature set; calculating a subspace region weight value and a Fisher value; calculating a Fisher value of each feature; and selecting an optimal feature set for radar target identification. According to the method, a feature space area is divided, the weight and Fisher value of each subspace area are calculated, and the Fisher value of each feature is obtained through weighting. According to the method, the calculation result of the Fisher value is more accurate, the target identification probability of the selected optimal feature set is higher, and the radar target identification rate is improved. Experiments also prove that the separability value calculation is more accurate, and the identification probability is better. The method is used for the recognition of aradar target.

Description

technical field [0001] The invention belongs to the technical field of radar, and particularly relates to radar target recognition, in particular to a Fisher feature selection method based on feature space distribution, which selects features with high separability for radar target recognition of radar. Background technique [0002] Radar target recognition includes target discrimination and target classification. Target identification technology is to further extract and select the characteristic information of the target on the basis of radar detection of the target, and finally realize the judgment of the target attribute and type. In the identification process, it is first necessary to perform feature extraction on the HRRP of objects in and out of the library. The distribution of different features is different, and the separability is different. If all features are used at the same time, on the one hand, the feature dimension will be very large, which will greatly re...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62G01S7/41
CPCG01S7/41G06F18/211G06F18/23213G06F18/213
Inventor 刘峥秦基凯王晶晶谢荣靳冰洋
Owner XIDIAN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap