High resolution SAR image classification method based on co-sparse model

A classification method and sparse model technology, applied in character and pattern recognition, instruments, computer components, etc., can solve problems such as insufficient utilization of learned analytic operators, limited accuracy of image classification, etc., and achieve fast speed and classification efficiency Improve the effect of accurate and detailed information

Active Publication Date: 2017-09-29
XIDIAN UNIV
View PDF4 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] Although the co-sparse analytical model has been applied to the field of image classification, and the method based on the co-sparse model has shown advantages in speed, but this type of method has not been applied in the field of high-resolution SAR image classification, and the image classification method solves When co-sparse coefficients are used, the soft threshold method is simply used, and the analytical operator of learning is not fully utilized, and the accuracy of image classification is limited.

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
  • High resolution SAR image classification method based on co-sparse model
  • High resolution SAR image classification method based on co-sparse model
  • High resolution SAR image classification method based on co-sparse model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0026] The present invention is a kind of high-resolution SAR image classification method based on co-sparse model, see figure 1 , including the following steps:

[0027] Step 1. Input the image to obtain the initial pixel value matrix: select a high-resolution SAR image to be classified, such as figure 2 For the image shown in (a), read the image with Matlab software. If the input image is a three-channel RGB image, it needs to be converted into a grayscale image. If it is a single-channel grayscale image, the image can be obtained directly. Gray value matrix, each element in the matrix corresponds to a pixel in the image.

[0028] Take all pixels as the center to intercept M 1 × M 1 For the convenience of subsequent data processing, each pixel block is expanded into a column. Traverse all pixels in the image to get the experimental sample X∈R M×LL , M is the number of pixels in a pixel block, LL is the number of pixels in the high-resolution SAR image to be classified,...

Embodiment 2

[0042] The method for classifying high-resolution SAR images based on the co-sparse model is the same as that in Embodiment 1, wherein in step 3, the method for combining the projected subgradient and the unified row norm compact frame is used to construct the analytical operator and includes the following steps:

[0043] 3.1. Construct the initial analytical operator Ω 0

[0044] Randomly generate an overcomplete matrix D=[d of size M×N 1 , d 2 ,...,d i ,...,d N ], where d i is a column vector, 1≤i≤N, N is the dimension of the analytical operator, the size of N is manually selected according to the needs but must be greater than M, and the initial analytical operator Ω is obtained 0 ,Ω 0 is the transpose matrix of the overcomplete matrix D, that is, Ω 0 =D T .

[0045] 3.2, by the subgradient projection formula Calculation parameter Ω G , 0≤i≤K max1 , K max1 In order to find the maximum number of iterations of the analytical operator, Ω i is the i-th analytical...

Embodiment 3

[0056] The method for classifying high-resolution SAR images based on the co-sparse model is the same as in Embodiment 1-2, wherein in step 4, the augmented Lagrangian method is used to solve the co-sparse coefficient Z problem: s.t.Z=ΩX, including the following steps:

[0057] 4.1. Given the initial pixel value matrix X 0 =X, analytical operator Ω, initial co-sparse coefficient Z 0 =ΩX 0 , the initial parameter matrix B 0 for size with Z 0 Equal zero matrix, constant parameter λ<1, constant coefficient γ<1.

[0058] 4.2. Calculate the pixel value matrix X according to the following formula i+1 , co-sparse coefficient Z i+1 and parameter matrix B i+1 , 0≤i≤K max2 , K max2 is the maximum number of iterations to find the co-sparse coefficients.

[0059]

[0060]

[0061] in

[0062] B i+1 =B i -(Z i+1 -ΩX i+1 )

[0063] 4.3. Repeat step 4.2 until ||Z i+1 -ΩX i+1 || 2 ≥ε or i>K max2 When the iterative process ends, ε is a constant error coefficient mu...

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 high resolution SAR image classification method based on a co-sparse model and solves a technical problem of high classification time complexity caused by utilizing only an integrated sparse model to represent images during SAR image classification in the prior art. The method comprises steps that an initial pixel value matrix X is selected in to-be-classified SAR images; an analysis operator is selected to learn an initial sample; a projection subgradient method and a unified row specification tight frame method are combined to learn an analysis operator omega; an augmented lagrangian method is utilized to solve a co-sparse coefficient Z; a co-sparse coefficient vector of a pixel block corresponding to each pixel point and a pixel value vector of the pixel block are combined to acquire a characteristic vector; classification is carried out based on an SVM classifier to acquire a prediction label of each pixel point characteristic vector of the whole graph; the prediction label result is displayed in a gray image. The method is advantaged in that sparse expressions of the images can be rapidly acquired, classification timeliness and classification accuracy of the SAR images can be guaranteed, and the method is for high resolution SAR image classification.

Description

technical field [0001] The invention belongs to the technical field of image processing, specifically for the classification of high-resolution SAR images, specifically a high-resolution SAR image classification method based on a co-sparse model, which is applied to the field of SAR image classification. Background technique [0002] With the gradual improvement of Synthetic Aperture Radar (SAR) imaging technology, more and more applications of SAR images in various fields have prompted further research and development of SAR image classification technology. Due to the coherent speckle noise in the original SAR image, the traditional image classification methods in the past are not applicable to SAR image classification. SAR image classification can be applied to resource detection, military reconnaissance, medical fields, crop growth monitoring, disaster hazard assessment, etc. The value and importance of SAR image applications, and SAR image classification methods need to ...

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
IPC IPC(8): G06K9/62
CPCG06F18/2411
Inventor 侯彪焦李成于竞竞王爽马晶晶马文萍冯婕张小华
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