Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

SAR image target recognition method based on multilayer probability statistics model

A probabilistic statistical model and target recognition technology, applied in the field of SAR image target recognition based on multi-layer probabilistic statistical model, can solve the non-negative feature extraction method that cannot obtain effect, does not use category information, and does not involve multi-layer probabilistic statistical model etc. to achieve the effect of reducing the complexity of network parameters, improving the performance and stability of target recognition, and good representation ability

Inactive Publication Date: 2017-12-01
XIDIAN UNIV
View PDF2 Cites 9 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In the current feature extraction and target recognition methods for SAR images, there is no non-negative feature extraction method based on a multi-layer probability statistical model; at the same time, the PGBN model is often used for text classification and topic extraction, and the PGBN model is an unsupervised multi- Layer probability statistics model does not use category information in training, so a large number of samples are needed to obtain more reliable results. In the case of a small number of samples, ideal results are often not obtained

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
  • SAR image target recognition method based on multilayer probability statistics model
  • SAR image target recognition method based on multilayer probability statistics model
  • SAR image target recognition method based on multilayer probability statistics model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0026] The common point of some existing methods for SAR image feature extraction is that when the input data is non-negative, the obtained global variables and latent variables still have negative values, while the pixels in the SAR image data are all positive values, so these The features extracted by the method cannot be well explained physically.

[0027] In addition, the existing method of approximately decomposing the input matrix into a dictionary and a non-negative weighted combination of hidden variables can extract the non-negative features inside the SAR image and increase the performance of SAR image target recognition, but a single-layer feature extraction method, The structure of the mined information is relatively simple.

[0028]In terms of modeling methods, traditional deep networks such as Deep Belief Network (DBN) use binary data modeling for hidden variables, and have limited modeling capabilities for other forms of data. The layer stacking strategy trains...

Embodiment 2

[0040] The SAR image target recognition method based on the multi-layer probability statistical model is the same as embodiment 1, the parameters of the Poisson gamma belief network are initialized in the step (2), that is, the parameters of the multi-layer probability statistical model are initialized, and the initialized variables have the global value of the model parameters, hidden variable parameters, and hyperparameters of each prior distribution, such as Φ (l,C) ,θ (l,C) ,r C ,γ 0 C ,c 0 C , η (l,C) ,p (2 ,C) ,c (3,C) ,a 0 C ,e 0 C ,b 0 C ,f 0 C ,K l C , hyperparameter a 0 C , b 0 C , e 0 C , f 0 C No need to update, other parameters need to be updated in the process of model training and testing, in this example for the global variable Φ (l,C) and latent variable θ (l,C) The initialization of includes the following steps:

[0041] Layer-l global variables for objects of class C Initialization proceeds as follows:

[0042] φ a (l,C) ~Di...

Embodiment 3

[0051] The SAR image target recognition method based on the multi-layer probability statistical model is the same as embodiment 1-2, the training multi-layer probability statistical model in step 3, see figure 2 , including the following steps:

[0052] (3a) Input various training samples, and set the number of network layers, the dimensions of the input layer and each hidden layer, and the number of iterations M of various training samples 1 .

[0053] (3b) Calculate each layer augmentation matrix, intra-layer augmentation matrix and inter-layer augmentation matrix in the process of uplink Gibbs sampling: use the Naive Bayesian criterion to independently train the multi-layer probability statistics model for each type of training data , in each iteration of various multi-layer probability statistical models, starting from the bottom layer, training layer by layer from the lower layer to the upper layer, and calculating the augmented matrix and intra-layer Augmented matrix an...

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 an SAR image target recognition method based on a multilayer probability statistics model, so as to solve the problem that when input SAR image data has non-negativity in the existing method, only single-layer non-negative features of an SAR image can be extracted. The method comprises steps: a training set and a test set for the multilayer probability statistics model are built; parameters of the multilayer probability statistics model are initialized; a joint learning mode is used to train the multilayer probability statistics model, and a global variable is saved; the joint learning mode is used to test the multilayer probability statistics model, and all parameters of the model are obtained for subsequent recognition; and the parameters obtained by test are used for target recognition. Based on the multilayer probability statistics model, the joint learning mode is used to train and test the model, the SAR image feature extraction and target recognition are carried out, the model parameter complexity is reduced, the multilayer non-negative features of the SAR image are extracted, the SAR image target recognition performance and the stability are improved, and the method of the invention is used for carrying out target recognition on the SAR image.

Description

technical field [0001] The invention belongs to the technical field of SAR image target recognition, and in particular relates to SAR image feature extraction, in particular to a SAR image target recognition method based on a multi-layer probability statistical model, which can be used for SAR image feature extraction and SAR image target recognition. Background technique [0002] SAR image target recognition has important military value and commercial value, and has been a hot research topic at home and abroad. Although the field of target recognition has been greatly developed in recent years, it is still a challenging task to complete SAR image target recognition accurately and quickly. Feature extraction is a very critical step in SAR image target recognition. The quality of the extracted features directly affects the recognition performance of SAR image targets. Therefore, it is necessary to carry out meaningful feature extraction on SAR images to improve the performanc...

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): G06K9/00G06K9/62
CPCG06V20/13G06F18/2411G06F18/2415
Inventor 陈渤张梦娇郭丹丹
Owner XIDIAN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com