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

Spectral response design method based on neural network

A spectral response, neural network technology, applied in the fields of photoelectric detection, spectral analysis, optoelectronic devices and optical imaging, it can solve the problem that the reverse design network is difficult to achieve the fitting accuracy of the forward prediction network, and avoid the dependence of completeness. , easy to evaluate, and the effect of improving design accuracy

Active Publication Date: 2020-08-18
ZHEJIANG UNIV +1
View PDF4 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Although there have been recent attempts to use deep learning for reverse design, its stability needs to be improved
Reverse design network based on deep learning is difficult to achieve the fitting accuracy of forward prediction network

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
  • Spectral response design method based on neural network
  • Spectral response design method based on neural network
  • Spectral response design method based on neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0053] A neural network-based spectral response design method implemented, the specific scheme is as follows:

[0054] (1) Determine the target spectral range λ min =420nm, λ max =720nm and spectral resolution Δλ=1nm, then λ min , lambda max The number of spectral channels between N=(λ max -λ min ) / Δλ=300. It is determined that the spectral modulator is an optical film filter, the number of which is M=50.

[0055](2) Construct a neural network, which is called a spectral perception network. The first layer of the network is set as a linear connection layer with an input unit number of 300 and an output unit number of 50, and the linear connection layer is called a hardware layer. Then the weight term w ji Represents the spectral response value of the j-th filter at the i-th spectral channel.

[0056] (3) Set the software layer network after the hardware layer. The specific setting method is to set a layer of batch normalization layer (BatchNormalization Layer), then s...

Embodiment 2

[0062] A neural network-based spectral response design method implemented, the specific scheme is as follows:

[0063] (1) Determine the target spectral range λ min =380nm, λ max =780nm and spectral resolution Δλ=2nm, then λ min , lambda max The number of spectral channels between N=(λ max -λ min ) / Δλ=200. It is determined that the spectral modulator is a photonic crystal device, the number of which is M=16.

[0064] (2) Construct an artificial neural network, and the first layer of the network is set as the number of input units is 200, the number of output units is a linear connection layer of 16, and the linear connection layer is called the hardware layer. Then the weight w between the i-th (i=1,2,…,200) input unit and the j-th (j=1,2,…,16) output unit of the hardware layer ji Represents the response value of the jth spectral modulator at the ith spectral channel.

[0065] (3) Set the software layer network after the hardware layer. The specific setting method is 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 spectral response design method based on a neural network. The combination of spectral modulators with different spectral responses can be regarded as a linear connection layer of the neural network. Meanwhile, another neural network is used for adding limiting conditions to generation of the linear connection layer parameters; through training of the whole neural network, the optimization of spectral response of the spectral modulator can be achieved, corresponding design parameters can be directly generated, and therefore the method is applied to scenes such as spectral recognition, spectral detection and spectral imaging. Compared with a traditional optimization method, the method has the advantages that the optimization target is clear, the advantages and disadvantages of the optimization result can be evaluated more easily, and the optimization effect is better; collaborative design of hardware and software is achieved, and design parameters can be flexibly adjusted to adapt to different application scenes; meanwhile, the design process excessively depending on experience is avoided, and a high-universality design framework is provided for reverse design of the spectrum modulation device.

Description

technical field [0001] The invention belongs to the fields of photoelectric detection, spectrum analysis, photoelectric devices and optical imaging. The invention can be applied to the spectral response design of various spectrometers, spectral imaging devices, and spectral detectors, so as to be applied to spectral analysis and high-resolution imaging, agricultural and agricultural product detection, and the fields of hygiene, medicine, and health. Background technique [0002] Spectral detection technologies such as spectrometers and spectral imaging provide great convenience for people to understand material properties and identify targets. Traditional spectral detection equipment is difficult to enter consumers' lives due to its complex structure, large size, bulky, expensive and other shortcomings. In recent years, with the introduction of software algorithms such as compressed sensing and machine learning into spectral detection, a large number of computational spectr...

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): G01J3/28G06N3/08G06N3/063
CPCG01J3/28G01J3/2823G06N3/063G06N3/08
Inventor 郝翔宋洪亚张文屹刘旭
Owner ZHEJIANG 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