[go: up one dir, main page]

CN118817635B - Multispectral channel gas classification and quantitative detection method and system thereof - Google Patents

Multispectral channel gas classification and quantitative detection method and system thereof Download PDF

Info

Publication number
CN118817635B
CN118817635B CN202411281550.5A CN202411281550A CN118817635B CN 118817635 B CN118817635 B CN 118817635B CN 202411281550 A CN202411281550 A CN 202411281550A CN 118817635 B CN118817635 B CN 118817635B
Authority
CN
China
Prior art keywords
gas
sub
spectral
band
beams
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202411281550.5A
Other languages
Chinese (zh)
Other versions
CN118817635A (en
Inventor
王建宇
王一婕
刘世界
李茼佟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Institute of Advanced Studies of UCAS
Original Assignee
Hangzhou Institute of Advanced Studies of UCAS
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hangzhou Institute of Advanced Studies of UCAS filed Critical Hangzhou Institute of Advanced Studies of UCAS
Priority to CN202411281550.5A priority Critical patent/CN118817635B/en
Publication of CN118817635A publication Critical patent/CN118817635A/en
Application granted granted Critical
Publication of CN118817635B publication Critical patent/CN118817635B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G02—OPTICS
    • G02B—OPTICAL ELEMENTS, SYSTEMS OR APPARATUS
    • G02B27/00—Optical systems or apparatus not provided for by any of the groups G02B1/00 - G02B26/00, G02B30/00
    • G02B27/10—Beam splitting or combining systems
    • G02B27/106—Beam splitting or combining systems for splitting or combining a plurality of identical beams or images, e.g. image replication
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/01—Arrangements or apparatus for facilitating the optical investigation
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
    • G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
    • G01N21/3504—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light for analysing gases, e.g. multi-gas analysis
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G—PHYSICS
    • G02—OPTICS
    • G02B—OPTICAL ELEMENTS, SYSTEMS OR APPARATUS
    • G02B27/00—Optical systems or apparatus not provided for by any of the groups G02B1/00 - G02B26/00, G02B30/00
    • G02B27/10—Beam splitting or combining systems
    • G02B27/12—Beam splitting or combining systems operating by refraction only
    • G02B27/123—The splitting element being a lens or a system of lenses, including arrays and surfaces with refractive power
    • G—PHYSICS
    • G02—OPTICS
    • G02B—OPTICAL ELEMENTS, SYSTEMS OR APPARATUS
    • G02B5/00—Optical elements other than lenses
    • G02B5/20—Filters
    • G02B5/201—Filters in the form of arrays
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/09—Supervised learning
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N2021/1765—Method using an image detector and processing of image signal

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Analytical Chemistry (AREA)
  • Chemical & Material Sciences (AREA)
  • Optics & Photonics (AREA)
  • Pathology (AREA)
  • Immunology (AREA)
  • Biochemistry (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Multimedia (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)
  • Spectrometry And Color Measurement (AREA)

Abstract

The invention belongs to the field of optical gas imaging detection, and discloses a multi-spectral channel gas classification and quantitative detection method and a system thereof, wherein an incident beam is divided into a plurality of sub-beams and filtered to obtain N-1 sub-beams with different narrowband spectral bands and a broadband all-pass band sub-beam; the method comprises the steps of converting sub-beams of N wave bands into digital image signals of multi-band channels for simultaneous imaging, carrying out image registration alignment on the digital image signals of the multi-band channels for simultaneous imaging and images of different channels, outputting a spectrum image cube, carrying out classification detection on the spectrum image cube to obtain pixel coordinate areas and type information of gas distribution, carrying out concentration equation quantitative calculation on the pixel coordinate areas and the type information of the gas distribution, carrying out pseudo-color enhancement on the gas, and outputting spatial distribution of gas concentration. The invention can obviously improve the application level of the optical gas detection technology and provide a more accurate and reliable solution for the fields of environmental monitoring and the like.

Description

Multispectral channel gas classification and quantitative detection method and system thereof
Technical Field
The invention relates to the field of optical gas imaging detection, in particular to a multispectral channel gas classification and quantitative detection method and a multispectral channel gas classification and quantitative detection system.
Background
Optical gas imaging detection technology has been widely used in the fields of environmental monitoring, industrial process control, medical health, etc. in recent years due to its characteristics of non-invasiveness, high sensitivity, and imaging and real-time response. Conventional optical gas detection methods are based primarily on the light absorption properties of gas molecules, and infer the presence and concentration of a gas by measuring the absorption intensity at a single specific wavelength. However, these methods have exposed many problems in practical use, particularly when applied to chemical sites, it is difficult to accurately distinguish the kind of leakage gas and quantify the concentration thereof. When the leaking gas is of unknown kind, conventional single wavelength or narrow band light absorption techniques are struggling to cope with this complex situation. Single wavelength detection makes it difficult to distinguish between individual gas components because the absorption spectra of different gases may overlap at certain wavelengths. In this case, error accumulation leads to inaccurate detection results, and even missing detection or false alarm may occur. In addition to the identification of gas species, accurate measurement of gas concentrations is also a challenge in optical gas detection techniques. There is no optical gas imaging detection technology capable of achieving concentration quantification in the industry.
Disclosure of Invention
The invention aims to provide a multispectral channel gas classification and quantitative detection method and a multispectral channel gas classification and quantitative detection system, so as to solve the problems in the background technology.
In order to achieve the above purpose, the present invention provides the following technical solutions:
a multi-spectral channel gas classification and quantitative detection method comprises the following steps:
Step 1, dividing an incident light beam into a plurality of sub-light beams through a multispectral light splitting module, and filtering to obtain N-1 sub-light beams with different narrowband spectrum bands and a broadband all-pass band sub-light beam;
step 2, converting the sub-beams of the N wave bands obtained in the step 1 into digital image signals of multi-band channels for simultaneous imaging through an infrared detector;
step 3, through an image distortion correction and registration unit, digital image signals which are simultaneously imaged by a multiband channel and are output by an infrared detector are subjected to distortion correction and are subjected to image registration alignment with images of different channels, and a spectrum image cube is output;
Step 4, carrying out pixel-level gas classification detection on the registered spectrum image cube output in the step 3 through a gas classification unit to obtain a pixel coordinate area and type information of gas distribution;
And 5, quantitatively solving a concentration equation of the pixel coordinate area and the type information of the gas distribution obtained in the step 4 through a gas quantitative unit, performing pseudo-color enhancement on the gas based on the gas concentration, and outputting the spatial distribution of the gas concentration.
Further, the step 1 specifically includes the following steps:
Step 1.1, an incident light beam is split into N sub-light beams through an aperture array module, wherein the aperture array module is a lens array formed by arranging N millimeter-level lenses;
Step 1.2, reserving one of N sub-beams, filtering the remaining N-1 sub-beams by using a filter array module to obtain N-1 sub-beams with different narrowband spectrum bands and a broadband all-pass band sub-beam, wherein the filter array module is a filter array formed by arranging N-1 narrowband band-pass filters with different center wavelengths.
Further, the step 3 specifically includes the following steps:
Step 3.1, cutting a digital image signal of simultaneous multi-band channel imaging into N sub-images, wherein the N sub-images comprise N-1 different narrowband spectrum band sub-images and one broadband all-pass band sub-image;
Step 3.2, registering the cut images by using an ORB algorithm;
And 3.3, arranging the registered multi-channel images into a spectrum image cube with N channels and outputting the spectrum image cube.
Further, the step 4 specifically includes the following steps:
step 4.1, introducing a normalized gas absorption spectrum database into a gas classification unit;
Step 4.2, calculating the registered spectrum image cube output in step 3 according to the following process:
The spectrum image cube is marked as I s, I s (x, y, c) represents a point with x-axis, y-axis and c-channel, the value of the point is a coordinate DN value, I s (x, y) represents a vector consisting of n DN values of I s(x,y,1)、Is(x,y,2)...Is (x, y, n), after the vector is normalized, the vector is classified by using a Bayesian optimization-based neural network, and the Bayesian optimization-based neural network outputs a gas type I to which each pixel point finally belongs and a probability p of an ith gas;
And 4.3, outputting pixel coordinate areas marked as gas distribution and type information.
The neural network further comprises a five-layer network structure, and comprises an input layer, three hidden layers and an output layer, wherein the number of neurons of the input layer is the number of spectrum channels, the number of neurons of the output layer is the number I of gas types needing to be classified, the Bayesian optimizer optimizes the super-parameters of the neural network based on initial values, the optimized super-parameters are as follows, the size of a first hidden layer is 71, the size of a second hidden layer is 164, regularization strength is 7.99425566291672e-06, an activation function is tanh, and a loss function is a cross entropy loss function.
Further, the step 5 specifically includes the following steps:
step 5.1, extracting DN values corresponding to all channel area points according to the pixel coordinate areas of the gas distribution output in the step 4;
And 5.2, carrying out difference on DN values of the gas pixels and DN values of surrounding background pixels to obtain delta DN (lambda), and selecting at least three narrow-band spectrum wavelength channels L a、Lb、Lc to obtain the following gas concentration equations respectively, and solving the inversion gas concentration by solving the following three equations:
Wherein e is a natural constant, λ a、λb、λc is the wavelength of the narrow-band spectral wavelength channel L a、Lb、Lc, τ s(λa)、τs(λb)、τs(λc) is the total band response of the narrow-band spectral wavelength channel L a、Lb、Lc, τ 2(λa)、τ2(λb)、τ2(λc) is the atmospheric transmittance of the wavelength λ a、λb、λc, which can be obtained by querying a Modtran radiation transmission model, K (λ a)、K(λb)、K(λc) is the absorption rate of the gas to be detected of the narrow-band spectral wavelength channel L a、Lb、Lc, which can be obtained by querying a Hitran database, ε (λ a)、ε(λb)、ε(λc) is the background emissivity of the narrow-band spectral wavelength channel L a、Lb、Lc, α (λ a)、α(λb)、α(λc) is the background emissivity of the narrow-band spectral wavelength channel L a、Lb、Lc, which is calculated by a planck formula of the wavelength λ a、λb、λc, β (λ a)、β(λb)、β(λc) is calculated by a planck formula of the wavelength λ a、λb、λc, and is the column density c·l of the gas, the background temperature T bg, the gas temperature T gas, which can be solved by substituting specific values into a ternary overrun equation after solving;
and 5.3, performing pseudo-color enhancement processing on the gas region of the broadband full-pass band subimage according to the concentration of the gas, and outputting the spatial distribution of the concentration of the gas.
The invention also provides a multispectral channel gas classification and quantitative detection system, which is used for realizing the multispectral channel gas classification and quantitative detection method, and comprises the following steps:
the multi-spectrum light splitting module is used for subdividing an incident light beam into sub-light beams with a plurality of wave bands;
The infrared detector is used for converting the sub-beams obtained by the multispectral light splitting module into digital image signals visible to naked eyes;
The gas classifying and quantifying module comprises an image distortion correcting and registering unit, a gas classifying unit and a gas quantifying unit, wherein the image distortion correcting and registering unit is used for correcting distortion of digital image signals output by an infrared detector and registering and aligning the digital image signals with images of different channels to output a multi-channel spectrum image cube, the gas classifying unit is used for classifying and detecting the gas of a pixel level of the registered spectrum image cube output by the image distortion correcting and registering unit to obtain pixel coordinate areas and type information of gas distribution, and the gas quantifying unit is used for quantitatively solving a concentration equation of the pixel coordinate areas and the type information of the gas distribution obtained by the output of the gas classifying unit, carrying out pseudo-color enhancement on the gas based on the gas concentration and outputting the spatial distribution of the gas concentration.
Further, the multispectral light splitting module comprises the following sub-modules:
the aperture array module is used for dividing an infrared incident beam into N sub-beams, and is a lens array formed by arranging N millimeter-level lenses;
the narrow-band bandpass filter array module is used for filtering the N-1 sub-beams to obtain N-1 sub-beams with different narrow-band spectrum bands, and the filter array module is a filter array formed by arranging N-1 narrow-band bandpass filters with different center wavelengths.
Compared with the prior art, the multi-spectral channel detection method has the advantages that the multi-spectral channel detection method can identify gas types and detect gas concentration compared with the traditional single-wavelength imaging detection method. The main innovation of multispectral channel detection is that by analyzing the absorption characteristics at different wavelengths, the presence of different gases can be distinguished more accurately. Even under the condition of overlapping absorption spectrums, the comprehensive utilization of multi-wavelength information can also obviously improve the identification accuracy. The redundant information of the multispectral channels enables the system to be more robust in the face of environmental noise and interference, and influences of non-target gases can be effectively filtered.
Through the innovations, the multispectral channel gas classification and quantitative detection method and the multispectral channel gas classification and quantitative detection system can remarkably improve the application level of the optical gas detection technology, and provide a more accurate and reliable solution for the fields of environmental monitoring, industrial control, public safety and the like.
Drawings
FIG. 1 is a flow chart of a multi-spectral channel gas classification and quantitative detection method of the present invention.
FIG. 2 is a schematic diagram of the operation of the gas classifying and quantifying module according to the present invention.
FIG. 3 is a schematic diagram of a multi-spectral channel gas classification and quantitative detection system according to the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1-2, a multi-spectral channel gas classification and quantitative detection method includes the following steps:
Step 1, dividing an incident light beam into a plurality of sub-light beams through a multispectral light splitting module, and filtering to obtain N-1 sub-light beams with different narrowband spectrum bands and a broadband all-pass band sub-light beam, wherein the method specifically comprises the following steps:
Step 1.1, an incident light beam is split into N sub-light beams through an aperture array module, wherein the aperture array module is a lens array formed by arranging N millimeter-sized lenses, and the N millimeter-sized lenses correspond to the N sub-light beams.
And 1.2, reserving one of N sub-beams, and filtering the remaining N-1 sub-beams by using a filter array module to obtain N-1 sub-beams with different narrowband spectrum bands and a broadband all-pass band sub-beam. The filter array module is a filter array formed by arranging N-1 narrow-band bandpass filters with different center wavelengths, and the rest N-1 sub-beams are filtered by the N-1 narrow-band bandpass filters with different center wavelengths in a one-to-one correspondence manner.
And 2, converting the sub-beams with the N wave bands obtained in the step 1 into digital image signals of multi-band channels for simultaneous imaging through an infrared detector.
Step 3, through the image distortion correction and registration unit, the digital image signals which are simultaneously imaged by the multiband channels and output by the infrared detector are subjected to distortion correction and are subjected to image registration alignment with images of different channels, and a spectrum image cube is output, and the method specifically comprises the following steps:
and 3.1, cutting the digital image signal of the multi-band channel simultaneous imaging into N sub-images, wherein the N sub-images comprise N-1 different narrow-band spectrum band sub-images and one broadband all-pass band sub-image.
Step 3.2, registering the cut images using ORB (Oriented FAST and Rotated BRIEF) algorithm.
And 3.3, arranging the registered multi-channel images into a spectrum image cube with N channels and outputting the spectrum image cube.
And 4, carrying out pixel-level gas classification detection on the registered spectrum image cube output in the step 3 through a gas classification unit to obtain a pixel coordinate area and type information of gas distribution, wherein the method specifically comprises the following steps of:
Step 4.1, introducing a normalized gas absorption spectrum database into the gas classification unit.
Step 4.2, calculating the registered spectrum image cube output in step 3 according to the following process:
The spectrum image cube is denoted as I s, I s (x, y, c) represents a point with x-axis, y-axis and c-channel, the value of which is a coordinate DN value, I s (x, y) represents a vector consisting of n DN values of I s(x,y,1)、Is(x,y,2)...Is (x, y, n), and after normalizing the vector, the vector is classified by using a Bayesian optimization-based neural network, and each pixel point is output to the gas class I and the probability p of the ith gas.
The neural network has a five-layer network structure and comprises an input layer, three hidden layers and an output layer, wherein the number of neurons of the input layer is the number of spectrum channels, the number of neurons of the output layer is the number I of gas types needing to be classified, the Bayesian optimizer optimizes the super-parameters of the neural network based on initial values, the optimized super-parameters are as follows, the size of a first hidden layer is 71, the size of a second hidden layer is 164, regularization intensity (lambda) is 7.99425566291672e-06, an activation function is tanh, and a loss function is a cross entropy loss function.
And 4.3, outputting pixel coordinate areas marked as gas distribution and type information, and outputting the result as shown in fig. 2.
And 5, quantitatively solving a concentration equation of the pixel coordinate area and the type information of the gas distribution obtained in the step 4 through a gas quantifying unit, performing pseudo-color enhancement on the gas based on the gas concentration, and outputting the spatial distribution of the gas concentration, wherein the method specifically comprises the following steps of:
and 5.1, extracting DN values corresponding to all channel area points according to the pixel coordinate areas of the gas distribution output in the step 4.
And 5.2, carrying out difference on DN values of the gas pixels and DN values of surrounding background pixels to obtain delta DN (lambda), and selecting at least three narrow-band spectrum wavelength channels L a、Lb、Lc to obtain the following gas concentration equations respectively, and solving the inversion gas concentration by solving the following three equations:
Wherein e is a natural constant, the value of which is about 2.718281828459045, λ a、λb、λc is the wavelength of the narrow-band spectral wavelength channel L a、Lb、Lc, τ s(λa)、τs(λb)、τs(λc) is the total band response of the narrow-band spectral wavelength channel L a、Lb、Lc, τ 2(λa)、τ2(λb)、τ2(λc) is the atmospheric transmittance of the wavelength λ a、λb、λc, which can be obtained by querying a Modtran radiation transmission model, K (λ a)、K(λb)、K(λc) is the absorption rate of the gas to be detected of the narrow-band spectral wavelength channel L a、Lb、Lc, which can be obtained by querying a Hitran database, ε (λ a)、ε(λb)、ε(λc) is the background emissivity of the narrow-band spectral wavelength channel L a、Lb、Lc, α (λ a)、α(λb)、α(λc) is the background emissivity of the narrow-band spectral wavelength channel L a、Lb、Lc, which is calculated by a planck formula of the wavelength λ a、λb、λc, β (λ a)、β(λb)、β(λc) is calculated by a planck formula of the wavelength λ a、λb、λc, and the unknown amounts are the column density c·l of the gas, the background temperature T bg, and the gas temperature T gas, and the column concentration, background temperature, of the gas and the gas temperature of the gas can be solved by introducing specific values into a solved ternary equation system.
And 5.3, performing pseudo-color enhancement processing on the gas region of the broadband full-pass band subimage according to the concentration of the gas, and outputting the spatial distribution of the concentration of the gas.
Referring to fig. 3, a multi-spectral channel gas classification and quantitative detection system is used for implementing the multi-spectral channel gas classification and quantitative detection method as described above, and includes a multi-spectral light-splitting module, an infrared detector, and a gas classification and quantitative module.
The multi-spectrum light splitting module is used for subdividing an incident light beam into sub-beams with a plurality of wave bands and comprises an aperture array module and a narrow-band pass filter array module, wherein the aperture array module is used for dividing an infrared incident light beam into N sub-beams, and the narrow-band pass filter array module is used for filtering N-1 sub-beams to obtain N-1 sub-beams with different narrow-band spectrum wave bands and a broadband all-pass wave band sub-beam. The aperture array module is a lens array formed by arranging N millimeter-sized lenses. The filter array module is an optical filter array formed by arranging N-1 narrow-band bandpass filters with different center wavelengths.
The infrared detector is used for converting the sub-beams obtained by the multispectral light splitting module into digital image signals visible to naked eyes.
The gas classifying and quantifying module is a computer program, is stored and operated in an upper computer, and comprises an image distortion correcting and registering unit, a gas classifying unit and a gas quantifying unit.
The image distortion correction and registration unit is used for correcting distortion of the digital image signal output by the infrared detector and performing image registration alignment with images of different channels, and outputs a spectrum image cube, and the image distortion correction and registration unit is a computer program, and specific implementation of the image distortion correction and registration unit can be seen in the step 3 and is not repeated here.
The gas classification unit is used for classifying and detecting the gas at the pixel level on the spectrum image cube after the registration output by the image distortion correction and registration unit, so as to obtain the pixel coordinate area and the kind information of the gas distribution, and the gas classification unit is a computer program, and the specific implementation of the gas classification unit can be seen in the step 4 and is not repeated here.
The gas quantifying unit is used for quantitatively solving a concentration equation of the pixel coordinate area and the type information of the gas distribution obtained by the output of the gas classifying unit, performing pseudo-color enhancement on the gas based on the gas concentration, and outputting the spatial distribution of the gas concentration, and the specific implementation of the gas quantifying unit is a computer program, and can be seen in the step 5 and is not repeated herein.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (5)

1.一种多光谱通道气体分类和定量检测方法，其特征在于，包括以下步骤:1. A multi-spectral channel gas classification and quantitative detection method, characterized in that it comprises the following steps: 步骤1，通过多光谱分光模块，将入射光束分为多个子光束并滤波，得到N-1个不同窄带光谱波段的子光束和一个宽带全通波段子光束，N为大于2的正整数；Step 1, using a multi-spectral spectrometer module, splitting the incident light beam into a plurality of sub-beams and filtering them to obtain N-1 sub-beams with different narrow-band spectral bands and one broadband all-pass band sub-beam, where N is a positive integer greater than 2; 步骤2，通过红外探测器将步骤1得到的N个波段的子光束转化为多波段通道同时成像的数字图像信号；Step 2, converting the N-band sub-beams obtained in step 1 into digital image signals simultaneously imaged by multi-band channels through an infrared detector; 步骤3，通过图像畸变矫正和配准单元，将红外探测器输出的多波段通道同时成像的数字图像信号进行畸变矫正并与不同通道的图像进行图像配准对齐，输出光谱图像立方体；Step 3, through the image distortion correction and registration unit, the digital image signal of the multi-band channels simultaneously imaged by the infrared detector is subjected to distortion correction and image registration and alignment with the images of different channels, and the spectral image cube is output; 步骤4，通过气体分类单元，将步骤3输出的配准后的光谱图像立方体进行像素级的气体的分类检测，得到气体分布的像素坐标区域和种类信息，具体包括以下过程：Step 4, through the gas classification unit, the registered spectral image cube output in step 3 is subjected to pixel-level gas classification detection to obtain the pixel coordinate area and type information of the gas distribution, which specifically includes the following processes: 步骤4.1，在气体分类单元中导入归一化的气体吸收光谱数据库；Step 4.1, import the normalized gas absorption spectrum database into the gas classification unit; 步骤4.2，对于步骤3输出的配准后的光谱图像立方体按如下过程计算：Step 4.2, the registered spectral image cube output from step 3 is calculated as follows: 将光谱图像立方体记为Is，则Is(x,y,c)表示横坐标为x，纵坐标为y，通道为c的点，其值是坐标DN值，Is(x,y)表示一个由Is(x,y,1)、Is(x,y,2)...Is(x,y,n) n个DN值组成的向量，将该向量归一化后，通过使用基于贝叶斯优化的神经网络对该向量进行分类，基于贝叶斯优化的神经网络输出每个像素点最终所属气体种类i和属于第i种气体的概率p；The spectral image cube is denoted as Is, then Is (x,y,c) represents a point with abscissa x, ordinate y, and channel c, and its value is the coordinate DN value. Is (x,y) represents a vector composed of n DN values, Is (x,y,1), Is (x,y,2)... Is (x,y,n). After normalizing the vector, the vector is classified by using a neural network based on Bayesian optimization. The neural network based on Bayesian optimization outputs the gas type i to which each pixel finally belongs and the probability p of belonging to the i-th gas. 步骤4.3，输出标记为气体分布的像素坐标区域和种类信息；Step 4.3, output pixel coordinate area and type information marked as gas distribution; 所述神经网络具有五层网络结构，包括一个输入层、三个隐藏层、一个输出层，其中输入层神经元数目为光谱通道数，输出层神经元数目为所需分类的气体种类数I，贝叶斯优化器基于初始值优化神经网络的超参数，优化后超参数如下：第一个隐藏层大小为71、第二个隐藏层大小为164；正则化强度为7.99425566291672e-06；激活函数为tanh，损失函数为交叉熵损失函数；The neural network has a five-layer network structure, including an input layer, three hidden layers, and an output layer, wherein the number of neurons in the input layer is the number of spectral channels, the number of neurons in the output layer is the number of gas types I required to be classified, and the Bayesian optimizer optimizes the hyperparameters of the neural network based on the initial values. The optimized hyperparameters are as follows: the size of the first hidden layer is 71, the size of the second hidden layer is 164; the regularization strength is 7.99425566291672e-06; the activation function is tanh, and the loss function is the cross entropy loss function; 步骤5，通过气体定量单元，将步骤4得到的气体分布的像素坐标区域和种类信息进行浓度方程定量解算，并基于气体浓度对气体进行伪彩色增强，输出气体浓度的空间分布，具体包括以下过程：Step 5, through the gas quantitative unit, the pixel coordinate area and type information of the gas distribution obtained in step 4 are quantitatively solved by the concentration equation, and the gas is pseudo-colored based on the gas concentration to output the spatial distribution of the gas concentration, which specifically includes the following processes: 步骤5.1，根据步骤4输出的气体分布的像素坐标区域提取对应各通道区域点的DN值；Step 5.1, extracting the DN value corresponding to each channel area point according to the pixel coordinate area of the gas distribution output in step 4; 步骤5.2，将气体像素的DN值与周围背景像素的DN值做差得到△DN(λ)，最少选择三个窄带光谱波长通道La、Lb、Lc，分别得到如下的气体浓度方程，通过求解以下三个方程解算反演气体浓度：Step 5.2, subtract the DN value of the gas pixel from the DN value of the surrounding background pixels to obtain △DN(λ), select at least three narrow-band spectral wavelength channels La , Lb , Lc , and obtain the following gas concentration equations respectively. The inverted gas concentration is calculated by solving the following three equations: ， , 其中，e为自然常数，λa、λb、λc分别为窄带光谱波长通道La、Lb、Lc的波长，τs(λa)、τs(λb)、τs(λc)分别为窄带光谱波长通道La、Lb、Lc的波段总响应率，τ2(λa)、τ2(λb)、τ2(λc)分别为波长λa、λb、λc的大气透过率，可以通过查询Modtran辐射传输模型得到，K(λa)、K(λb)、K(λc)为分别窄带光谱波长通道La、Lb、Lc的待测气体吸收率，可以通过查询Hitran数据库得到，ε(λa)、ε(λb)、ε(λc)分别为窄带光谱波长通道La、Lb、Lc的背景发射率，α(λa)、α(λb)、α(λc)分别通过波长λa、λb、λc普朗克公式计算得到，β(λa)、β(λb)、β(λc)分别通过波长λa、λb、λc普朗克公式计算得到，未知量为气体的柱密度c·l、背景温度Tbg、气体温度Tgas，将具体数值带入后可以通过求解的三元超越方程组，解出气体的柱密度、背景温度和气体温度；Wherein, e is a natural constant, λ a , λ b , λ c are the wavelengths of the narrow-band spectral wavelength channels La , L b , L c respectively, τ s (λ a ), τ s (λ b ), τ s (λ c ) are the band total response rates of the narrow-band spectral wavelength channels La , L b , L c respectively, τ 2 (λ a ), τ 2 (λ b ), τ 2 (λ c ) are the atmospheric transmittances of the wavelengths λ a , λ b , λ c respectively, which can be obtained by querying the Modtran radiation transfer model, K(λ a ), K(λ b ), K(λ c ) are the gas absorptivity of the narrow-band spectral wavelength channels La , L b , L c respectively, which can be obtained by querying the Hitran database, ε(λ a ), ε(λ b ), ε(λ c ) are the background emissivities of the narrow-band spectral wavelength channels La , L b , L c respectively, α(λ a ), α(λ b ), α(λ c ) are the atmospheric transmittances of the wavelengths λ a , λ b , λ c are calculated by Planck's formula, β(λ a ), β(λ b ), β(λ c ) are calculated by Planck's formula for wavelengths λ a , λ b , λ c respectively, the unknown quantities are the gas column density c·l, background temperature T bg , and gas temperature T gas . After substituting the specific values into the gas column density, background temperature, and gas temperature, the three-variable transcendental equation group can be solved; 步骤5.3，根据气体的浓度，对宽带全通波段子图像的气体区域进行伪彩色增强处理，输出气体浓度的空间分布。Step 5.3, according to the gas concentration, the gas area of the broadband all-pass band sub-image is subjected to pseudo-color enhancement processing, and the spatial distribution of the gas concentration is output. 2.根据权利要求1所述的一种多光谱通道气体分类和定量检测方法，其特征在于，所述步骤1具体包括以下过程：2. A multi-spectral channel gas classification and quantitative detection method according to claim 1, characterized in that step 1 specifically includes the following process: 步骤1.1，通过孔径阵列模块将入射光束分光为N个子光束，孔径阵列模块为由N个毫米级透镜排列而成的透镜阵列；Step 1.1, splitting an incident light beam into N sub-beams by an aperture array module, where the aperture array module is a lens array composed of N millimeter-level lenses; 步骤1.2，保留N个子光束中的一束，剩下的N-1个子光束通过滤光片阵列模块进行子光束滤波，得到N-1个不同窄带光谱波段的子光束和一个宽带全通波段子光束，所述滤光片阵列模块为由N-1个中心波长不同的窄带带通滤光片排列而成的滤光片阵列。Step 1.2, retaining one of the N sub-beams, and performing sub-beam filtering on the remaining N-1 sub-beams through a filter array module to obtain N-1 sub-beams with different narrowband spectral bands and one broadband all-pass band sub-beam, wherein the filter array module is a filter array composed of N-1 narrowband bandpass filters with different central wavelengths. 3.根据权利要求1所述的一种多光谱通道气体分类和定量检测方法，其特征在于，所述步骤3具体包括以下过程：3. A multi-spectral channel gas classification and quantitative detection method according to claim 1, characterized in that step 3 specifically includes the following process: 步骤3.1，将多波段通道同时成像的数字图像信号切割为N张子图像，其中包含N-1个不同窄带光谱波段子图像和一个宽带全通波段子图像；Step 3.1, cutting the digital image signal of the multi-band channel simultaneous imaging into N sub-images, including N-1 different narrow-band spectral band sub-images and one broadband all-pass band sub-image; 步骤3.2，使用ORB算法将切割后的图像配准；Step 3.2, use the ORB algorithm to register the cut images; 步骤3.3，将配准后的多通道图像排列成为具有N个通道的光谱图像立方体，输出。Step 3.3, arrange the registered multi-channel images into a spectral image cube with N channels and output it. 4.一种多光谱通道气体分类和定量检测系统，其特征在于，用于实现如权利要求1-3中任一所述的多光谱通道气体分类和定量检测方法，包括：4. A multi-spectral channel gas classification and quantitative detection system, characterized in that it is used to implement the multi-spectral channel gas classification and quantitative detection method as described in any one of claims 1 to 3, comprising: 多光谱分光模块，所述多光谱分光模块用于将入射光束细分为多个波段的子光束；A multi-spectral spectrometer module, which is used to subdivide an incident light beam into sub-beams of multiple wavelength bands; 红外探测器，所述红外探测器用于将多光谱分光模块得到的子光束转化为肉眼可见的数字图像信号；An infrared detector, which is used to convert the sub-beams obtained by the multi-spectral spectrometer into digital image signals visible to the naked eye; 气体分类定量模块，所述气体分类定量模块包括图像畸变矫正和配准单元、气体分类单元和气体定量单元，所述图像畸变矫正和配准单元用于将红外探测器输出的数字图像信号进行畸变矫正并与不同通道的图像进行图像配准对齐，输出多通道的光谱图像立方体；所述气体分类单元用于将图像畸变矫正和配准单元输出的配准后的光谱图像立方体进行像素级的气体的分类检测，得到气体分布的像素坐标区域和种类信息；所述气体定量单元用于将气体分类单元的输出得到的气体分布的像素坐标区域和种类信息进行浓度方程定量解算，并基于气体浓度对气体进行伪彩色增强，输出气体浓度的空间分布。A gas classification and quantification module, the gas classification and quantification module includes an image distortion correction and registration unit, a gas classification unit and a gas quantification unit, the image distortion correction and registration unit is used to perform distortion correction on the digital image signal output by the infrared detector and perform image registration and alignment with images of different channels, and output a multi-channel spectral image cube; the gas classification unit is used to perform pixel-level gas classification detection on the registered spectral image cube output by the image distortion correction and registration unit, and obtain the pixel coordinate area and type information of the gas distribution; the gas quantification unit is used to perform quantitative concentration equation solution on the pixel coordinate area and type information of the gas distribution obtained from the output of the gas classification unit, and perform pseudo-color enhancement on the gas based on the gas concentration, and output the spatial distribution of the gas concentration. 5.根据权利要求4所述的一种多光谱通道气体分类和定量检测系统，其特征在于，所述多光谱分光模块包含如下子模块：5. A multi-spectral channel gas classification and quantitative detection system according to claim 4, characterized in that the multi-spectral spectrometry module comprises the following sub-modules: 孔径阵列模块，所述孔径阵列模块用于将红外入射光束分为N个子光束，所述孔径阵列模块为由N个毫米级透镜排列而成的透镜阵列；An aperture array module, the aperture array module is used to divide the incident infrared light beam into N sub-beams, and the aperture array module is a lens array composed of N millimeter-level lenses; 窄带带通滤光片阵列模块，所述窄带带通滤光片阵列模块用于将N-1个子光束滤波，得到N-1个不同窄带光谱波段的子光束，所述滤光片阵列模块为由N-1个中心波长不同的窄带带通滤光片排列而成的滤光片阵列。A narrowband bandpass filter array module, which is used to filter N-1 sub-beams to obtain N-1 sub-beams with different narrowband spectral bands. The filter array module is a filter array composed of N-1 narrowband bandpass filters with different central wavelengths.
CN202411281550.5A 2024-09-13 2024-09-13 Multispectral channel gas classification and quantitative detection method and system thereof Active CN118817635B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202411281550.5A CN118817635B (en) 2024-09-13 2024-09-13 Multispectral channel gas classification and quantitative detection method and system thereof

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202411281550.5A CN118817635B (en) 2024-09-13 2024-09-13 Multispectral channel gas classification and quantitative detection method and system thereof

Publications (2)

Publication Number Publication Date
CN118817635A CN118817635A (en) 2024-10-22
CN118817635B true CN118817635B (en) 2025-01-07

Family

ID=93073325

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202411281550.5A Active CN118817635B (en) 2024-09-13 2024-09-13 Multispectral channel gas classification and quantitative detection method and system thereof

Country Status (1)

Country Link
CN (1) CN118817635B (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111272687A (en) * 2020-03-27 2020-06-12 东北大学 A real-time detection device for dangerous gas based on infrared multispectral imaging
CN113260835A (en) * 2018-12-14 2021-08-13 光谱Md公司 System and method for high precision multi-aperture spectral imaging

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111413290B (en) * 2020-03-25 2021-02-05 北京理工大学 Passive infrared filtering imaging sparse spectrum gas identification device and method
CN113848537B (en) * 2020-06-25 2025-07-29 深圳奥锐达科技有限公司 Chromatic dispersion spectrum photosensitive assembly, receiving end and laser radar system
CN113554578B (en) * 2021-07-23 2024-05-31 奥比中光科技集团股份有限公司 Method, device, terminal and storage medium for determining spectral image
CN113777056A (en) * 2021-08-13 2021-12-10 昆山格博尔智能科技有限责任公司 Ink color matching detection method based on multispectral CMOS image chip

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113260835A (en) * 2018-12-14 2021-08-13 光谱Md公司 System and method for high precision multi-aperture spectral imaging
CN111272687A (en) * 2020-03-27 2020-06-12 东北大学 A real-time detection device for dangerous gas based on infrared multispectral imaging

Also Published As

Publication number Publication date
CN118817635A (en) 2024-10-22

Similar Documents

Publication Publication Date Title
Paulus et al. Technical workflows for hyperspectral plant image assessment and processing on the greenhouse and laboratory scale
Kumar et al. Deep remote sensing methods for methane detection in overhead hyperspectral imagery
CN104697943B (en) A kind of lossless detection method of rape water stress
CN105203465B (en) A kind of ultraphotic spectrum infrared-imaging gas monitoring device and its monitoring method
CN111398176B (en) Water body water color abnormity remote sensing identification method and device based on pixel scale characteristics
CN113109317A (en) Raman spectrum quantitative analysis method and system based on background subtraction extraction peak area
CN118858223B (en) A water quality detection method, device and system for marine water quality background fluctuations
CN115561193A (en) Fourier infrared spectrometer data processing and analyzing system
CN110108658B (en) A method and system for identifying polluted gas infrared spectrum
CN115731191A (en) Narrow-band spectral imaging method based on neural network
CN111413290B (en) Passive infrared filtering imaging sparse spectrum gas identification device and method
JP2015180895A (en) Method for designing discrimination filter, discrimination method, discrimination filter set, discrimination device, and program
CN119715423A (en) A method and system for predicting organic acid content in traditional Chinese medicine based on hyperspectral imaging
CN118817635B (en) Multispectral channel gas classification and quantitative detection method and system thereof
CN118351431A (en) A rapid detection method for forestry seedling nutrient information based on specific band hyperspectral imaging equipment
CN116046698A (en) Potato detection method, device, equipment, storage medium and program product
CN113566971B (en) Multispectral high-temperature transient measurement system based on neural network
CN111222543A (en) Substance identification method and apparatus, and computer-readable storage medium
CN113945531A (en) A dual-channel imaging gas quantitative detection method
CN117871450B (en) Cross interference processing method in non-dispersive infrared gas detection
WO2014201117A1 (en) Fluorescence detection device, system, and process
CN119935935A (en) Multispectral infrared imaging gas detection method based on microlens and filter array
CN115270061B (en) A method for inverting gas pollutant concentration under complex background based on infrared hyperspectral
CN116412911A (en) Broadband Spectral Coding Method and Spectral System Based on Triangular Matrix Properties
CN108593104B (en) A small hand-held spectral detection system with high signal-to-noise ratio

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant