CN102331402B - Chemical gas detection and identification method by hyperspectral imaging - Google Patents
Chemical gas detection and identification method by hyperspectral imaging Download PDFInfo
- Publication number
- CN102331402B CN102331402B CN 201110146379 CN201110146379A CN102331402B CN 102331402 B CN102331402 B CN 102331402B CN 201110146379 CN201110146379 CN 201110146379 CN 201110146379 A CN201110146379 A CN 201110146379A CN 102331402 B CN102331402 B CN 102331402B
- Authority
- CN
- China
- Prior art keywords
- chemical gas
- pixel
- execution
- gas
- chemical
- 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
Links
Images
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a chemical gas detection and identification method by hyperspectral imaging and relates to the chemical gas detection and identification method. By the chemical gas detection and identification method, the technical problems that on-site sampling is required and the detection speed is low in the conventional chemical gas detection and identification method based on a chemical method can be solved. The chemical gas detection and identification method comprises the following steps of: shooting a spectral image of a detected place when any chemical gas does not exist in the detected place so as to acquire a pixel vector of a background; acquiring a spectral vector of a detected chemical gas from a chemical gas spectrum library; shooting another spectral image of the detected place and filtering the second spectral image by using a filter; sequencing output values of the filter from large to small; setting a threshold of the chemical gas; inputting the output values of the filter and the threshold of the chemical gas into a judger; recording the output values, greater than the threshold, of the filter so as to acquire a chemical gas-containing pixel set; and judging pixels containing the chemical gas by using a minimum markov distance method so as to identify the type of the gas. The chemical gas detection and identification method can be used in the fields of detection for pollution gas emission of chemical factories and environment monitoring.
Description
Technical field
The invention belongs to chemical gas and detect recognition methods.
Background technology
High light spectrum image-forming is the process of obtaining the image information of high spectral resolution at the spectral band of electromagnetic spectrum.High spectrum image has comprised spatial information and the spectral information in the observation scene simultaneously, has the characteristic of " collection of illustrative plates unification ".Each pixel in the image is a high dimension vector, and every one dimension of vector is illustrated in a spectral band.Because high spectrum image has advantages of that spectral resolution is high, contains much information, can greatly improve the ability of in the target detection process, target being carried out qualitative and quantitative analysis, aspect civilian and military field many, embodied huge using value.The chemical gas incident of leakage happens occasionally all over the world, existing based on the method for chemical method to the detection and Identification of chemical gas, all to arrive leak scene sampling, staff's health is damaged, and detection speed is slow, is unfavorable for the timely processing to accident.
Summary of the invention
The present invention will solve existingly need to arrive on-site sampling and the slow technical matters of detection speed based on chemical method to the method for the detection and Identification of chemical gas, and provides a kind of high light spectrum image-forming chemical gas to detect recognition methods.
A kind of high light spectrum image-forming chemical gas of the present invention detects recognition methods to carry out according to the following steps:
Step 1: when being detected the area without chemical gas, adopting wavelength with the infrared pick-up device is that at least three different-wavebands in 8 μ m~13 μ m are taken the images with K pixel, each wave band is taken an image, is obtained K pixel vectors x of background by the pixel of the correspondence of these images
k, k=1 wherein, 2,3 ..., K; K is the number of pixels of entire image;
Step 2: use
Calculate the pixel vectors x of background
kThe maximum likelihood estimated value of average
Use again
Calculate the pixel vectors x of background
kThe maximum likelihood estimated matrix of variance
Wherein
Be pixel vectors x
kTransposed vector;
Step 3: the target chemical gas for N kind needs detect according to the data in the chemical gas library of spectra, obtains chemical gas spectrum vector s in the gaseous spectrum storehouse
i, i=1 wherein, 2,3 ..., N;
Step 4: in detected area, adopting wavelength with the infrared pick-up device is that at least three different-wavebands in 8 μ m~13 μ m are taken the images with K pixel, each wave band is taken an image, is obtained K the pixel vectors z in detected area by the pixel of the correspondence of these images
j, j=1 wherein, 2,3 ..., K;
Step 5: shilling i=1;
Step 6: utilize
K the pixel vectors z that step 4 is obtained
jThe output valve y of calculating filter
I, 1, y
I, 2, y
I, 3... y
I, j-1, y
I, j, y
I, j+1..., y
I, kAnd arranged sequentially by from big to small, the maximum likelihood estimated value E[Y of the average of [α K] individual output valve before calculating again] and the maximum likelihood estimated value Var[Y of variance], simultaneously [α K] individual output valve is designated as u
iWherein
Be matrix
Inverse matrix; α=5%~10%;
Step 7: utilize
With
Calculate parameter σ and ξ; Be used further to detect the thresholding of i kind chemical gas
P wherein
FABe false-alarm probability, P
FA=10
-3~10
-5
Step 8: the output valve y of each wave filter that step 6 is calculated
I, 1, y
I, 2, y
I, 3... y
I, j-1, y
I, j, y
I, j+1..., y
I, kThe thresholding of the i kind chemical gas that calculates with step 7
Together input decision device and compare, judge whether the output valve greater than the wave filter of thresholding, if not, then make i add 1, execution in step six; If so, execution in step nine;
Step 9: record is greater than pixel vectors z corresponding to the output valve of thresholding
iAnd corresponding i kind chemical gas;
Step 10: judge then whether i equals N, if not, make i add 1, execution in step six; If so, execution in step 11 then;
Step 11: the pixel vectors z that the output valve greater than thresholding that records in the step 8 is corresponding
cExpression, c=1,2 ..., C; The corresponding spectrum vector s of chemical gas in the gaseous spectrum storehouse
dExpression, d=1 wherein, 2,3 ..., D; Make c=1, d=1; Wherein C represents the maximal value greater than the number of pixel vectors corresponding to the output valve of thresholding, and D represents the maximal value of corresponding chemical gas kind;
Step 12: utilize
Calculate horse formula distance, delta
C, d,
Wherein,
Be the maximum likelihood estimated value of the average of background,
The maximum likelihood estimated matrix of the variance of background
Inverse matrix,
The transposition of the spectrum vector of d kind chemical gas in the gaseous spectrum storehouse;
Step 13: judge whether d is 1, if make d add 1, execution in step 12; If not, execution in step 14 then;
Step 14: judge Δ
C, d≤ Δ
C, d-1If,, record d kind chemical gas; If not, execution in step 15;
Step 15: judge whether d equals D, if not, then make d add 1, return step 12; If, execution in step 16;
Step 10 six: c gas that monitors threat is judged into the chemical gas of last record and made c add 1;
Step 10 seven: judge whether c equals C, if not, make c add 1, execution in step 12; If then execution in step 18;
Step 10 eight: calculate the corresponding volume coordinate of each pixel that detects chemical gas, finish the detection identification of chemical gas.
Because atmosphere is relatively transparent in LONG WAVE INFRARED zone (8-13 μ m), utilize the high spectrum image of this wave band to carry out remote detection and Identification to the chemical gas of leakage or discharge in violation of regulations.Compare with the gas detection based on chemical method, utilize the high light spectrum image-forming technology that gas is detected and have more real-time and more secure to personal safety.In long wave infrared region, the imaging model of chemical gas can be constructed as x=as+v, and wherein x is a pixel vectors in the high spectrum image, and s is the spectrum vector of chemical gas, and v is the background vector, and a is the coefficient relevant with thickness with the concentration of this gas.Therefore, the impact that just must remove background clutter in order to detect the pixel that contains chemical gas designs a matched filter and evades this interference, and whether the output of wave filter can cooperate decision threshold to detect chemical gas to exist.But this need to make up a rational statistical model so that the output of wave filter can meet, theory according to Manolakis, the high spectrum image that the background clutter that covers without chemical gas becomes in long wave infrared region, through behind the matched filter, its output valve meets broad sense Pareto and distributes, and obtains a CFAR decision threshold by the mode of setting false-alarm probability.The detection of chemical gas is by designing a matched filter and associated decision threshold is finished.Mahalanobis distance utilizes the similarity that the statistical property of data can two samples of fine measurement, and therefore on the problem of chemical gas identification, the pixel that the chemical gas that detects in the image is existed is judged to the with it chemical gas of mahalanobis distance minimum.Find P the minimum value in the mahalanobis distance, corresponding gaseous species namely is the chemical gas that this location of pixels is revealed or discharged.The design concept schematic diagram of the inventive method integrates above various piece as shown in Figure 1, is exactly the system of automatic chemical gas detection and Identification.
High light spectrum image-forming chemical gas of the present invention detects the at a distance area of monitoring chemical gas discharging or leakage of recognition methods, has avoided arriving on-site sampling, and has worked by computer operation, and detection speed is fast.
The present invention can be used for detection and the environmental monitoring field to the discharging of chemical plant dusty gas.
Description of drawings
The design concept schematic diagram of Fig. 1 the inventive method; Fig. 2 is the process flow diagram that high light spectrum image-forming chemical gas of the present invention detects recognition methods.
Embodiment
Embodiment one: a kind of high light spectrum image-forming chemical gas of present embodiment detects recognition methods to carry out according to the following steps:
Step 1: when being detected the area without chemical gas, adopting wavelength with the infrared pick-up device is that at least three different-wavebands in 8 μ m~13 μ m are taken the images with K pixel, each wave band is taken an image, is obtained K pixel vectors x of background by the pixel of the correspondence of these images
k, k=1 wherein, 2,3 ..., K; K is the number of pixels of entire image;
Step 2: use
Calculate the pixel vectors x of background
kThe maximum likelihood estimated value of average
Use again
Calculate the pixel vectors x of background
kThe maximum likelihood estimated matrix of variance
Wherein
Be pixel vectors x
kTransposed vector;
Step 3: the target chemical gas for N kind needs detect according to the data in the chemical gas library of spectra, obtains chemical gas spectrum vector s in the gaseous spectrum storehouse
i, i=1 wherein, 2,3 ..., N;
Step 4: in detected area, adopting wavelength with the infrared pick-up device is that at least three different-wavebands in 8 μ m~13 μ m are taken the images with K pixel, each wave band is taken an image, is obtained K the pixel vectors z in detected area by the pixel of the correspondence of these images
j, j=1 wherein, 2,3 ..., K;
Step 5: shilling i=1;
Step 6: utilize
K the pixel vectors z that step 4 is obtained
jThe output valve y of calculating filter
I, 1, y
I, 2, y
I, 3... y
I, j-1, y
I, j, y
I, j+1..., y
I, kAnd arranged sequentially by from big to small, the maximum likelihood estimated value E[Y of the average of [α K] individual output valve before calculating again] and the maximum likelihood estimated value Var[Y of variance], simultaneously [α K] individual output valve is designated as u
iWherein
Be matrix
Inverse matrix; α=5%~10%;
Step 7: utilize
With
Calculate parameter σ and ξ; Be used further to detect the thresholding of i kind chemical gas
P wherein
FABe false-alarm probability, P
FA=10
-3~10
-5
Step 8: the output valve y of each wave filter that step 6 is calculated
I, 1, y
I, 2, y
I, 3... y
I, j-1, y
I, j, y
I, j+1..., y
I, kThe thresholding of the i kind chemical gas that calculates with step 7
Together input decision device and compare, judge whether the output valve greater than the wave filter of thresholding, if not, then make i add 1, execution in step six; If so, execution in step nine;
Step 9: record is greater than pixel vectors z corresponding to the output valve of thresholding
iAnd corresponding i kind chemical gas;
Step 10: judge then whether i equals N, if not, make i add 1, execution in step six; If so, execution in step 11 then;
Step 11: the pixel vectors z that the output valve greater than thresholding that records in the step 8 is corresponding
cExpression, c=1,2 ..., C; The corresponding spectrum vector s of chemical gas in the gaseous spectrum storehouse
dExpression, d=1 wherein, 2,3 ..., D; Make c=1, d=1; Wherein C represents the maximal value greater than the number of pixel vectors corresponding to the output valve of thresholding, and D represents the maximal value of corresponding chemical gas kind;
Step 12: calculate
Calculate horse formula distance, delta
C, d, wherein,
Be the maximum likelihood estimated value of the average of background,
The maximum likelihood estimated matrix of the variance of background
Inverse matrix,
The transposition of the spectrum vector of d kind chemical gas in the gaseous spectrum storehouse;
Step 13: judge whether d is 1, if make d add 1, execution in step 12; If not, execution in step 14 then;
Step 14: judge Δ
C, d≤ Δ
C, d-1If,, record d gas; If not, execution in step 15;
Step 15: judge whether d equals D, if not, then make d add 1, return step 12; If, execution in step 16;
Step 10 six: c gas that monitors threat is judged into the chemical gas of last record and made c add 1;
Step 10 seven: judge whether c equals C, if not, make c add 1, execution in step 12; If then execution in step 18;
Step 10 eight: calculate the corresponding volume coordinate of each pixel that detects chemical gas, finish the detection identification of chemical gas.
[α K] expression in the step 5 of present embodiment rounds calculating with α K.
In the present embodiment, because atmosphere is relatively transparent in LONG WAVE INFRARED zone (8-13 μ m), utilize the high spectrum image of this wave band to carry out remote detection and Identification to the chemical gas of leakage or discharge in violation of regulations.Compare with the gas detection based on chemical method, utilize the high light spectrum image-forming technology that gas is detected and have more real-time and more secure to personal safety.In long wave infrared region, the imaging model of chemical gas can be constructed as x=as+v, and wherein x is a pixel vectors in the high spectrum image, and s is the spectrum vector of chemical gas, and v is the background vector, and a is the coefficient relevant with thickness with the concentration of this gas.Therefore, the impact that just must remove background clutter in order to detect the pixel that contains chemical gas designs a matched filter and evades this interference, and whether the output of wave filter can cooperate decision threshold to detect chemical gas to exist.But this need to make up a rational statistical model so that the output of wave filter can meet, theory according to Manolakis, the high spectrum image that the background clutter that covers without chemical gas becomes in long wave infrared region, through behind the matched filter, its output valve meets broad sense Pareto and distributes, and obtains a CFAR decision threshold by the mode of setting false-alarm probability.The detection of chemical gas is by designing a matched filter and associated decision threshold is finished.Mahalanobis distance utilizes the similarity that the statistical property of data can two samples of fine measurement, and therefore on the problem of chemical gas identification, the pixel that the chemical gas that detects in the image is existed is judged to the with it chemical gas of mahalanobis distance minimum.Find P the minimum value in the mahalanobis distance, corresponding gaseous species namely is the chemical gas that this location of pixels is revealed or discharged.Present embodiment integrates above various piece, is exactly the system of automatic chemical gas detection and Identification.The at a distance monitoring chemical gas discharging of present embodiment or the area of leakage all have very high practicality at aspects such as the detection of chemical plant dusty gas discharging and environmental monitorings.
Embodiment two: what present embodiment and embodiment one were different is: calculating each method that detects the corresponding volume coordinate of pixel of chemical gas in the step 10 eight is: according to spatial resolution and the camera site of high spectrum sensor, calculate the corresponding volume coordinate of each pixel that detects chemical gas:
(u
0, v
0) be the volume coordinate of a known pixels on the image, take it as reference point, (u, v) be volume coordinate corresponding to pixel, (e, f) is the transverse and longitudinal number of pixels coordinate that the pixel that detects chemical gas deviates from benchmark pixel, and θ is the spatial resolution of sensor.Other is identical with embodiment one.
Embodiment three: the method for a kind of high light spectrum image-forming chemical gas detection and Identification of present embodiment is carried out according to the following steps:
Step 1: when being detected the area without chemical gas, adopting wavelength with the infrared pick-up device is that 10 different-wavebands in 8 μ m~13 μ m are taken the images with 10000 pixels, each wave band is taken an image, obtain altogether 10 images, obtained 10000 pixel vectors x of background by the pixel of the correspondence of these 10 images
k, k=1 wherein, 2,3 ..., 10000;
Step 2: use
Calculate the maximum likelihood estimated value of the average of background
Use again
Calculate the maximum likelihood estimated matrix of the variance of background
Wherein
Be pixel vectors x
kTransposed vector;
Step 3: the target chemical gas for 10 kinds of needs detect according to the data in the chemical gas library of spectra, obtains i kind chemical gas spectrum vector s in the gaseous spectrum storehouse
i, i=1 wherein, 2,3 ..., 10;
Step 4: in detected area, adopting wavelength with the infrared pick-up device is the image that the 10 different-wavebands shooting in 8 μ m~13 μ m has 10000 pixels, each wave band is taken an image, obtain altogether 10 images, obtained 10000 pixel vectors z in detected area by the pixel of the correspondence of these 10 images
j, j=1 wherein, 2,3 ..., 10000;
Step 5: shilling i=1;
Step 6: utilize
10000 pixel vectors z that step 4 is obtained
jThe output valve y of calculating filter
I, 1, y
I, 2, y
I, 3... y
I, j-1, y
I, j, y
I, j+1..., y
I, 10000And arranged sequentially by from big to small, calculate again the maximum likelihood estimated value E[Y of the average of front 500 output valves] and the maximum likelihood estimated value Var[Y of variance], simultaneously the 500th output valve is designated as u
iWherein
Be matrix
Inverse matrix;
Step 7: utilize
With
Calculate parameter σ and ξ; Be used further to detect the thresholding of i kind chemical gas
P wherein
FABe false-alarm probability, P
FA=10
-4
Step 8: the output valve y of each wave filter that step 6 is calculated
I, 1, y
I, 2, y
I, 3... y
I, j-1, y
I, j, y
I, j+1..., y
I, kThe thresholding of the i kind chemical gas that calculates with step 7
Together input decision device and compare, judge whether the output valve greater than the wave filter of thresholding, if not, then make i add 1, execution in step six; If so, execution in step nine;
Step 9: record is greater than pixel vectors z corresponding to the output valve of thresholding
iAnd corresponding i kind chemical gas;
Step 10: judge then whether i equals N, if not, make i add 1, execution in step six; If so, execution in step 11 then;
Step 11: the pixel vectors z that the output valve greater than thresholding that records in the step 8 is corresponding
cExpression, c=1,2 ..., C; The corresponding spectrum vector s of chemical gas in the gaseous spectrum storehouse
dExpression, d=1 wherein, 2,3 ..., D; Make c=1, d=1; Wherein C represents the maximal value greater than the number of pixel vectors corresponding to the output valve of thresholding, and D represents the maximal value of corresponding chemical gas kind;
Step 12: calculate
Calculate horse formula distance, delta
C, d, wherein,
Be the maximum likelihood estimated value of the average of background,
The maximum likelihood estimated matrix of the variance of background
Inverse matrix,
The transposition of the spectrum vector of d kind chemical gas in the gaseous spectrum storehouse;
Step 13: judge whether d is 1, if make d add 1, execution in step 12; If not, execution in step 14 then;
Step 14: judge Δ
C, d≤ Δ
C, d-1If,, record d gas; If not, execution in step 15;
Step 15: judge whether d equals D, if not, then make d add 1, return step 12; If, execution in step 16;
Step 10 six: c gas that monitors threat is judged into the chemical gas of last record and made c add 1;
Step 10 seven: judge whether c equals C, if not, make c add 1, execution in step 12; If then execution in step 18;
Step 10 eight: according to spatial resolution and the camera site of high spectrum sensor, calculate the corresponding volume coordinate of each pixel that detects chemical gas:
(u
0, v
0) be the volume coordinate of a known pixels on the image, take it as reference point, (u, v) be volume coordinate corresponding to pixel, (e, f) be the transverse and longitudinal number of pixels coordinate that the pixel that detects chemical gas deviates from benchmark pixel, θ is the spatial resolution of sensor, finishes the detection identification of chemical gas.
Wavelength in the present embodiment in step 1 and the step 4 is that 10 different-wavebands in 8 μ m~13 μ m are got like this:
8.0μm~8.25μm | 8.26μm~8.50μm | 8.51μm~8.75μm | 8.51μm~8.75μm | 8.76μm~9.00μm |
9.05μm~9.25μm | 9.26μm~9.50μm | 9.51μm~9.75μm | 9.76μm~10.00μm | 10.01μm~10.25μm |
The dimension of pixel vectors in the high spectrum image, identical with the number of the wave band of getting, the pixel vectors x of background in the step 1 of present embodiment
kDimension be 10 the dimension, the pixel vectors z in the step 4
jDimension also be 10 the dimension.
In the present embodiment, because atmosphere is relatively transparent in LONG WAVE INFRARED zone (8-13 μ m), utilize the high spectrum image of this wave band to carry out remote detection and Identification to the chemical gas of leakage or discharge in violation of regulations.Compare with the gas detection based on chemical method, utilize the high light spectrum image-forming technology that gas is detected and have more real-time and more secure to personal safety.In long wave infrared region, the imaging model of chemical gas can be constructed as x=as+v, and wherein x is a pixel vectors in the high spectrum image, and s is the spectrum vector of chemical gas, and v is the background vector, and a is the coefficient relevant with thickness with the concentration of this gas.Therefore, the impact that just must remove background clutter in order to detect the pixel that contains chemical gas designs a matched filter and evades this interference, and whether the output of wave filter can cooperate decision threshold to detect chemical gas to exist.But this need to make up a rational statistical model so that the output of wave filter can meet, theory according to Manolakis, the high spectrum image that the background clutter that covers without chemical gas becomes in long wave infrared region, through behind the matched filter, its output valve meets broad sense Pareto and distributes, and obtains a CFAR decision threshold by the mode of setting false-alarm probability.The detection of chemical gas is by designing a matched filter and associated decision threshold is finished.Mahalanobis distance utilizes the similarity that the statistical property of data can two samples of fine measurement, and therefore on the problem of chemical gas identification, the pixel that the chemical gas that detects in the image is existed is judged to the with it chemical gas of mahalanobis distance minimum.Find P the minimum value in the mahalanobis distance, corresponding gaseous species namely is the chemical gas that this location of pixels is revealed or discharged.Present embodiment integrates above various piece, is exactly the system of automatic chemical gas detection and Identification.The at a distance monitoring chemical gas discharging of present embodiment or the area of leakage all have very high practicality at aspects such as the detection of chemical plant dusty gas discharging and environmental monitorings.
Claims (2)
1. a high light spectrum image-forming chemical gas detects recognition methods, it is characterized in that the high light spectrum image-forming chemical gas detects recognition methods and carries out according to the following steps:
Step 1: when being detected the area without chemical gas, adopting wavelength with the infrared pick-up device is that at least three different-wavebands in 8 μ m~13 μ m are taken the images with K pixel, each wave band is taken an image, is obtained K pixel vectors x of background by the pixel of the correspondence of these images
k, k=1 wherein, 2,3 ..., K; K represents the number of pixels of entire image;
Step 2: use
Calculate the pixel vectors x of background
kThe maximum likelihood estimated value of average
Use again
Calculate the pixel vectors x of background
kThe maximum likelihood estimated matrix of variance
Wherein
Be pixel vectors x
kTransposed vector;
Step 3: the target chemical gas for N kind needs detect according to the data in the chemical gas library of spectra, obtains chemical gas spectrum vector s in the gaseous spectrum storehouse
i, i=1 wherein, 2,3 ..., N;
Step 4: in detected area, adopting wavelength with the infrared pick-up device is that at least three different-wavebands in 8 μ m~13 μ m are taken the images with K pixel, each wave band is taken an image, is obtained K the pixel vectors z in detected area by the pixel of the correspondence of these images
j, j=1 wherein, 2,3 ..., K;
Step 5: shilling i=1;
Step 6: utilize
K the pixel vectors z that step 4 is obtained
jThe output valve y of calculating filter
I, 1, y
I, 2, y
I, 3... y
I, j-1, y
I, j, y
I, j+1..., y
I, kAnd arranged sequentially by from big to small, the maximum likelihood estimated value E[Y of the average of [α K] individual output valve before calculating again] and the maximum likelihood estimated value Var[Y of variance], simultaneously [α K] individual output valve is designated as u
iWherein
Be matrix
Inverse matrix; α=5%~10%;
Step 7: utilize
With
Calculate parameter σ and ξ; Be used further to detect the thresholding of i kind chemical gas
P wherein
FABe false-alarm probability, P
FA=10
-3~10
-5
Step 8: the output valve y of each wave filter that step 6 is calculated
I, 1, y
I, 2, y
I, 3... y
I, j-1, y
I, j, y
I, j+1..., y
I, kThe thresholding of the i kind chemical gas that calculates with step 7
Together input decision device and compare, judge whether the output valve greater than the wave filter of thresholding, if not, then make i add 1, execution in step six; If so, execution in step nine;
Step 9: record is greater than pixel vectors z corresponding to the output valve of thresholding
iAnd corresponding i kind chemical gas;
Step 10: judge then whether i equals N, if not, make i add 1, execution in step six; If so, execution in step 11 then;
Step 11: the pixel vectors z that the output valve greater than thresholding that records in the step 8 is corresponding
cExpression, c=1,2 ..., C; The corresponding spectrum vector s of chemical gas in the gaseous spectrum storehouse
dExpression, d=1 wherein, 2,3 ..., D; Make c=1, d=1; Wherein C represents the maximal value greater than the number of pixel vectors corresponding to the output valve of thresholding, and D represents the maximal value of corresponding chemical gas kind;
Step 12: utilize
Calculate horse formula distance, delta
C, d, wherein,
Be the maximum likelihood estimated value of the average of background,
The maximum likelihood estimated matrix of the variance of background
Inverse matrix,
The transposition of the spectrum vector of d kind chemical gas in the gaseous spectrum storehouse;
Step 13: judge whether d is 1, if make d add 1, execution in step 12; If not, execution in step 14 then;
Step 14: judge Δ
C, d≤ Δ
C, d-1If,, record d kind chemical gas; If not, execution in step 15;
Step 15: judge whether d equals D, if not, then make d add 1, execution in step 12; If, execution in step 16;
Step 10 six: c gas that monitors threat is judged into the chemical gas of last record and made c add 1;
Step 10 seven: judge whether c equals C, if not, make c add 1, execution in step 12; If then execution in step 18;
Step 10 eight: calculate the corresponding volume coordinate of each pixel that detects chemical gas, finish the detection identification of chemical gas.
2. a kind of high light spectrum image-forming chemical gas according to claim 1 detects recognition methods, it is characterized in that calculating in the step 10 eight each method that detects the corresponding volume coordinate of pixel of chemical gas is: according to spatial resolution and the camera site of high spectrum sensor, calculate the corresponding volume coordinate of each pixel that detects chemical gas:
(u
0, v
0) be the volume coordinate of a known pixels on the image, take it as reference point, (u, v) be volume coordinate corresponding to pixel, (e, f) is the transverse and longitudinal number of pixels coordinate that the pixel that detects chemical gas deviates from benchmark pixel, and θ is the spatial resolution of sensor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110146379 CN102331402B (en) | 2011-06-01 | 2011-06-01 | Chemical gas detection and identification method by hyperspectral imaging |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110146379 CN102331402B (en) | 2011-06-01 | 2011-06-01 | Chemical gas detection and identification method by hyperspectral imaging |
Publications (2)
Publication Number | Publication Date |
---|---|
CN102331402A CN102331402A (en) | 2012-01-25 |
CN102331402B true CN102331402B (en) | 2013-05-01 |
Family
ID=45483253
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 201110146379 Active CN102331402B (en) | 2011-06-01 | 2011-06-01 | Chemical gas detection and identification method by hyperspectral imaging |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102331402B (en) |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106018316B (en) * | 2016-05-23 | 2018-11-09 | 湖北久之洋红外系统股份有限公司 | A kind of gas detection method based on EO-1 hyperion infrared image processing |
CN110192098B (en) * | 2016-11-14 | 2022-01-04 | 奥普加尔光电工业有限公司 | System and method for quantifying gas leaks |
CN107451590B (en) * | 2017-07-19 | 2020-09-25 | 哈尔滨工程大学 | Gas detection identification and concentration representation method based on hyperspectral infrared image |
CN110673628B (en) * | 2019-09-20 | 2020-09-29 | 北京航空航天大学 | Inspection method for oil-gas pipeline of composite wing unmanned aerial vehicle |
CN111862541A (en) * | 2020-06-28 | 2020-10-30 | 安徽旭帆信息科技有限公司 | Poisonous and harmful gas monitoring system based on infrared imaging technology |
CN111965115B (en) * | 2020-07-21 | 2022-05-20 | 天津大学 | Ship tail gas measuring system and method based on hyperspectrum |
CN111965116A (en) * | 2020-07-21 | 2020-11-20 | 天津大学 | Hyperspectrum-based airport gas detection system and method |
CN115082796A (en) * | 2022-07-06 | 2022-09-20 | 中国科学院合肥物质科学研究院 | Telemetering identification method, system, equipment, device and storage medium for gas imaging |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101059425A (en) * | 2007-05-29 | 2007-10-24 | 浙江大学 | Method and device for identifying different variety green tea based on multiple spectrum image texture analysis |
CN101144861A (en) * | 2007-11-02 | 2008-03-19 | 北京航空航天大学 | High spectrum sub-pixel target detection method and device |
CN100507603C (en) * | 2007-10-16 | 2009-07-01 | 哈尔滨工业大学 | Hyperspectral image abnormal point detection method based on selective kernel principal component analysis |
WO2010056254A1 (en) * | 2008-11-17 | 2010-05-20 | Raytheon Company | Hyperspectral image dimension reduction system and method |
CN101832926A (en) * | 2010-03-19 | 2010-09-15 | 江南大学 | Method for performing apple powder materialization non-destructive inspection by using hyper-spectral image technique |
-
2011
- 2011-06-01 CN CN 201110146379 patent/CN102331402B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101059425A (en) * | 2007-05-29 | 2007-10-24 | 浙江大学 | Method and device for identifying different variety green tea based on multiple spectrum image texture analysis |
CN100507603C (en) * | 2007-10-16 | 2009-07-01 | 哈尔滨工业大学 | Hyperspectral image abnormal point detection method based on selective kernel principal component analysis |
CN101144861A (en) * | 2007-11-02 | 2008-03-19 | 北京航空航天大学 | High spectrum sub-pixel target detection method and device |
WO2010056254A1 (en) * | 2008-11-17 | 2010-05-20 | Raytheon Company | Hyperspectral image dimension reduction system and method |
CN101832926A (en) * | 2010-03-19 | 2010-09-15 | 江南大学 | Method for performing apple powder materialization non-destructive inspection by using hyper-spectral image technique |
Non-Patent Citations (2)
Title |
---|
贺霖等.高光谱图像目标检测研究进展.《电子学报》.2009,第37卷(第9期),第2016-2024页. |
高光谱图像目标检测研究进展;贺霖等;《电子学报》;20090930;第37卷(第9期);第2016-2024页 * |
Also Published As
Publication number | Publication date |
---|---|
CN102331402A (en) | 2012-01-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102331402B (en) | Chemical gas detection and identification method by hyperspectral imaging | |
CN100507603C (en) | Hyperspectral image abnormal point detection method based on selective kernel principal component analysis | |
CN111398176B (en) | Water body water color abnormity remote sensing identification method and device based on pixel scale characteristics | |
CN101807301A (en) | High spectral image target detection method based on high order statistic | |
CN104732186B (en) | Single sample face recognition method based on Local Subspace rarefaction representation | |
CN102903114A (en) | Hyperspectral remotely-sensed data dimensionality reduction method based on improved hierarchical clustering | |
CN115690564A (en) | Outdoor fire smoke image detection method based on Recursive BIFPN network | |
CN116977909B (en) | Deep learning fire intensity recognition method and system based on multi-modal data | |
CN106096505A (en) | The SAR target identification method of expression is worked in coordination with based on Analysis On Multi-scale Features | |
CN117649123B (en) | Wisdom garden management system | |
CN105760857A (en) | High spectral remote sensing image object detection method | |
CN115830516B (en) | Computer neural network image processing method for battery deflagration detection | |
Yang et al. | Research on farmland crop classification based on UAV multispectral remote sensing images | |
CN116704241A (en) | Full-channel 3D convolutional neural network hyperspectral remote sensing image classification method | |
CN106018316A (en) | Gas detection method based on hyperspectral infrared image processing | |
CN111126230A (en) | Smoke concentration quantitative evaluation method and electronic equipment applying same | |
CN114629047A (en) | Method, device and equipment for detecting slippage of damper | |
Koltunov et al. | GOES early fire detection (GOES-EFD) system prototype | |
Roy | Hybrid algorithm for hyperspectral target detection | |
Basener | An automated method for identification and ranking of hyperspectral target detections | |
Zhang et al. | Forest Fire Smoke Detection Method Based on MoAm-YOLOv4 Algorithm | |
Das et al. | Crime Pattern Detection Using Data Mining | |
CN111814787B (en) | Lock hole detection method for visible light image | |
Pollard | Development and Evaluation of Machine Learning Models for Fugitive Methane Detection and Intensity Prediction | |
CN117094477B (en) | Environment evaluation method and system in rare earth mining process |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20200327 Address after: 150001 No. 118 West straight street, Nangang District, Heilongjiang, Harbin Patentee after: Harbin University of technology high tech Development Corporation Address before: 150001 Harbin, Nangang, West District, large straight street, No. 92 Patentee before: HARBIN INSTITUTE OF TECHNOLOGY |