CN103207015A - Spectrum reconstruction method and spectrometer device - Google Patents
Spectrum reconstruction method and spectrometer device Download PDFInfo
- Publication number
- CN103207015A CN103207015A CN2013101320290A CN201310132029A CN103207015A CN 103207015 A CN103207015 A CN 103207015A CN 2013101320290 A CN2013101320290 A CN 2013101320290A CN 201310132029 A CN201310132029 A CN 201310132029A CN 103207015 A CN103207015 A CN 103207015A
- Authority
- CN
- China
- Prior art keywords
- spectrum
- matrix
- initial
- optical filter
- reconstruct
- 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.)
- Pending
Links
Images
Landscapes
- Spectrometry And Color Measurement (AREA)
Abstract
The invention discloses a spectrum reconstruction method. The method comprises the following steps of: initializing an optical filter, namely acquiring a transmission spectrum curve of each optical filter channel on the optical filter; acquiring a spectrum, namely acquiring initial spectrum information after light is transmitted into the optical filter; and performing spectrum reconstruction, namely reconstructing the initial spectrum information to acquire the reconstructed spectrum information by using the transmission spectrum curves through a non-negative matrix factorization (NMF) method. The spectrum information of probing light is acquired by reconstructing the spectrum information acquired by an image sensor. The invention also discloses a spectrometer device.
Description
Technical field
The invention belongs to field of spectral analysis technology, relate in particular to a kind of spectrum reconstructing method and spectrometer device thereof.
Background technology
The application of spectral technique has almost covered all scientific domains, comprises medicine, chemistry, geology, physics and uranology etc., and to remote universe, spectrometer is collected the material characterization information of world around for people from the bottom of ocean.Because spectrometer can obtain a large amount of information relevant with the structure of matter, be of wide application, as concentration analysis, desulfurization and sulphur recovery analysis, tails assay, flue gas analysis, pollution source analysis, colorimetric analysis and nephelometric analysis etc., so be widely used in environmental monitoring, legal medical expert's evaluation, biomedical, scientific and technological agricultural, traffic, resource exploration and scientific research field and industrial sectors such as exploitation, anti-counterfeiting detection (best brand of product, public security, customs, finance), military analysis and industrial flow monitoring.Because existing most spectral instruments all belong to large-scale precision instrument, what adopt all is light splitting modes such as grating, prism or Fourier transform, beam splitting system is bulky, anti-seismic performance is low, can only use in relatively-stationary place in laboratory and workshop etc., be difficult to satisfy the application demand of aspects such as on-the-spot and open-air detection, more can't popularization and application in daily life.
All equipment with camera function all are to obtain extraneous optical signalling by built-in CCD chip in the prior art, but the CCD chip can't directly obtain extraneous spectral information.In existing spectrum reconstruction processing method, adopt compressed sensing to solve spectrum reconstruct problem, but defective is main body optimization step wherein, namely ask the minimized speed of l1 norm (efficient) lower, and this method has been given tacit consent to the sparse base of spectrum.Adopt lucky big vast promise husband regularization to add the number that a regular terms is controlled the solution of non-well-posed problem at original objective function in addition, and simultaneously this mathematical model is become convex model, but defective is owing to having added that regular terms makes that last degree of accuracy as a result is lower, and has increased the quantity of the parameter in the algorithm.Simultaneously, determine a large amount of experience or the calculation step of the needs of value preferably of these parameters.Also there are NNLS, SVD equal matrix decomposition method in the prior art, but all need to use specific method to handle the negative element of the matrix after decomposing, and the method for simply handling the negative element can cause the removal of the useful information in the algorithm operational process and cause the degree of accuracy of net result to reduce.
Summary of the invention
The present invention has overcome in the prior art that the spectral instrument volume is excessive, spectrum reconstructing method computing complexity and the not high defective of reconstruct degree of accuracy, has proposed a kind of spectrum reconstructing method and spectrometer device thereof.
The present invention proposes a kind of spectrum reconstructing method, comprising:
The optical filter initialization step obtains the transmission spectrum curve of each optical filter passage on the optical filter;
The spectrum obtaining step obtains the initial spectrum information after light sees through described optical filter;
The spectrum reconstruction step utilizes described transmission spectrum curve with the described initial spectrum information of nonnegative matrix full rank decomposition method reconstruct, the spectral information after the acquisition reconstruct.
Wherein, described spectrum reconstruction step comprises:
Reconstruction parameter is set step, sets the largest loop number;
The matrix decomposition step generates initial matrix by described transmission spectrum curve, adopts nonnegative matrix full rank decomposition method to decompose described initial matrix according to described largest loop number, generates basis matrix and matrix of coefficients;
The matrix computations step is calculated the pseudo inverse matrix that obtains described initial matrix according to described basis matrix, matrix of coefficients and initial matrix;
The reconstruction result calculation procedure, according to described pseudo inverse matrix and initial spectrum information calculations reconstruction result, the spectral information after the acquisition reconstruct.
Wherein, in the described matrix computations step, the computing formula of pseudo inverse matrix is as described below:
inv(T)＝C
T(B
TTC
T)
-1B
T；
In the formula, C represents matrix of coefficients, and B represents basis matrix, and T represents described initial matrix.
Wherein, in the described reconstruction result calculation procedure, obtain spectral information after the reconstruct according to following formula:
X＝inv(T)Y；
In the formula, X represents the spectral information after the reconstruct, the pseudo inverse matrix of inv (T) expression initial matrix, and Y represents initial spectrum information.
Wherein, further comprise: repeat that described spectrum reconstructing method obtains a plurality of spectral informations and with respect to difference value and the error rate of real spectrum, the spectral information of the spectral information of choosing described difference value and described error rate minimum after as reconstruct.
The invention allows for a kind of spectrometer device, comprising:
The optical filter parts, its light to different wave length has different transmissivities;
Imageing sensor, it is arranged on the rear of described optical filter parts, is used for surveying the initial spectrum information of the light that sees through described optical filter parts;
The spectrum calculating unit, it receives the initial spectrum information of described imageing sensor and to reconstruct, generates the spectral information after the reconstruct.
Wherein, further comprise: the focus lamp parts, it is arranged on the place ahead of described optical filter parts, on the described imageing sensor with light focusing.
Wherein, described optical filter parts are integrated narrow-band filter.
Wherein, described imageing sensor is the CCD chip.
Wherein, further comprise display unit, it is connected with described spectrum calculating unit, to show the spectral information after the described reconstruct.
The present invention arranges one deck integrated narrow-band filter array by the surface to the CCD chip, by according to transmittances different on the integrated narrow-band filter array image that the CCD chip obtains being combined, can calculate the spectral information that obtains image.The present invention compares with spectral instrument of the prior art, and its cost of manufacture is very cheap.The present invention is particularly useful for the equipment that mobile phone etc. has CCD chip and data storage and processing capacity, can realize that spectral instrument changes on a small scale by mobile phone, promotes spectral instrument popularizing in use.
Description of drawings
What Fig. 1 showed is the process flow diagram of spectrum reconstructing method.
What Fig. 2 showed is the process flow diagram of spectrum reconstruction step.
What Fig. 3 showed is reconstruct spectrogram successively.
What Fig. 4 showed is the structural drawing of spectrometer device.
What Fig. 5 showed is the structural drawing of narrow band pass filter and CCD chip.
Embodiment
In conjunction with following specific embodiments and the drawings, the present invention is described in further detail.Implement process of the present invention, condition, experimental technique etc., except the following content of mentioning specially, be universal knowledege and the common practise of this area, the present invention is not particularly limited content.
As shown in Figure 1, spectrum reconstructing method of the present invention obtains spectral information by optical filter initialization step, spectrum obtaining step and spectrum reconstruction step.
In the step S1 optical filter initialization step, can adopt the visible light grating spectrograph in advance, incident and the exit slit of spectrometer are transferred to high resolving power, be transferred to 0.1nm in the present embodiment, then with the illumination of outgoing in the exit slit on imageing sensor, the grating of rotating shutter spectrometer makes the optical wavelength of exit slit outgoing change length by length, every one step of variation, just record next width of cloth intensity distribution on the CCD, thereby obtain the meticulous spectral value of each optical filter passage on the optical filter.In the present embodiment, integrated optical filter has 32 optical filter passages, and wherein each optical filter passage can obtain the transmission spectrum curve of self, the data point of total m accurate spectral value on each bar transmission spectrum curve by this testing process.These 32 transmission spectrum curves will become each row of initial matrix T, form the initial matrix T on 32*m rank for this reason.
In the step S2 spectrum obtaining step, light make the light of different wave length be dispersed in the zones of different in space, so integrated narrow-band filter has just replaced traditional beam splitter rotating grating or prism through after the integrated narrow-band filter.Because integrated narrow-band filter is positioned at before the image detector, so, the light that penetrates different wave length from the integrated narrow-band filter diverse location is radiated in the different photosensitive unit of light-sensitive surface of imageing sensor, the different photosensitive unit of imageing sensor provides the corresponding intensity of light of different wave length, initial spectrum information is obtained in realization, shown in Fig. 3 (b).
Innovative point of the present invention is to adopt nonnegative matrix full rank decomposition method, and (Non-negative Matrix Factorization NMF) is reconstructed to obtain spectral information after the reconstruct to the spectral information after the mating plate after filtration.In the present embodiment, the nonnegative matrix full rank is decomposed initial 32*m rank matrix T, the j of initial matrix T row are exactly the meticulous spectral value on the transmission spectrum curve of j optical filter, m is exactly a meticulous spectral value number that transmission spectrum curve is contained, in the physical sense, the value of m can be determined by the instrument parameter of the spectrometer of photometry spectrum, so the precision of m can be done very highly, for example the size of m can be 32 more than ten times, or even thousands of times up to a hundred.The nonnegative matrix full rank is decomposed initial parameter k=min (m, 32)=32.As shown in Figure 2, step S3 spectrum reconstruction step comprises: step S31 reconstruction parameter is set step, step S32 matrix decomposition step, step S33 matrix computations step and step S34 reconstruction result calculation procedure.
Wherein, step S31 reconstruction parameter is set in the step, sets the maximum algorithm period of total algorithm max_FES, sets the nonnegative matrix full rank and decomposes the fineness that MM algorithm largest loop is counted max_NMF and horizontal ordinate.The largest loop number is set in the 100-10000 underrange, and the fineness of horizontal ordinate is set in the 3-3000 times of scope of horizontal ordinate fineness in the initial spectrum information.
In the step S32 matrix decomposition step, use the nonnegative matrix full rank to decompose the MM algorithm, operation parameter k, initial matrix T and largest loop are counted max_NMF as input, export two matrixes at last, and dimension is respectively basis matrix B and the matrix of coefficients C of 32*32 and 32*m.
In the step S33 matrix computations step, according to basis matrix B and the matrix of coefficients C of above calculating, utilize the pseudo inverse matrix inv (T) of computing formula calculating initial matrix T:
inv(T)＝C
T(B
TTC
T)
-1B
T
In the formula, C represents matrix of coefficients, and B represents basis matrix, and T represents initial matrix.
In the S34 reconstruction result calculation procedure, utilize formula to calculate to generate spectral information after the reconstruct according to pseudo inverse matrix inv (T):
X＝inv(T)Y
Wherein, X represents reconstruct spectral information later, the pseudo inverse matrix of inv (T) expression initial matrix, and Y represents initial spectrum information, i.e. the measured value of the spectrum gathered of imageing sensor is 32 points at the curve of this Y.
During the spectral information of the present invention after calculating reconstruct, also calculate its difference value Error_rate and error rate Negative_rate.Error_rate refers to the F norm error between the optimum solution of spectral information after the reconstruct of the present invention and spectrum reconstructing method of the present invention.Negative_rate represents the ratio of the negative element of the spectral information after the reconstruct of the present invention, does not just have the ratio of the negative element of physical significance fully.The present invention obtains a plurality of spectral informations and difference value and error rate after repeatedly carrying out the spectrum reconstructing method.Choose minimum as Optimal error Best_error according to the size of comparing difference value and error rate, this Optimal error refers to the optimum solution of the spectral information after the reconstruct and the minimum value of the error between the actual result, namely with the immediate spectral information of real spectrum.The present invention adopts this method to compare and distinguish for the different solution that the inventive method draws, thereby reconstructs the highest spectral information of degree of accuracy.
In the present embodiment, be used for spectrum matrix such as Fig. 3 (a) of reconstruct, this spectrum matrix is unique, and the spectral information after the spectrum reconstructing method of the present invention reconstruct is a probabilistic result, namely can draw different results under the situation of identical input.Owing to be limited by the restriction of image acquisition instrument, the spectral information that collects by imageing sensor, shown in Fig. 3 (b), this initial spectrum information and have only 32 spectroscopic data points, i.e. matrix Y, its with Fig. 3 (a) between the very big difference of existence.By among the spectrum reconstructing method use initial matrix T of the present invention matrix Y being reconstructed.
(guarantee that namely Fig. 3 (a) waveform can be completely reconstructed for the integrality that guarantees matrix among Fig. 3 (a) in the spectrum restructuring procedure, the part curve of its waveform both sides can not disappear), spectrum reconstructing method of the present invention uses some precision to choose data, present embodiment is example with 401 precision, carry out universe and choose 401 data points equably in 1 to 10001 (10001 are full accuracy), just the point under 25 full accuracies is got a data point.By just can obtaining the matrix Y that precision is 401 optimum after such get, and corresponding matrix T also will evenly extract 401 row as 32 * 401 matrix T under spectrum restructing algorithm 401 precision of the present invention in 10001 row successively.
The present invention is based on the spectrum of NMF and be reconstructed, at first carry out the nonnegative matrix full rank for matrix T and decompose, matrix T is a non-singular matrix, and the order of matrix T is 32 in the present embodiment.Obtain basis matrix B and matrix of coefficients C by the decomposition of nonnegative matrix full rank, basis matrix B and matrix of coefficients C are the intermediate variables of spectrum reconstructing method of the present invention.Then, ask for the pseudo inverse matrix of matrix T by the full rank finding the inverse matrix.At last, directly obtain the matrix of consequence X of reconstruct by the pseudo inverse matrix of this matrix T.
Wherein, because the nonnegative matrix full rank is decomposed part and is needed random initializtion in the spectrum reconstructing method, basis matrix B and matrix of coefficients C are that random initializtion obtains, so the matrix X of spectrum reconstructing method of the present invention after according to the final reconstruct that obtains after the homogeneous initialization not is different.So also to filtering out the matrix X of an optimum, for other different matrix X, calculate itself and the difference value of optimum matrix X and error rate separately among the present invention, select matrix X after the spectrum reconstruct spectral information after as reconstruct with this.Spectral information after the reconstruct is shown in Fig. 3 (c).As seen from the figure, the degree of agreement of the spectrum shown in Fig. 3 (c) and spectrogram 3 (a) is obviously better than the degree of agreement of the initial spectrum information shown in the Fig. 3 (b) that directly measures and spectrogram 3 (a), although there are some shakes in mild place, Fig. 3 (c) curve both sides after handling through reconstructing method of the present invention, but two peaks are all by reconstruction render, and the degree of accuracy that the present invention has improved has improved the performance of spectrometer device among the present invention simultaneously.
As shown in Figure 4, spectrometer device of the present invention comprises optical filter parts 1, imageing sensor 2, spectrum calculating unit 3 focus lamp parts 4 and display unit 6.Focus lamp parts 4 are arranged on the place ahead of optical filter parts 1, and light passes optical filter parts 1 by focus lamp parts 4 and is mapped to then on the imageing sensor 2.Imageing sensor 2 obtains the spectral information of light, spectral information is sent in the spectrum calculating unit 3 be reconstructed, and by the spectral information after the display unit 6 demonstration reconstruct.
In the present embodiment, optical filter parts 1 are integrated narrow-band filter.The initial film architecture of integrated narrow-band filter array is (LH)
5(xL) (HL)
5, wherein lower membrane is (LH)
5, wall is xL, upper layer film is (HL)
5Upper layer film system and lower membrane system form the minute surface symmetry.H represents that thickness is λ
0/ 4 high refractive index layer, L represent that thickness is λ
0/ 4 low-index film, λ
0Be the centre wavelength of initial narrow band pass filter, 800nm.High refractive index layer and low-index film can be respectively tantalum pentoxide rete and silica coating.Unique difference between each unit of integrated narrow-band filter array is exactly the space layer difference, because the logical peak position of the band of narrow band pass filter is directly proportional with the thickness of wall, the thickness difference of wall is the logical peak position difference of band of corresponding narrow band pass filter then, realize that by changing x the variation range of x is 1＜x＜5.Interval between concrete scope and each the narrow-band-filter blade unit can depend on the circumstances, and present embodiment is got and is spaced apart 0.025, x is 1.60～2.40, and the logical peak position of the correspondence band of narrow band pass filter is 785～816nm.In the present embodiment, present embodiment is 32 kinds of equidistant narrow band pass filters, and wall is strip, and each bar is corresponding with a kind of narrow band pass filter.
In the present embodiment, imageing sensor 2 is the CCD chip.Adopt methods such as vacuum coating or reactive magnetron sputtering to be coated with lower membrane system and 2.40L wall at the CCD chip.End plated film then, the sample that has plated lower membrane system and wall is taken out, adopt ion etching process subregion conventional in the semiconductor technology to carry out etching, form 32 wall arrays that thickness does not wait.Then carry out being coated with of upper layer film system on this basis, though space layer has nothing in common with each other on the slice, thin piece that plates this moment, but rate of sedimentation is all the same everywhere during plated film, the upper layer film system of therefore plating on the zone of different interval layer thickness is all identical, and to have only space layer difference, other films be identical integrated narrow-band filter array so just formed on the CCD chip.Like this, just an integrated narrow-band filter array has been transferred on the CCD chip effectively.As shown in Figure 5, the bar shaped composed pattern is represented the integrated narrow-band filter array, and the grid composed pattern is represented the CCD chip, and each grid is represented a picture dot.After transferring to the integrated narrow-band filter array on the CCD chip as stated above, a row picture dot of CCD chip is all corresponding a kind of narrow band pass filter, this row picture dot detects the size that the signal sum is exactly this narrow band pass filter light transmission capacity for this reason.Final spectrum is made up of each row device signal size of CCD chip.Usually all greater than 256*256, will there be the picture dots in the 8 row CCD chips corresponding with it behind the narrow band pass filter according to the scale of present CCD chip.Like this, the picture dot correspondence in 32 8*256 piece CCD chips of CCD chip 32 narrow band pass filters, final initial spectrum information is to be rearranged according to the order of sequence by the signal magnitude that 32 picture dot pieces in the 256X256 scale CCD chip obtain.Spectrum calculating unit 3 is with spectrum reconstructing method reconstruct spectral information, by display unit 5 to show the spectral information after the reconstruct.
Protection content of the present invention is not limited to above embodiment.Under the spirit and scope that do not deviate from inventive concept, variation and advantage that those skilled in the art can expect all are included in the present invention, and are protection domain with the appending claims.
Claims (10)
1. a spectrum reconstructing method is characterized in that, comprising:
The optical filter initialization step obtains the transmission spectrum curve of each optical filter passage on the optical filter;
The spectrum obtaining step obtains the initial spectrum information after light sees through described optical filter;
The spectrum reconstruction step utilizes described transmission spectrum curve with the described initial spectrum information of nonnegative matrix full rank decomposition method reconstruct, the spectral information after the acquisition reconstruct.
2. spectrum reconstructing method as claimed in claim 1 is characterized in that, described spectrum reconstruction step comprises:
Reconstruction parameter is set step, sets the largest loop number;
The matrix decomposition step generates initial matrix by described transmission spectrum curve, adopts nonnegative matrix full rank decomposition method to decompose described initial matrix according to described largest loop number, generates basis matrix and matrix of coefficients;
The matrix computations step is calculated the pseudo inverse matrix that obtains described initial matrix according to described basis matrix, matrix of coefficients and initial matrix;
The reconstruction result calculation procedure, according to described pseudo inverse matrix and initial spectrum information calculations reconstruction result, the spectral information after the acquisition reconstruct.
3. spectrum reconstructing method as claimed in claim 2 is characterized in that, in the described matrix computations step, the computing formula of pseudo inverse matrix is as described below:
inv(T)＝C
T(B
TTC
T)
-1B
T；
In the formula, C represents matrix of coefficients, and B represents basis matrix, and T represents described initial matrix.
4. spectrum reconstructing method as claimed in claim 2 is characterized in that, in the described reconstruction result calculation procedure, obtains spectral information after the reconstruct according to following formula:
X＝inv(T)Y；
In the formula, X represents the spectral information after the reconstruct, the pseudo inverse matrix of inv (T) expression initial matrix, and Y represents initial spectrum information.
5. spectrum reconstructing method according to claim 1, it is characterized in that, further comprise: repeat that described spectrum reconstructing method obtains a plurality of spectral informations and with respect to difference value and the error rate of real spectrum, the spectral information of the spectral information of choosing described difference value and described error rate minimum after as reconstruct.
6. a spectrometer device is characterized in that, comprising:
Optical filter parts (1), its light to different wave length has different transmissivities;
Imageing sensor (2), it is arranged on the rear of described optical filter parts (1), is used for surveying the initial spectrum information of the light that sees through described optical filter parts (1);
Spectrum calculating unit (3), it receives the initial spectrum information of described imageing sensor (2) and to reconstruct, generates the spectral information after the reconstruct.
7. spectrometer device as claimed in claim 6 is characterized in that, further comprises: focus lamp parts (4), it is arranged on the place ahead of described optical filter parts (1), on the described imageing sensor (2) with light focusing.
8. spectrometer device as claimed in claim 6 is characterized in that, described optical filter parts are integrated narrow-band filter.
9. spectrometer device as claimed in claim 6 is characterized in that, described imageing sensor (2) is the CCD chip.
10. spectrometer device as claimed in claim 6 is characterized in that, further comprises display unit (6), and it is connected with described spectrum calculating unit (3), to show the spectral information after the described reconstruct.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013101320290A CN103207015A (en) | 2013-04-16 | 2013-04-16 | Spectrum reconstruction method and spectrometer device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013101320290A CN103207015A (en) | 2013-04-16 | 2013-04-16 | Spectrum reconstruction method and spectrometer device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103207015A true CN103207015A (en) | 2013-07-17 |
Family
ID=48754309
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2013101320290A Pending CN103207015A (en) | 2013-04-16 | 2013-04-16 | Spectrum reconstruction method and spectrometer device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103207015A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104359555A (en) * | 2014-10-16 | 2015-02-18 | 中国科学院上海技术物理研究所 | Portable hyperspectral reconstitution device based on digital micro reflector |
| CN109557034A (en) * | 2018-11-14 | 2019-04-02 | 南京工程学院 | Compressed sensing based specific gas derived components spectroscopic analysis methods and device |
| CN110031098A (en) * | 2019-03-29 | 2019-07-19 | 中国科学院上海技术物理研究所 | A kind of spectrographic detection chip and reconstructing method based on integrated microcavity |
| CN110658179A (en) * | 2019-10-09 | 2020-01-07 | 上海理工大学 | Laser Raman gas concentration detection method based on multi-signal superposition and pseudo-inverse method |
| CN111366573A (en) * | 2020-03-27 | 2020-07-03 | 合肥金星机电科技发展有限公司 | Evaluation method based on LIBS spectral component analysis result |
| CN112461366A (en) * | 2020-12-16 | 2021-03-09 | 四川长虹电器股份有限公司 | Method for realizing near-infrared spectrometer based on random filter array |
| CN113324920A (en) * | 2021-05-27 | 2021-08-31 | 西安电子科技大学 | Spectral reconstruction method based on micro-nano structure optical filter modulation and sparse matrix transformation |
| CN114779467A (en) * | 2022-04-27 | 2022-07-22 | 吉林大学 | Novel spectrometer membrane system combination selection method based on detector characteristics |
| CN115931130A (en) * | 2022-12-09 | 2023-04-07 | 化学与精细化工广东省实验室 | Algorithm reconstruction narrow-band spectrum detection method and equipment based on pass-band adjustable optical filter |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2005040739A2 (en) * | 2003-10-22 | 2005-05-06 | Softmax, Inc. | System and method for spectral analysis |
| CN101441872A (en) * | 2007-11-19 | 2009-05-27 | 三菱电机株式会社 | Denoising acoustic signals using constrained non-negative matrix factorization |
| CN102466520A (en) * | 2010-11-11 | 2012-05-23 | 香港纺织及成衣研发中心 | Multispectral imaging color measurement system and imaging signal processing method thereof |
| CN102759408A (en) * | 2011-04-25 | 2012-10-31 | 中国科学院空间科学与应用研究中心 | Single-photon counting imaging system and method of same |
-
2013
- 2013-04-16 CN CN2013101320290A patent/CN103207015A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2005040739A2 (en) * | 2003-10-22 | 2005-05-06 | Softmax, Inc. | System and method for spectral analysis |
| CN101441872A (en) * | 2007-11-19 | 2009-05-27 | 三菱电机株式会社 | Denoising acoustic signals using constrained non-negative matrix factorization |
| CN102466520A (en) * | 2010-11-11 | 2012-05-23 | 香港纺织及成衣研发中心 | Multispectral imaging color measurement system and imaging signal processing method thereof |
| CN102759408A (en) * | 2011-04-25 | 2012-10-31 | 中国科学院空间科学与应用研究中心 | Single-photon counting imaging system and method of same |
Non-Patent Citations (2)
| Title |
|---|
| 周昌军等: "基于NMF图像重构的人脸识别", 《计算机工程》 * |
| 颜建军等: "基于NMF的多光谱图像和全色图像融合方法", 《计算机工程》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104359555A (en) * | 2014-10-16 | 2015-02-18 | 中国科学院上海技术物理研究所 | Portable hyperspectral reconstitution device based on digital micro reflector |
| CN109557034A (en) * | 2018-11-14 | 2019-04-02 | 南京工程学院 | Compressed sensing based specific gas derived components spectroscopic analysis methods and device |
| CN110031098A (en) * | 2019-03-29 | 2019-07-19 | 中国科学院上海技术物理研究所 | A kind of spectrographic detection chip and reconstructing method based on integrated microcavity |
| CN110658179A (en) * | 2019-10-09 | 2020-01-07 | 上海理工大学 | Laser Raman gas concentration detection method based on multi-signal superposition and pseudo-inverse method |
| CN110658179B (en) * | 2019-10-09 | 2021-11-19 | 上海理工大学 | Laser Raman gas concentration detection method based on multi-signal superposition and pseudo-inverse method |
| CN111366573B (en) * | 2020-03-27 | 2022-12-20 | 合肥金星智控科技股份有限公司 | Evaluation method based on LIBS spectral component analysis result |
| CN111366573A (en) * | 2020-03-27 | 2020-07-03 | 合肥金星机电科技发展有限公司 | Evaluation method based on LIBS spectral component analysis result |
| CN112461366A (en) * | 2020-12-16 | 2021-03-09 | 四川长虹电器股份有限公司 | Method for realizing near-infrared spectrometer based on random filter array |
| CN112461366B (en) * | 2020-12-16 | 2021-12-21 | 四川长虹电器股份有限公司 | Method for realizing near-infrared spectrometer based on random filter array |
| CN113324920A (en) * | 2021-05-27 | 2021-08-31 | 西安电子科技大学 | Spectral reconstruction method based on micro-nano structure optical filter modulation and sparse matrix transformation |
| CN113324920B (en) * | 2021-05-27 | 2022-05-17 | 西安电子科技大学 | Spectral reconstruction method based on micro-nano structure optical filter modulation and sparse matrix transformation |
| CN114779467A (en) * | 2022-04-27 | 2022-07-22 | 吉林大学 | Novel spectrometer membrane system combination selection method based on detector characteristics |
| CN115931130A (en) * | 2022-12-09 | 2023-04-07 | 化学与精细化工广东省实验室 | Algorithm reconstruction narrow-band spectrum detection method and equipment based on pass-band adjustable optical filter |
| CN115931130B (en) * | 2022-12-09 | 2023-10-24 | 化学与精细化工广东省实验室 | Method and equipment for detecting algorithm reconstruction narrowband spectrum based on passband-adjustable filter |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103207015A (en) | Spectrum reconstruction method and spectrometer device | |
| Choma et al. | Graph neural networks for icecube signal classification | |
| Du et al. | Estimation of total phosphorus concentration using a water classification method in inland water | |
| CN110308547B (en) | Dense sample lens-free microscopic imaging device and method based on deep learning | |
| Xu et al. | Deep Learning Application for Predicting Soil Organic Matter Content by VIS‐NIR Spectroscopy | |
| CN109632693A (en) | A kind of tera-hertz spectra recognition methods based on BLSTM-RNN | |
| CN106788714B (en) | A kind of sparse solution mixing method based on optical computing | |
| CN108593569A (en) | EO-1 hyperion water quality parameter quantitative inversion method based on spectrum morphological feature | |
| CN107894288B (en) | Method and system for measuring vortex beam topological charge under partial coherent light condition | |
| CN113203694B (en) | MSI lake eutrophication index remote sensing estimation method | |
| Kim et al. | Fabrication of 2D thin-film filter-array for compressive sensing spectroscopy | |
| US10458843B2 (en) | Spectrometry apparatus and spectrometry method | |
| CN112881308B (en) | Spectrum camera based on wide spectrum coding and deep learning | |
| CN104142178B (en) | Based on spectral measurement device and the method for filtering membrane array | |
| Cai et al. | Estimating Chemical Oxygen Demand in estuarine urban rivers using unmanned aerial vehicle hyperspectral images | |
| Zhang et al. | Automatic classification of marine plankton with digital holography using convolutional neural network | |
| Ermolli et al. | The prototype RISE-PSPT instrument operating in Rome | |
| Fang et al. | Recent progress and applications of Raman spectrum denoising algorithms in chemical and biological analyses: A review | |
| Di Frischia et al. | Enhanced data augmentation using gans for Raman spectra classification | |
| CN115931123A (en) | Spectrum reconstruction calculation method and spectrum system based on self-adaptive optimization sparse dictionary | |
| Jiao et al. | Deep learning baseline correction method via multi-scale analysis and regression | |
| CN108827882A (en) | Water quality parameter spectroscopic analysis system | |
| Zhang et al. | Developing New Colored Dissolved Organic Matter Retrieval Algorithms Based on Sparse Learning | |
| CN117233116B (en) | Water quality analysis method and system based on machine vision | |
| WO2014201117A1 (en) | Fluorescence detection device, system, and 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 | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20130717 |