CN109993155A - For the characteristic peak extracting method of low signal-to-noise ratio uv raman spectroscopy - Google Patents
For the characteristic peak extracting method of low signal-to-noise ratio uv raman spectroscopy Download PDFInfo
- Publication number
- CN109993155A CN109993155A CN201910327282.9A CN201910327282A CN109993155A CN 109993155 A CN109993155 A CN 109993155A CN 201910327282 A CN201910327282 A CN 201910327282A CN 109993155 A CN109993155 A CN 109993155A
- Authority
- CN
- China
- Prior art keywords
- raman
- spectrum
- region
- peak
- signal
- 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.)
- Granted
Links
Classifications
-
- 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
- G01N21/62—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light
- G01N21/63—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light optically excited
- G01N21/65—Raman scattering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
- G06F2218/10—Feature extraction by analysing the shape of a waveform, e.g. extracting parameters relating to peaks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
- G06F2218/14—Classification; Matching by matching peak patterns
Landscapes
- Health & Medical Sciences (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)
Abstract
Characteristic peak extracting method disclosed by the invention for low signal-to-noise ratio uv raman spectroscopy, belongs to spectrographic detection and spectral processing techniques and field of signal processing.The present invention acquires ultraviolet Raman signal in real time, counts to acquisition Raman signal data, predicts Raman signal effective peak.Each frame Raman spectrum of acquisition is handled, for each frame Raman spectrum, by each frame Raman spectrum spectrum cutting is each piece of region by effective low ebb of acquisition, judges each piece of area attribute of cutting respectively.For each noise region, wherein each point is allowed to be equal to the minimum value in the region, then, by Raman signal region and treated that noise region carries out split.Later, treated N+1 frame spectrum along time shaft is spliced into 2D image, by the bilateral filtering method of iteration, which is filtered, N+1 spectrum is overlapped and is normalized along time shaft after filtering, obtains clean raman characteristic peak spectrum picture.
Description
Technical field
The present invention relates to a kind of characteristic peak extracting method for low signal-to-noise ratio uv raman spectroscopy more particularly to a kind of needles
The method of clean to the acquisition of the ultraviolet Raman signal of real-time low signal-to-noise ratio and effective raman characteristic peak image, belongs to spectrographic detection
With spectral processing techniques and field of signal processing.
Background technique
Raman spectroscopy is a kind of inelastic optical scattering (i.e. energy of incident laser based on high-order harmonics spectrum
Amount/frequency changes) not damaged optical spectrum detecting method.It (is drawn by measuring the tested specific Raman spectrum of molecular system
Graceful finger-print), Non-contact nondestructive detection and quantitative analysis quickly, simple, repeatable, required time can be carried out to sample
It is short, the features such as amount of samples is small, and measurement result is accurate.But the sensitivity of Raman spectroscopic detection it is lower (Rayleigh scattering line it is strong
Degree only has the 10 of incident intensity-3, raman spectrum strength only about Rayleigh line 10-3), especially for some small-sized, just
It takes or easy uv raman spectroscopy detection system, since laser power is not high, laser is not sufficiently stable, non-brake method CCD makes an uproar
Too big etc. factors of sound, detection result influence of noise is larger, and Raman signal signal-to-noise ratio is low, it is difficult to identify raman characteristic peak.
Some traditional signal processing methods, can remove some random noises, to make raman characteristic peak significant, still
For might not be ideal for Raman spectrum, because many faint Raman peaks are either from amplitude or waveform,
With noise all very close to.So can also lose many faint Raman useful signals, signal-to-noise ratio during removing random noise
It is not improved too many.So the method for some novel filtering in time-domain is suggested, can retain some with not
The small-signal of the certain waveform changed over time, such as predictive filtering method and Bayesian forecasting filtering algorithm on t-x-y.
Only single sample is detected since Raman system is usually disposable, and the Raman signal that single sample generates removes
Amplitude variation is outer, and Raman signatures peak position fixes, and noise has randomness.In this way in the sample of one sample of acquisition in real time
When spectrum picture, to distinguish the noise signal for not stopping variation at any time, it can be filtered out by counting indeclinable peak position at any time
These noise signals, interested effective Raman signal are highlighted.2D spectrum picture construction method in time-domain is utilized simultaneously
, the spectrum in x-axis is subjected to longitudinal expansion along t axis, the t axis is time shaft, and the x-axis is wavelength or Raman shift
Axis, the spectrum picture for then choosing the effective noise region counted is area-of-interest, selects to have and protects the smooth of side characteristic
Filtering method is filtered area-of-interest, the noise random at any time near effective Raman peaks is filtered out, to obtain out
Clean and effective Raman peaks image, so as to operations such as the identification matchings of subsequent progress, to improve the identification of rear end Raman signal
Accuracy rate.
In conclusion after by the way that Raman detection signal is expanded in time-domain, the statistical classification of progress and emerging for feeling
The 2D filtering method in the t-x spectrum picture region of interest, is that can effectively extract characteristic peak, it is distortionless to get clean peak value
Raman characteristic peak image, solves that small-sized, portable or easy uv raman spectroscopy detection system Raman signal signal-to-noise ratio is low to be made
It obtains faint raman characteristic peak and is difficult to the effective ways extracted.
Summary of the invention
Characteristic peak extracting method technical problems to be solved disclosed by the invention for low signal-to-noise ratio uv raman spectroscopy
It is: is extracted after processing for real-time raman spectral signal especially for Raman signal-to-noise ratio low Raman spectroscopic detection system
Clean out and effective raman characteristic peak image carries out the operations such as identification matching so as to subsequent, to improve the knowledge of rear end Raman signal
Other accuracy rate.The low Raman spectroscopic detection system of the Raman signal-to-noise ratio refers in particular to small-sized, portable or easy ultraviolet drawing
Graceful spectrum investigating system.
The purpose of the present invention is what is be achieved through the following technical solutions.
Characteristic peak extracting method disclosed by the invention for low signal-to-noise ratio uv raman spectroscopy, acquires ultraviolet Raman in real time
Signal counts acquisition Raman signal data, and then predicts Raman signal effective peak, and the effective peak includes effective
Low ebb and peak.Each frame Raman spectrum of acquisition is handled, for each frame Raman spectrum, passes through the effective of acquisition
Each frame Raman spectrum spectrum cutting is each piece of region by low ebb, judges each piece of area attribute of cutting respectively, and the region belongs to
Property refers to noise region attribute or effective Raman signal area attribute.For each noise region, allow wherein each point be equal to the area
The minimum value in domain, then, by Raman signal region, noise region carries out split with treated.It later, will treated N+1 frame
Spectrum is spliced into 2D image along time shaft, by the bilateral filtering method of iteration, is filtered to the 2D image, after filtering along when
Between axis N+1 spectrum is overlapped and is normalized, obtain clean raman characteristic peak spectrum picture.
Characteristic peak extracting method disclosed by the invention for low signal-to-noise ratio uv raman spectroscopy, includes the following steps:
Step 1: acquiring ultraviolet Raman signal in real time, counts to acquisition Raman signal data, and then predicts Raman letter
Number effective peak, the effective peak includes effective low ebb and peak.
By obtaining ultraviolet Raman signal in real time in the spectrometer in the Raman detection system of front end, Raman signal will be obtained in real time
It is transmitted to signal buffer area, the signal buffer area is every to obtain the new spectrum of a frame for keeping access N+1 frame spectrum, can delete
Except N+1 frame spectroscopic data before the, Raman signal data acquisition is completed.Low ebb and peak statistics are carried out to N+1 frame image, listed low
Paddy netlist and peak netlist.Default statistical classification tolerance interval step-length, then entire spectrum can be divided into M tolerance interval along x-axis,
The x-axis be wavelength axis or Raman shift axis, if the peak netlist of N frame spectrum have in each tolerance interval peak and
The absolute value of the slope at section left or right end is greater than preset threshold Kmax, that is, there is N frame spectrum to be all satisfied following formula condition,
There are f'(x in tolerance intervalm) > Kmax&&f'(xm+1)≤0,m∈Cm, preset threshold K described in m=1,2 ..., Mmax
Equal to greatest gradient absolute value caused by random noise energy, then it is labeled as effective peak, effective peak is the high drawing of intensity
Man Feng records its position.
And for the low ebb netlist of N+1 frame spectrum, the Rule of judgment of low valley point is given by,
There are f'(x in tolerance intervalm)≤0&&f'(xm+1)≥0,m∈Cm, m=1,2 ..., M
Then, real spectrum low valley point is predicted by the following method, is divided into following three kinds of situations and is predicted real spectrum
Low valley point: situation one, in a tolerance interval step-length, it is not then effective that being less than or equal to (N+1)/2 frame spectrum, there are low valley points
Low ebb；Situation two, in a tolerance interval step-length, it is then effective low ebb that more than or equal to N frame spectrum, there are low valley points, according to
Average value or weight equal value are demarcated this effective low ebb position and are recorded；Situation three, in a tolerance interval step-length, be greater than (N
+ 1)/2 it is less than N frame spectrum the case where there are low valley points, the situation is possible as low-intensity peak value shape affected by noise
At, effective low ebb is judged whether it is by default predicted condition.Real spectrum low valley point is predicted by above-mentioned three kinds of situations,
And then obtain out low ebb netlist.Each effective peak position of Raman signal is predicted by the low ebb netlist and peak netlist of statistics.
It is as follows for judging whether it is the specific judgment method of default predicted condition of effective low ebb described in step 1: observation
In tolerance interval step-length region similar in the low ebb value or so, if there are the high Raman peaks of the intensity recorded in the netlist of peak,
If it exists it is effective low ebb, records into low ebb netlist, if it does not exist, be then overlapped N+1 frame spectrum, if superposition spectrum exists
There are low ebbs in the tolerance interval step-length, then are effective low ebb, record into low ebb netlist.
Statistical classification tolerance interval step-length is preferably 3 described in step 1.
Step 2: being directed to each frame Raman spectrum, and the effective low ebb obtained by step 1 is by each frame Raman spectrum light
Spectrum segmentation is each piece of region, judges that each piece of area attribute of cutting, the area attribute refer to noise region attribute or effective respectively
Raman signal area attribute.For each noise region, wherein each point is allowed to be equal to the minimum value in the region, then, by Raman
Noise region carries out split with treated for signal area.
By the low ebb netlist for each frame Raman spectrum that step 1 predicts, each frame Raman light that step 1 is obtained
Spectrum segmentation is each piece of region, carries out determined property to each region, determines each piece of area attribute respectively, i.e., determine each piece of area respectively
Domain is noise region, or effectively Raman signal region.For single region, there are two judgment criterias: the first judgment criteria
Are as follows: when there are the Raman peaks that the intensity recorded in the peak netlist obtained in step 1 is high in region, it is determined that be effective Raman
Signal area；Second of judgment criteria are as follows: p tolerance interval length in selecting step one is step-length, carries out least square method
Smoothly, then the smooth rear standard deviation of data and the ratio of average value in the region are judged, if the ratio of standard deviation and average value is big
In a preset threshold, it is determined that for effective Raman signal region be then otherwise noise region.The preset threshold is all
It is step-length that region, which takes p tolerance interval length in step 1, the ratio of standard deviation and average value after carrying out least square method smoothly
The half of the average value of value.The first judgment criteria is for detecting strong peak region；Second of judgment criteria is for being directed to
Weak peak is judged.
After determining each piece of area attribute, for each noise region, wherein each point is allowed to be equal to the minimum value in the region, note
Record all wavelength for being judged as noise attribute region or Raman shift location information.Then, by effective Raman signal region and filter
Noise region after wave carries out split.
P quantity is preferably 3 in p tolerance interval described in step 2.
The principle of second of judgment criteria is that the Raman width of weak peak is greater than sampling width, and noise width is that sampling is wide
Degree estimates step-length according to weak peak mean breadth and carries out least squares filtering, and noise smoothing is near linear by least squares filtering
Region, and weak peak can be effectively retained, the ratio of smooth rear near linear regional standard and average value can be significantly less than weak peak area
The standard deviation in domain and the ratio of average value are judged by the standard deviation in the calculating region with toaverage ratio size later
It is noise region or effective Raman signal region.
Step 3: each frame Raman spectrum that step 1 acquires to be carried out to the processing of step 2, after step 2 is handled
N+1 frame spectrum be spliced into 2D image along time shaft, by the bilateral filtering method of iteration, which is filtered, filter
N+1 spectrum is overlapped and is normalized along time shaft after wave, obtains clean raman characteristic peak spectrum picture.
Due to the effective not processed mistake in Raman signal region judged in step 2, so still containing noise；Spectrum
2D building and filtering principle on t-x are spectrum to be launched into time t-x image, since noise exists on a timeline
Randomness, and effectively peak position is constant on a timeline for Raman signal.So smothing filtering is carried out to t-x image, it can not only be
In conventional x-axis, i.e., on wavelength or Raman shift axis, filtering out partial noise makes curve smoothing, and the spectrum recombined is allowed to become
It is uniform continuous, and can filter out the noise changed over time on t axis, i.e., on time shaft, can reduce in step 2 not
The noise in effective Raman signal region of processing is without influencing Raman signal feature peak-to-peak value itself.
Each frame Raman spectrum that step 1 acquires is subjected to the processing of step 2, by step 2 treated N+1 frame
Spectrum is spliced into 2D image along time shaft, and x-axis is wavelength or Raman shift axis at this time, and t axis is time shaft, and z-axis is that Raman is strong
Degree.By the bilateral filtering method of iteration, which is filtered, after filtering, along time shaft, N+1 spectrum is folded
Adduction normalization, obtains clean Raman peak values spectrum picture.
The principle of the bilateral filtering method are as follows: in filtering algorithm, the pixel value on target point is usually where it
The value of one small local neighbor pixel around on position is determined.Specific implementation in 2D gaussian filtering is i.e. to surrounding
Pixel value in preset range is assigned to different Gauss weighted values respectively, and the most termination of current point is obtained after weighted average
Fruit.And the Gauss weight factor is generated using the space length relationship between two pixels.Its description one formulated
As it is as described below:
C therein is the Gauss weight based on space length, and kd(x) it is used to carry out result unitization.
Gaussian filtering only considers the relationship on the spatial position between pixel, therefore the result filtered in low-pass filtering algorithm
The information at edge can be lost, the marginal information is exactly effective Raman signal that peak position is stable at any time.And bilateral filtering is just
It is that joined an other weight branch in gaussian filtering to solve the problems, such as this.For the holding at edge in bilateral filtering
It is realized by following expressions:
S therein is the Gauss weight based on similarity degree between pixel, kr(x) same unitization for being carried out to result.
The bilateral filtering based on space length, similarity degree comprehensive consideration can be obtained by being combined to the two:
The comprehensive two kinds of Gauss weights of unitization branch k (x) in above formula are obtained in together, and c therein and s are calculated in detail
It is described as follows:
And there is d (ξ, x)=d (ξ-x)=‖ ξ-x ‖
And there is σ (φ, f)=σ (φ-f)=‖ φ-f ‖
The above-mentioned expression formula provided is unlimited integral spatially, and needs to carry out it in the image of pixelation
Discretization.Distance is more than that the pixel of predetermined extent actually influences very little to current object pixel, be can ignore that.Restriction office
Discretization simplified formula after portion's subregion is following form:
The operating procedure of the bilateral filtering method of the iteration are as follows: (1) set bilateral filtering function space domain sigma because
Sub- sigma_s, the area of space refer to the region t-x, the i.e. region 2D of time and wavelength or Raman shift composition, the sky
Between domain sigma predictor selection should include as far as possible more multiframe spectrum, but need to be less than Raman signatures peak value width；(2) it sets bilateral
Filter function pixel coverage domain sigma factor sigma_r, the pixel coverage region refer to every frame spectrally each wavelength or Raman
Raman scattering intensity range at intensity, preset step-length can be halved or subtract by returning to setting sigma_r, sigma_r every time, until
Less than set minimum Smin, then keep sigma_r=Smin constant, sigma_r initial value is chosen, and is obtained according to step 2
All noise regions standard deviation half and divided by all noise region average value sums as a result, retain two-decimal after obtain
?；(3) bilateral filtering is carried out to the 2D image of building；(4) it according to the location information of the noise region recorded in step 2, calculates
The standard deviation and toaverage ratio size of each noise region of filtered image return if the ratio is greater than preset threshold value
(2).(5) iteration result after finally filtering is obtained.
The advantage of the bilateral filtering method is, realizes and is being effectively retained the stable effective Raman letter of peak position at any time
On the basis of number, substantially reduce the noise size changed at random at any time.It, can be very and by adjusting sigma_r parameter size
Weight is set well, by iteration, improves single smoothed precision, will not be obscured as disposably smoothly in smooth out noise
Effective Raman signal.
The utility model has the advantages that
1, the characteristic peak extracting method disclosed by the invention for low signal-to-noise ratio uv raman spectroscopy is believed according to ultraviolet Raman
Number and random noise physical characteristic, using statistical classification, 2D building and iteration bilateral filtering method, compared to common spectrum
Superposition or characteristic matching extracting method can more rapidly and effectively obtain clean and distortionless raman characteristic peak and extract knot
Fruit spectrum picture, and this method is easy to transplant in various systems.
2, the characteristic peak extracting method disclosed by the invention for low signal-to-noise ratio uv raman spectroscopy is made an uproar at random by analysis
The physical characteristic of sound is effectively predicted and judges effective low ebb and peak, and is effectively classified after spectrum is divided, and individually right
Noise region is handled, this method can the interested effective raman characteristic peak signal location of quick lock in, and quickly pre- place
Reason falls most uninterested random noise region, compared to the commonly not extraction raman characteristic peak image side of statistical classification
Method, this method is capable of providing effective target, more efficient, and making will not be by uninterested noise region when extracting peak value
It influences, the noise region of statistics and effective Raman signal region are also capable of providing good random noise statistical property and signal system
Characteristic is counted, provides reference for the parameter preset or threshold value of subsequent algorithm.
3, existing Raman spectrum filtering method focuses primarily upon the filtering method of spatial domain, disclosed by the invention to be directed to
The characteristic peak extracting method of low signal-to-noise ratio uv raman spectroscopy, due to effective Raman signal change over time peak position stablize, when
Between limbus straight line is formed on axis, and noise changes at random at any time, forms random noise on a timeline, draws on time shaft
Enter more physical features and is more conducive to Filtering Analysis.
4, the characteristic peak extracting method disclosed by the invention for low signal-to-noise ratio uv raman spectroscopy, is calculated using bilateral filtering
The guarantor side characteristic of method, not only in the Raman spectral image of 2D building, removal does not constitute the noise at edge, retains and constitutes edge
Raman signatures peak-to-peak signal, and the Gauss weighted factor of very little can be set in bilateral filtering, by alternative manner, can guarantee to filter
Wave precision is not in disposably to filter out the case where too much noise reduction spectral results being distorted.
Detailed description of the invention
Fig. 1 is the characteristic peak extracting method main-process stream schematic diagram for low signal-to-noise ratio uv raman spectroscopy of the invention.
Fig. 2 is step one flow diagram of the invention.
Fig. 3 is step two flow diagram of the invention.
Fig. 4 is step three flow diagram of the invention, and wherein Fig. 4 a is that 2D spectrum picture constructs schematic diagram, and Fig. 4 b is repeatedly
The bilateral filtering method processing flow schematic diagram in generation.
Fig. 5 is of the invention to the real-time uv raman spectroscopy processing result example schematic of heroin, and wherein Fig. 5 a is not
One frame Raman spectrum original image of processing, Fig. 5 b are temporary 10 frame Raman spectrum original images to be processed, and Fig. 5 c is Raman after processing
Characteristic peak image zooming-out result.
Fig. 6 be it is of the invention to the real-time uv raman spectroscopy processing result example schematic of other samples, wherein Fig. 6 a is
The original image of untreated frame paracetamol tablets Raman spectrum, Fig. 6 b are treated paracetamol tablets Raman signatures
For peak image zooming-out as a result, Fig. 6 c is the original image of untreated frame cefixime dispersible tablet Raman spectrum, Fig. 6 d is that treated
Cefixime dispersible tablet raman characteristic peak image zooming-out result.
Specific embodiment
Objects and advantages in order to better illustrate the present invention with reference to the accompanying drawing do further summary of the invention with example
Explanation.
Embodiment 1:
Now using laboratory from grinding portable 266nm laser excitation Raman spectrum system, to heroin Heroin sample into
Row Raman detection, acquired integrated time are 2 seconds, open baseline correction function, that is, can remove spectrum when obtaining every frame Raman spectrum
Fluorescence bottom is made an uproar, and N=9 is chosen, that is, temporary to have preceding 9 frame spectrum when receiving the new spectrum of a frame, x-axis is Raman shift, and sample, which is stablized, to be put
Start to carry out algorithm process after setting 6 seconds, for characteristic peak extracting method process flow such as Fig. 1 of low signal-to-noise ratio uv raman spectroscopy
It is shown.
As shown in Figure 1, the disclosed characteristic peak extracting method for being directed to low signal-to-noise ratio uv raman spectroscopy of the present embodiment, specifically
Implementation method is as follows:
Step 1: as shown in Fig. 2, in real time acquisition the ultraviolet Raman signal of heroin sample, to acquisition Raman signal data into
Row statistics, and then predict Raman signal effective peak, the effective peak includes effective low ebb and peak.
By obtaining the ultraviolet Raman signal of heroin sample in the spectrometer in the Raman detection system of front end in real time, pass through USB
The real-time Raman signal that obtains is transmitted to pcs signal buffer area by communication, and the signal buffer area is for keeping 10 frame light of access
Spectrum, every to obtain the new spectrum of a frame, 10 frame spectroscopic datas before will be deleted the complete Raman signal data acquisition.To this 10 frame figure
As carrying out low ebb and peak statistics, low ebb netlist and peak netlist are listed.Default statistical classification tolerance interval step-length is 3, if 9 frames
The peak netlist of spectrum is greater than 240 at the section peak Nei Douyou that step-length is 3 and in the absolute value of the slope at section left or right end,
It is then labeled as effective peak, effective peak is the high Raman peaks of intensity, records its position.And for the low ebb of 10 frame spectrum
Netlist predicts real spectrum low valley point by the following method, is divided into following three kinds of situations and predicts real spectrum low valley point: feelings
Condition one has low ebb less than or equal to 5 frame spectrum in step-length 3, then is not effective low ebb；Situation two is more than or equal to 9 frame spectrum one
There is low ebb in a step-length 3, is then effective low ebb, records the Raman shift position of the low ebb；Situation three, be greater than 5 and less than 9 frame light
Spectrum has the case where low ebb in a step-length 3, observes the close step-length of the low ebb value or so and is in 3 regions, if there are peak nets
The high Raman peaks of the intensity recorded in table are effective low ebb if it exists, are recorded into low ebb netlist, if it does not exist, then by 10 frames
Spectrum is overlapped, if there are low ebbs in 3 section records for effective low ebb into low ebb netlist, by upper for superposition spectrum
It states three kinds of situations and predicts real spectrum low valley point, and then obtain out low ebb netlist.Pass through the low ebb netlist and peak net of statistics
Table predicts each effective peak position of Raman signal.
Step 2: as shown in figure 3, being directed to each frame Raman spectrum, the effective low ebb obtained by step 1 is by each frame
Raman spectrum spectrum cutting is each piece of region, judges that each piece of area attribute of cutting, the area attribute refer to noise region respectively
Attribute or effective Raman signal area attribute.For each noise region, wherein each point is allowed to be equal to the minimum value in the region, so
Afterwards, by Raman signal region, noise region carries out split with treated.
By the low ebb netlist for each frame Raman spectrum that step 1 predicts, each frame Raman light that step 1 is obtained
Spectrum segmentation is each piece of region, carries out determined property to each region, determines each piece of area attribute respectively, i.e., determine each piece of area respectively
Domain is noise region, or effectively Raman signal region.For single region, there are two judgment criterias: the first judgment criteria
Are as follows: when there are the Raman peaks that the intensity recorded in the peak netlist obtained in step 1 is high in region, it is determined that be effective Raman
Signal area；Second of judgment criteria are as follows: selection length 9 is step-length, it is smooth to carry out Least Square Method, then judge the region
The interior smooth rear standard deviation of data and the ratio of average value, if the ratio of standard deviation and average value is greater than 0.19, it is determined that have
Otherwise imitating Raman signal region is then noise region.
After determining each piece of area attribute, for each noise region, wherein each point is allowed to be equal to the minimum value in the region, note
Record all wavelength for being judged as noise attribute region or Raman shift location information.Then, by effective Raman signal region and filter
Noise region after wave carries out split.
Step 3: as shown in fig. 4 a, each frame Raman spectrum that step 1 acquires to be carried out to the processing of step 2, will be walked
Rapid two treated 10 frame spectrum are spliced into 2D image along time shaft, pass through the bilateral filtering side of iteration as shown in Figure 4 b
Method is filtered the 2D image, after filtering, along time shaft, 10 spectrum is overlapped and is normalized, obtain clean Raman
Characteristic peak spectrum picture.
Each frame Raman spectrum that step 1 acquires is subjected to the processing of step 2, by step 2 treated 10 frame light
Spectrum, is spliced into 2D image along time shaft, and x-axis is wavelength or Raman shift axis at this time, and t axis is time shaft, and z-axis is raman scattering intensity.
By the bilateral filtering method of iteration, which is filtered.
As shown in Figure 4 b, the operating procedure of the bilateral filtering method of the iteration are as follows: (1) it is empty to set bilateral filtering function
Between domain sigma factor sigma_s=10；(2) bilateral filtering function pixel coverage domain the sigma factor sigma_r, sigma_ are set
R initial value is chosen for 0.16, returns to setting sigma_r every time, sigma_r can halve, until less than 0.02, then keeping
Sigma_r=0.02 is constant；(3) bilateral filtering is carried out to the 2D image of building；(4) currently processed result images and upper one are calculated
The mean difference of each position intensity of secondary processing result image returns to (2) if the difference is greater than 0.0001.(5) it obtains final
Iteration result after filtering.
After filtering, along time shaft, 10 frame spectrum is overlapped and is normalized, clean raman characteristic peak spectrogram is obtained
Picture.
Fig. 5 a, illustrates untreated frame heroin Raman spectrum original image, and Fig. 5 b illustrates temporary to be processed 10
Frame heroin Raman spectrum original image, Fig. 5 c are illustrated by this patent method treated heroin raman characteristic peak image zooming-out
As a result.
Embodiment 2:
Using the characteristic peak extracting method for low signal-to-noise ratio uv raman spectroscopy of this patent respectively to acetyl ammonia
The real-time ultraviolet Raman detection signal of base phenol piece and cefixime dispersible tablet is handled, the acquisition side of real-time ultraviolet Raman signal
Method and condition and embodiment 1 are consistent.The characteristic peak extracting method for low signal-to-noise ratio uv raman spectroscopy of this patent is implemented
When paracetamol tablets, the slope absolute value threshold value for the section left end that step 1 is chosen is 35, step 2 Plays difference peace
The threshold value of the ratio of mean value is 0.175, and the initialization value of sigma_r chooses 0.04, Smin and is chosen for 0.016 in step 3.This
The characteristic peak extracting method for low signal-to-noise ratio uv raman spectroscopy of patent is implemented in cefixime dispersible tablet, step 1 choosing
The slope absolute value threshold value of the section left end taken is 50, and the threshold value of the ratio of step 2 Plays difference and average value is 0.12, step
The initialization value of sigma_r chooses 0.1, Smin and is chosen for 0.015 in rapid three.
Fig. 6 a illustrates the original image of untreated frame paracetamol tablets Raman spectrum, Fig. 6 b, after illustrating processing
Paracetamol tablets raman characteristic peak image zooming-out result；Fig. 6 c illustrates untreated frame cefixime dispersible tablet
The original image of Raman spectrum, Fig. 6 d, the cefixime dispersible tablet Raman raman characteristic peak image zooming-out result that illustrates that treated.
Above-described specific descriptions have carried out further specifically the purpose of invention, technical scheme and beneficial effects
It is bright, it should be understood that the above is only a specific embodiment of the present invention, the protection model being not intended to limit the present invention
It encloses, all within the spirits and principles of the present invention, any modification, equivalent substitution, improvement and etc. done should be included in the present invention
Protection scope within.
Claims (8)
1. being directed to the characteristic peak extracting method of low signal-to-noise ratio uv raman spectroscopy, it is characterised in that: include the following steps,
Step 1: acquiring ultraviolet Raman signal in real time, counts to acquisition Raman signal data, and then predicts that Raman signal has
Peak value is imitated, the effective peak includes effective low ebb and peak；
Step 2: being directed to each frame Raman spectrum, is cut each frame Raman spectrum spectrum by effective low ebb that step 1 obtains
It is divided into each piece of region, judges that each piece of area attribute of cutting, the area attribute refer to noise region attribute or effective Raman respectively
Signal area attribute；For each noise region, wherein each point is allowed to be equal to the minimum value in the region, then, by Raman signal
Noise region carries out split with treated in region；
Step 3: each frame Raman spectrum that step 1 acquires is carried out the processing of step 2, by step 2 treated N+1
Frame spectrum is spliced into 2D image along time shaft, by the bilateral filtering method of iteration, is filtered to the 2D image, edge after filtering
N+1 spectrum is overlapped and is normalized by time shaft, obtains clean raman characteristic peak spectrum picture.
2. being directed to the characteristic peak extracting method of low signal-to-noise ratio uv raman spectroscopy as described in claim 1, it is characterised in that: step
A rapid implementation method is,
By obtaining ultraviolet Raman signal in real time in the spectrometer in the Raman detection system of front end, Raman signal transmission will be obtained in real time
To signal buffer area, the signal buffer area is every to obtain the new spectrum of a frame for keeping access N+1 frame spectrum, N before deleting the
+ 1 frame spectroscopic data completes Raman signal data acquisition；Low ebb and peak statistics are carried out to N+1 frame image, list low ebb netlist
With peak netlist；Default statistical classification tolerance interval step-length, entire spectrum are divided into M tolerance interval along x-axis, and the x-axis is
Wavelength axis or Raman shift axis, if the peak netlist of N frame spectrum has peak in each tolerance interval and in section left or right
The absolute value of the slope at end is greater than preset threshold Kmax, that is, there is N frame spectrum to be all satisfied following formula condition,
There are f'(x in tolerance intervalm) > Kmax&&f'(xm+1)≤0,m∈Cm, preset threshold K described in m=1,2 ..., MmaxIt is equal to
Greatest gradient absolute value caused by random noise energy is then labeled as effective peak, and effective peak is the high Raman peaks of intensity,
Record its position；
And for the low ebb netlist of N+1 frame spectrum, the Rule of judgment of low valley point is given by,
There are f'(x in tolerance intervalm)≤0&&f'(xm+1)≥0,m∈Cm, m=1,2 ..., M
Then, real spectrum low valley point is predicted by the following method, is divided into following three kinds of situations and is predicted real spectrum low ebb
Point: situation one, in a tolerance interval step-length, being less than or equal to (N+1)/2 frame spectrum, there are low valley points, then are not effectively low
Paddy；Situation two, in a tolerance interval step-length, it is then effective low ebb that more than or equal to N frame spectrum, there are low valley points, according to flat
Mean value or weight equal value are demarcated this effective low ebb position and are recorded；Situation three, in a tolerance interval step-length, be greater than (N+
1)/2 it is less than N frame spectrum the case where there are low valley points, the situation is possible as low-intensity peak value shape affected by noise
At, effective low ebb is judged whether it is by default predicted condition；Real spectrum low valley point is predicted by above-mentioned three kinds of situations,
And then obtain out low ebb netlist；Each effective peak position of Raman signal is predicted by the low ebb netlist and peak netlist of statistics.
3. being directed to the characteristic peak extracting method of low signal-to-noise ratio uv raman spectroscopy as claimed in claim 2, it is characterised in that: step
Rapid two implementation method is,
By the low ebb netlist for each frame Raman spectrum that step 1 predicts, each frame Raman spectrum that step 1 obtains is cut
It is divided into each piece of region, determined property is carried out to each region, each piece of area attribute is determined respectively, i.e., determines that each piece of region is respectively
Noise region, or effectively Raman signal region；For single region, there are two judgment criterias: the first judgment criteria are as follows: when
There are the Raman peaks that the intensity recorded in the peak netlist obtained in step 1 is high in region, it is determined that is effective Raman signal area
Domain；Second of judgment criteria are as follows: p tolerance interval length in selecting step one is step-length, and progress least square method is smooth, then
Judge the smooth rear standard deviation of data and the ratio of average value in the region, is preset if the ratio of standard deviation and average value is greater than one
Threshold value, it is determined that for effective Raman signal region be then otherwise noise region；The preset threshold is that all areas take step
P tolerance interval length in rapid one is step-length, and the ratio of standard deviation and average value is averaged after carrying out least square method smoothly
The half of value；The first judgment criteria is for detecting strong peak region；Second of judgment criteria is used to carry out for weak peak
Judgement；
After determining each piece of area attribute, for each noise region, wherein each point is allowed to be equal to the minimum value in the region, record institute
There are the wavelength for being judged as noise attribute region or Raman shift location information；It then, will be behind effective Raman signal region and filtering
Noise region carry out split.
4. being directed to the characteristic peak extracting method of low signal-to-noise ratio uv raman spectroscopy as claimed in claim 3, it is characterised in that: step
Rapid three implementation method is,
Each frame Raman spectrum that step 1 acquires is subjected to the processing of step 2, by step 2 treated N+1 frame spectrum,
Along time shaft, it is spliced into 2D image, x-axis is wavelength or Raman shift axis at this time, and t axis is time shaft, and z-axis is raman scattering intensity；It is logical
The bilateral filtering method for crossing iteration is filtered the 2D image, after filtering, along time shaft, N+1 spectrum is overlapped and is returned
One changes, and obtains clean Raman peak values spectrum picture.
5. the characteristic peak extracting method for low signal-to-noise ratio uv raman spectroscopy as described in claim 1,2,3 or 4, feature
Be: as follows for judging whether it is the specific judgment method of default predicted condition of effective low ebb described in step 1: observation should
In tolerance interval step-length region similar in low ebb value or so, if there are the high Raman peaks of the intensity recorded in the netlist of peak, if
It in the presence of being effective low ebb, records into low ebb netlist, if it does not exist, is then overlapped N+1 frame spectrum, if superposition spectrum is at this
There are low ebbs in tolerance interval step-length, then are effective low ebb, record into low ebb netlist.
6. being directed to the characteristic peak extracting method of low signal-to-noise ratio uv raman spectroscopy as claimed in claim 5, it is characterised in that: step
A rapid statistical classification tolerance interval step-length is selected as 3.
7. the characteristic peak extracting method for low signal-to-noise ratio uv raman spectroscopy as described in claim 1,2,3 or 4, feature
Be: p quantity is selected as 3 in p tolerance interval described in step 2；
The principle of second of judgment criteria is that the Raman width of weak peak is greater than sampling width, and noise width is sampling width, root
Step-length, which to be estimated, according to weak peak mean breadth carries out least squares filtering, noise smoothing is near linear region by least squares filtering,
And weak peak can be effectively retained, the ratio of smooth rear near linear regional standard and average value can be significantly less than the mark in weak peak region
The ratio of quasi- difference and average value, later by calculating standard deviation and the toaverage ratio size in the region to judge it is noise
Region or effectively Raman signal region.
8. the characteristic peak extracting method for low signal-to-noise ratio uv raman spectroscopy as described in claim 1,2,3 or 4, feature
It is: the operating procedure of the bilateral filtering method of the iteration are as follows: (1) set the bilateral filtering function space domain sigma factor
Sigma_s, the area of space refer to the region t-x, the i.e. region 2D of time and wavelength or Raman shift composition, the space
Domain sigma predictor selection should include as far as possible more multiframe spectrum, but need to be less than Raman signatures peak value width；(2) bilateral filter is set
Wave function pixel coverage domain sigma factor sigma_r, the pixel coverage region refers to every frame, and spectrally each wavelength or Raman are strong
Raman scattering intensity range at degree, preset step-length, Zhi Dao little can be halved or subtract by returning to setting sigma_r, sigma_r every time
In set minimum Smin, then keep sigma_r=Smin constant, sigma_r initial value is chosen, and is obtained according to step 2
The half of the standard deviation of all noise regions and divided by all noise region average value sums as a result, retain two-decimal after obtain
?；(3) bilateral filtering is carried out to the 2D image of building；(4) it according to the location information of the noise region recorded in step 2, calculates
The standard deviation and toaverage ratio size of each noise region of filtered image return if the ratio is greater than preset threshold value
(2)；(5) iteration result after finally filtering is obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910327282.9A CN109993155B (en) | 2019-04-23 | 2019-04-23 | Characteristic peak extraction method for low signal-to-noise ratio ultraviolet Raman spectrum |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910327282.9A CN109993155B (en) | 2019-04-23 | 2019-04-23 | Characteristic peak extraction method for low signal-to-noise ratio ultraviolet Raman spectrum |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109993155A true CN109993155A (en) | 2019-07-09 |
| CN109993155B CN109993155B (en) | 2021-01-01 |
Family
ID=67135082
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910327282.9A Active CN109993155B (en) | 2019-04-23 | 2019-04-23 | Characteristic peak extraction method for low signal-to-noise ratio ultraviolet Raman spectrum |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109993155B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112161966A (en) * | 2020-09-29 | 2021-01-01 | 中国科学院长春光学精密机械与物理研究所 | Method and device for separating Raman spectrum of sample containing fluorescence spectrum |
| WO2021149760A1 (en) * | 2020-01-24 | 2021-07-29 | パナソニックＩｐマネジメント株式会社 | Raman spectroscopy method and raman spectroscopy suppport device |
| CN113567604A (en) * | 2021-07-22 | 2021-10-29 | 华谱科仪(大连)科技有限公司 | Detection and analysis method of chromatographic spectrogram and electronic equipment |
| CN113640445A (en) * | 2021-08-11 | 2021-11-12 | 贵州中烟工业有限责任公司 | Characteristic peak identification method based on image processing, computing equipment and storage medium |
| CN114113035A (en) * | 2021-11-18 | 2022-03-01 | 北京理工大学 | Transgenic soybean oil identification method |
| CN114280002A (en) * | 2021-12-16 | 2022-04-05 | 宜宾五粮液股份有限公司 | Abnormal fermented grain spectrum screening method based on characteristic peak determination |
| CN115001571A (en) * | 2022-05-11 | 2022-09-02 | 青岛诺克通信技术有限公司 | Self-adaptive digital filtering method of optical time domain reflectometer and optical time domain reflectometer |
| CN115078616A (en) * | 2022-05-07 | 2022-09-20 | 天津国科医工科技发展有限公司 | Multi-window spectral peak identification method, device, medium and product based on signal-to-noise ratio |
| CN115389481A (en) * | 2022-08-16 | 2022-11-25 | 华中科技大学 | Method for detecting uniformity of biomass surface coating based on Raman spectrum surface scanning |
| WO2023272749A1 (en) * | 2021-06-29 | 2023-01-05 | 江苏大学 | Feature information extraction method and portable detection system for small molecule volatile substance |
| CN115078616B (en) * | 2022-05-07 | 2024-06-07 | 天津国科医疗科技发展有限公司 | Multi-window spectrum peak identification method, equipment, medium and product based on signal to noise ratio |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103743717A (en) * | 2013-12-03 | 2014-04-23 | 中国工程物理研究院化工材料研究所 | Explosive CL-20 crystal form quantitative analysis Raman feature region determination method |
| CN105138826A (en) * | 2015-08-10 | 2015-12-09 | 厦门大学 | Raman signal reconstruction method under strong noise background |
| CN107179310A (en) * | 2017-06-01 | 2017-09-19 | 温州大学 | Raman spectrum characteristic peak recognition methods based on robust noise variance evaluation |
| CN107764797A (en) * | 2017-09-21 | 2018-03-06 | 天津大学 | A kind of Raman spectral image data preprocessing method based on low-rank tensor algorithm |
| CN107818298A (en) * | 2017-09-27 | 2018-03-20 | 厦门大学 | General Raman spectral characteristics extracting method for machine learning material recognition |
| CN108918499A (en) * | 2018-06-28 | 2018-11-30 | 华南师范大学 | The method of Raman baseline drift is removed in Raman map |
-
2019
- 2019-04-23 CN CN201910327282.9A patent/CN109993155B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103743717A (en) * | 2013-12-03 | 2014-04-23 | 中国工程物理研究院化工材料研究所 | Explosive CL-20 crystal form quantitative analysis Raman feature region determination method |
| CN105138826A (en) * | 2015-08-10 | 2015-12-09 | 厦门大学 | Raman signal reconstruction method under strong noise background |
| CN107179310A (en) * | 2017-06-01 | 2017-09-19 | 温州大学 | Raman spectrum characteristic peak recognition methods based on robust noise variance evaluation |
| CN107764797A (en) * | 2017-09-21 | 2018-03-06 | 天津大学 | A kind of Raman spectral image data preprocessing method based on low-rank tensor algorithm |
| CN107818298A (en) * | 2017-09-27 | 2018-03-20 | 厦门大学 | General Raman spectral characteristics extracting method for machine learning material recognition |
| CN108918499A (en) * | 2018-06-28 | 2018-11-30 | 华南师范大学 | The method of Raman baseline drift is removed in Raman map |
Non-Patent Citations (2)
| Title |
|---|
| 柏顺陈樵，范文涛: "基于多尺度分解的双边滤波图像增强", 《重庆理工大学学报（自然科学）》 * |
| 王昕 等: "基于匹配追踪的拉曼光谱信号重构算法", 《光谱学与光谱分析》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021149760A1 (en) * | 2020-01-24 | 2021-07-29 | パナソニックＩｐマネジメント株式会社 | Raman spectroscopy method and raman spectroscopy suppport device |
| CN112161966A (en) * | 2020-09-29 | 2021-01-01 | 中国科学院长春光学精密机械与物理研究所 | Method and device for separating Raman spectrum of sample containing fluorescence spectrum |
| WO2023272749A1 (en) * | 2021-06-29 | 2023-01-05 | 江苏大学 | Feature information extraction method and portable detection system for small molecule volatile substance |
| CN113567604A (en) * | 2021-07-22 | 2021-10-29 | 华谱科仪(大连)科技有限公司 | Detection and analysis method of chromatographic spectrogram and electronic equipment |
| CN113640445A (en) * | 2021-08-11 | 2021-11-12 | 贵州中烟工业有限责任公司 | Characteristic peak identification method based on image processing, computing equipment and storage medium |
| CN113640445B (en) * | 2021-08-11 | 2024-06-11 | 贵州中烟工业有限责任公司 | Characteristic peak identification method based on image processing, computing device and storage medium |
| CN114113035A (en) * | 2021-11-18 | 2022-03-01 | 北京理工大学 | Transgenic soybean oil identification method |
| CN114113035B (en) * | 2021-11-18 | 2024-02-02 | 北京理工大学 | Identification method of transgenic soybean oil |
| CN114280002A (en) * | 2021-12-16 | 2022-04-05 | 宜宾五粮液股份有限公司 | Abnormal fermented grain spectrum screening method based on characteristic peak determination |
| CN115078616A (en) * | 2022-05-07 | 2022-09-20 | 天津国科医工科技发展有限公司 | Multi-window spectral peak identification method, device, medium and product based on signal-to-noise ratio |
| CN115078616B (en) * | 2022-05-07 | 2024-06-07 | 天津国科医疗科技发展有限公司 | Multi-window spectrum peak identification method, equipment, medium and product based on signal to noise ratio |
| CN115001571A (en) * | 2022-05-11 | 2022-09-02 | 青岛诺克通信技术有限公司 | Self-adaptive digital filtering method of optical time domain reflectometer and optical time domain reflectometer |
| CN115001571B (en) * | 2022-05-11 | 2023-09-12 | 青岛诺克通信技术有限公司 | Adaptive digital filtering method of optical time domain reflectometer and optical time domain reflectometer |
| CN115389481A (en) * | 2022-08-16 | 2022-11-25 | 华中科技大学 | Method for detecting uniformity of biomass surface coating based on Raman spectrum surface scanning |
| CN115389481B (en) * | 2022-08-16 | 2024-05-24 | 华中科技大学 | Method for detecting uniformity of biomass surface coating based on Raman spectrum surface scanning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109993155B (en) | 2021-01-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109993155A (en) | For the characteristic peak extracting method of low signal-to-noise ratio uv raman spectroscopy | |
| US11842483B2 (en) | Systems for cell shape estimation | |
| JP7273215B2 (en) | Automated assay evaluation and normalization for image processing | |
| CN110032988A (en) | Uv raman spectroscopy system real-time noise-reducing Enhancement Method | |
| Pilling et al. | High-throughput quantum cascade laser (QCL) spectral histopathology: a practical approach towards clinical translation | |
| CN108169213B (en) | Automatic identification method for peak elements of laser-induced breakdown spectroscopy | |
| CN111448569A (en) | Method for storing and retrieving digital pathology analysis results | |
| CN106815819B (en) | More strategy grain worm visible detection methods | |
| CN104990892B (en) | The spectrum picture Undamaged determination method for establishing model and seeds idenmtification method of seed | |
| Shu et al. | A semi-automatic image analysis tool for biomarker detection in immunohistochemistry analysis | |
| CN102706813B (en) | Poa pratensis variety identification method based on hyper-spectral image | |
| CN101692052A (en) | Hyperspectrum image technique-based method and hyperspectrum image technique-based device for identifying true and false famous tea | |
| AU2014235921A1 (en) | Method and system for analyzing biological specimens by spectral imaging | |
| AU2015372563A1 (en) | Vessel analysis in multiplexed images | |
| CN105891230A (en) | Fruit appearance detection method based on spectral image analysis | |
| CN113008805A (en) | Radix angelicae decoction piece quality prediction method based on hyperspectral imaging depth analysis | |
| Deepak et al. | Computer assisted pap smear analyser for cervical cancer screening using quantitative microscopy | |
| CN111523587A (en) | Woody plant species spectrum identification method based on machine learning | |
| CN113129281B (en) | Wheat stem section parameter detection method based on deep learning | |
| CN104076003B (en) | Extraction method of mineral spectrum absorption characteristic parameters | |
| JP6006036B2 (en) | Spectral spectrum analysis method | |
| JP2011145264A (en) | Biological specimen analyzer | |
| EP1155381B1 (en) | Minimally invasive apparatus and method for testing lesions of the oral cavity and similar epithelium | |
| CN100487395C (en) | Astronomical spectrum automatic sorting and red shift measuring method based on similarity measure | |
| Hao et al. | Wavelet-based threshold denoising for imaging hyperspectral data |
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 |