JP3095970B2 - Deconvolution processing method and apparatus - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a deconvolution processing method and apparatus.

[0002]

BACKGROUND OF THE INVENTION As is well known, when performing various spectroscopic measurements, light emitted passively or actively from a sample is spectrally separated using a spectroscope, and a predetermined light component is separated from the spectrometer. The light is received by a photodetector arranged on the emission side of the device, converted into a predetermined electric signal there, sent to a signal processing device, and analyzed.

The above-mentioned spectroscope and photodetector are generally used for a spectrum measuring method in which a single detector such as a photomultiplier is combined with a monochromator, or a multichannel detector such as a CCD or PDA and a polychromator. There is a multi-channel spectrum measurement method combined with a meter.

In the above-described spectroscope, slits are arranged on the entrance side and the exit side, and the spectral resolution is increased as the slit width becomes smaller within the limit of the diffraction grating to be used. However, since it is impossible to make the slit width zero, there is a limit to increasing the resolution inevitably, and the actual spectrum waveform is detected with a certain degree of dullness. . As an example, even if a spectrum exists only at a certain wavelength (fraction) such as a single trigger (bright line spectrum) as shown in FIG. 13A, the slit function of the slit used is In the case as shown in FIG. 7B, waveform data as shown in FIG. If the detector used at this time is a single-channel detector such as a photomultiplier, the obtained waveform data is continuous.
By performing an arithmetic process called deconvolution, the wavelength resolution can be improved and the original spectral data (FIG. 1A) can be obtained.

[0005] As described above, the resolution can be improved to some extent by reducing the slit width, and when higher resolution is required,
As the preprocessing, the above deconvolution processing may be performed.

In the case of a multi-channel detector, even if the slit width is reduced, the resolution does not exceed the resolution determined by the channel width on the detector side. Therefore, in order to improve the resolution, it is necessary to manufacture and prepare a detector having such a narrow channel width. However, this cannot be performed as easily as narrowing the slit width, and there is a limit.

Therefore, the present inventor tried to improve the resolution by arithmetic processing by deconvolution used in a single detector. However, the resolution cannot be improved by directly applying the above-described conventional deconvolution processing. This is for the following reason.

First, as for the effect obtained by the deconvolution, the more the number of data used for the calculation (the higher the data point density), the more accurately the original spectrum data can be reproduced. However, in the case of a multi-channel detector, the data obtained is discrete and the data point density is low. Therefore, in some cases, the number of data points for obtaining the slit function is several at the peak, and in the worst case FIG.
As shown in (A), there is a possibility that one point of the peak portion (indicated by a black circle) may be obtained. Then, the function that can be expressed by this is a crisp function as shown by the two-dot chain line in FIG.
Even when deconvolution processing is performed, no improvement in resolution is seen at all.

Further, even if the spectrum data emitted from the slit is as shown by a solid line in FIG. 1B, the data detected by the detector is discrete as shown by a black circle in FIG. In addition, the sampling interval is widened. Therefore, the detected spectrum data that can be reproduced from the obtained data (black circles) is, for example, as shown by a two-dot line in the figure, and may be far from the actual data. If the data itself contains a large error, it is almost impossible to reproduce data close to the actual spectrum data even after deconvolution.

Although the deconvolution is based on the premise that the sampling intervals are equal, the data detected in each channel of the multi-channel detector is physically equal, but the Since the wave number intervals between the channels are different for each channel, the actually obtained spectrum data is sampled at irregular intervals with respect to the wave numbers. Therefore, since the above assumptions are different, accurate deconvolution cannot be performed in that respect as well.

For the above various reasons, deconvolution has been performed only for a single detector so far, and has not been applied to a multi-channel detector for many years. The fact was not mentioned.

[0012] The present invention has been made in view of the above background, and an object of the present invention is to break down the conventional fixed concept described above and to perform deconvolution processing even in spectrometry using a multi-channel detector. Provide a deconvolution processing method and apparatus that can improve the resolution, enable accurate analysis with a small spectroscope, and exhibit the same effect on spectra other than spectrum or images in space. Is to do.

[0013]

In order to achieve the above-mentioned object, in the deconvolution processing method according to the present invention, a slit function is separately measured for measurement spectrum data obtained from a multi-channel detector, and the slit function is measured. When performing the deconvolution processing using the above, the following procedure was used. (1) First, a plurality of data are measured a plurality of times while shifting the center wave number of the spectrum data detected by the multi-channel detector, and a large number of obtained data are rearranged in the order of wave numbers to form a data sequence having an improved data point density. Generate. (2) Next, the generated data sequence is divided into a plurality of small sections, and a polynomial fitting is performed for each of the small sections to obtain a function type, thereby obtaining a desired data point density and equal wave number intervals. By substituting such a wave number into the function type, the step of obtaining a pseudo data sequence is performed on the measurement spectrum based on the sample and the spectrum for obtaining the slit function, respectively. Extract each pseudo data string for the function. (3) Perform deconvolution based on each of the extracted pseudo data strings.

At this time, preferably, the sampling interval for extracting the pseudo data sequence for the slit function is set wider than the sampling interval used for extracting the pseudo data sequence for the target spectrum data. It is.

The application of the present invention is not limited to the spectrum described above. In other words, the processing target can be replaced with the spectral data to be an aerial image. In this case, a point light source such as scattering of a bead sphere is used instead of the slit function.

A deconvolution processing apparatus according to the present invention suitable for carrying out the above method is a deconvolution processing apparatus mounted on a spectrometer having a spectroscope with a multi-channel detector. Movement control means for moving at least one of the diffraction grating in the spectroscope and the multi-channel detector by a predetermined amount, and issues a control command to the movement control means to obtain spectral data detected by the multi-channel detector. Data interpolating means for changing the center wave number by a predetermined amount, receiving an output from the multi-channel detector, and generating a data sequence having a data point density equal to or less than the channel width of the multi-channel detector; and Based on the generated data sequence, the desired Means for generating a pseudo data sequence having a data point density and having equal wave number intervals, and a means for performing deconvolution based on the pseudo data sequence generated by the uniform data generation unit. And the equal-interval data generating means is configured as follows.

That is, the equally-spaced data generating means includes:
A given data sequence is divided into small sections, and a polynomial fitting is performed based on the data present in the small sections to generate a spectrum for each of the small sections. The deconvolution data extracting means for extracting each data point constituting the pseudo data sequence using the information on the generated polynomial-fitted function type that defines the spectrum.

In this case, the sampling interval at the time of data extraction by the deconvolution data extracting means is variable, and the sampling interval for the slit function is set to be smaller for the apparent slit function in order to reduce the apparent width of the slit function. Preferably, it can be made wider than the sampling interval.

Here, the "wave number" referred to in the present invention means not only a generally used wave number but also an equivalent (convertible).
This concept includes various wavelengths. In addition, the process of dividing into small sections to be performed as preprocessing when extracting a pseudo data string may be performed so that each small section does not overlap, or such that adjacent small sections partially overlap. The present invention includes both cases.

[0020]

When a multi-channel detector is used, its resolution depends on the channel width. Therefore, the resolution increases as the detector width decreases,
In practice, it is difficult and expensive to manufacture such a detector. Therefore, first, measurement is performed a plurality of times (n times) while shifting the center wave number of the spectrum data detected by the detector. As a result, data of n times the number of data obtained by one measurement is obtained, so that the data point density per channel is also improved by n times. In addition, the interval at which the wave number is shifted may be set without concern for the channel width. Then, assuming that the shift amount is Δν and the channel width is d, d / Δν is not an integer. That is, there are many cases where one channel cannot be divided at equal intervals. There is no problem, and the point is that the number of data required to reproduce characteristics close to the original data can be obtained.
It should be noted that each data constituting the data sequence obtained by this processing is data based on actual measurement.

The data string having a high data point density extracted by the above processing is divided into small sections, polynomial fitting is performed for each small section, and data emitted from a spectroscope or the like and incident on a detector is obtained. Reproduce. Since the data is divided into small sections, the original data can be accurately reproduced with a relatively low-dimensional polynomial. At this time, in practice, it is better to express the function type in consideration of the post-processing.

In order to perform deconvolution with high accuracy, (a) the data point density must be higher than a certain level, and (b) the data points sampled must be equally spaced (for example, in wave number expression). is necessary. Therefore, the wave numbers satisfying the above conditions (a) and (b) are extracted from the spectrum reproduced as described above (expressed in functional form for each small section), and the above-mentioned reproduced spectrum at the wave number is extracted. Find the data points of. Specifically, it can be easily obtained by substituting the wave number into a function sequence. Each data obtained in this way is
This is pseudo data obtained by calculation based on a spectrum generated by polynomial fitting. Therefore, in the present specification, a data string composed of this data is called a pseudo data string.

The above processing is performed on each of a measured spectrum based on an actual sample and a spectrum for obtaining a slit function, and respective pseudo data strings are obtained, and deconvolution is performed using each pseudo data string. Original data (without dulling due to slits or the like) can be reproduced.

[0024]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A deconvolution processing method and apparatus according to the present invention will be described below in detail with reference to the accompanying drawings.
First, the basic concept of the present invention will be described. (1) First, while moving the relative position of the spectroscope and the multi-channel detector, data is acquired each time, and the wave number detected in each channel of the multi-channel detector is shifted. Then, the spectrum data detected each time is rearranged into one spectrum according to the wave number, thereby improving the data point density. With this, if it is moved n times,
Since the data point density is also improved by n times, there is a high possibility that the low data point density, which is one of the problems of the multi-channel detector, can be eliminated.

(2) On the other hand, even if the data point density is improved as described above, there is a possibility that the data point density required for performing deconvolution accurately cannot be obtained. Even if is obtained, it is not sampled at equal intervals with respect to the wave number, so even if deconvolution is performed using the detected data as it is, it will still accurately reproduce the original (before slit passage) data I can't.

Therefore, a large number of data obtained as described above are divided into fixed sections and polynomial fitting is performed for each section, so that a curve passing over or near data existing in a short section is obtained. Is obtained and one continuous data is generated by connecting a function (curve) based on a polynomial formed for each section.

As described above, by performing processing only for a relatively short section, a curve approximating the characteristic (curve) of the spectrum data emitted from the spectroscope can be reproduced. Consecutive data can also be obtained that approximates the original data after passing through the slit.

Then, the reproduced approximate data is sampled at equal intervals based on the wave number, and data having a data point density required for deconvolution is obtained. Note that this data acquisition is performed for both the data to be measured and the calculation of the slit function.

(3) On the basis of the data extracted by the above process, a single detector performs a normal deconvolution process conventionally performed. Then, by the processing up to the above (2), the data point density can be reduced even if the channel width of the multi-channel detector is not so narrow,
The spectrum data emitted from the slit can be accurately reproduced. In addition, since deconvolution is performed based on the data sampled at the required data point density at equal wave number intervals, the spectrum data before passing through the slit can be accurately reproduced, and the wave number resolution can be improved. Therefore, high-performance spectroscopy and analysis can be performed by performing the subsequent predetermined processing.

Next, an example of a spectrometer including a specific processing device for performing the above processing will be described. FIG.
Shows an example in which this apparatus is applied to a Raman scattering measurement apparatus. As shown in FIG. 1, a light source (for example, an Ar laser) 1
Is transmitted through the half mirror 2,
The light is converged on the sample cell 4 by the condenser lens 3. In the direction orthogonal to the direction of irradiation on the sample cell 4, an imaging lens 5 and a polychromator 6 as a spectroscope are arranged in this order.
The Raman scattered light emitted from the sample in the sample cell 4 is imaged by the imaging lens 5 through the entrance slit 6 a in the polychromator 6.

As is well known, a diffraction grating and other optical components are arranged in a predetermined positional relationship in the polychromator 6, and by setting the diffraction grating to a predetermined rotation angle position, the output side of the polychromator 6 Outputs a spectrum of a predetermined wave number component, and C is set at the exit focal plane.
Light is received by each channel of the multi-channel detector 7 composed of a CD or the like.

The multi-channel detector 7 converts the received optical signal into an electric signal, and transfers the detected data to the computer system 9 via the multi-channel detector driver 8 connected to the output side. The computer system 9 performs predetermined arithmetic processing such as deconvolution which is a main part of the present invention.

Here, in the present invention, the diffraction grating in the polychromator 6 is connected to rotation driving means such as a stepping motor so as to be rotatable, and spectral data of a desired wave number is received by each channel of the multi-channel detector 7. So that it can be done. The rotation control of the diffraction grating is performed by rotating the stepping motor by a predetermined angle in a predetermined direction based on a control signal from the wavelength scanning driver 10.

In this embodiment, the diffraction grating is rotated. However, the present invention is not limited to this, and the multi-channel detector 7 may be moved in parallel.
As a mechanism for moving the motor, a means for converting the direction of motion from rotation to linear reciprocation may be connected to the output of the motor, or a mechanism using a piezoelectric element may be used. . Note that such a piezoelectric element can also be applied to the power for rotating the diffraction grating described above.

Further, a sub-light source 11 for emitting a bright line spectrum such as a Ne lamp is provided for the half mirror 2,
The light (bright line spectrum) emitted from the sub light source 11 is reflected by the half mirror 2 so that the sample cell 4 can be irradiated via the condenser lens 3. This is a light source used for obtaining the slit function. When obtaining the slit function, an appropriate scattering medium is placed in the sample cell 4.

Note that a scattering plate may be provided instead of the sample cell 4 to which the scattering medium is supplied. Further, a normal mirror may be installed instead of the half mirror 2. However, in that case, the mirror is made movable in a predetermined direction, and during normal measurement, the mirror is moved from the state shown in the figure to collect the light emitted from the light source 1 as it is (without being disturbed by the mirror). When the light is condensed on the sample cell 4 via the lens 3 and the slit function is obtained, it is necessary to return to the state shown in the figure so that the light emitted from the sub light source 11 can be guided to the sample cell 4.

Next, an internal configuration of the computer system 9 which is a main part of the present invention will be described. As shown in FIG. 2, on the input side, the data interpolation unit 20 (the above-described processing (1))
And the data interpolating unit 20 increases the data point density from the density based on the channel width of the detector, and determines the wave number required to perform deconvolution based on a large number of interpolated data. The equally-spaced data generating unit 21 for extracting the equally-spaced data (the part for executing the process (2) described above) and the executing unit 22 for executing the deconvolution based on the data output from the generating unit 21 (as described above) (A part for executing the process (3)).

Although not shown, the result of the deconvolution is transmitted to an arithmetic processing unit for a predetermined analysis (not shown), and the arithmetic processing unit is also mounted in the computer system 9.

The data interpolator 20 acquires the spectrum data a plurality of times while changing the center wave number of the spectrum data detected by the multi-channel detector 7, and obtains the spectrum data (channel width) obtained by one detection. Is detected), another spectral data present in
The data point density is improved by performing the interpolation processing, and specific processing functions are as shown in FIG.

Before describing the specific processing, the operation principle of the data interpolation unit 20 will be described. For example, as shown in FIG. 4, when the channel width of the multi-channel detector is d (cm ⁻¹ ), the width d is obtained in one measurement.
It is not possible to obtain a resolution higher than (cm ^-1 ). Therefore, as described above, the center wave number of each channel is set to Δν (cm ⁻¹ ).
By measuring a plurality of times (n times) while shifting each other and rearranging them based on the wave number, the channel width d
A plurality of measurement data can be present in (cm ^-1 ), thereby improving the data point density. And 1
In the case of performing data interpolation in one channel, generally, the inside of the channel is divided into n at equal intervals, and data (boundary or center, etc.) representing each of the divided sections is acquired (example shown in the figure). Then it is divided into three).

That is, it is necessary that d = n × Δν. However, n needs to be an integer because it corresponds to the number of measurements, and furthermore, the wave number is shifted by rotating the diffraction grating by a predetermined angle. Δν
There is also a limit on the value that can be taken as (cm ⁻¹ ). Also, since d itself differs for each channel, the condition of the equation is satisfied only for a specific channel.

Therefore, the process of finding each value satisfying the above conditions is not only complicated, but also it is not always possible to obtain the required number of data points, and it is difficult to accurately reproduce the original spectrum. Become.

Therefore, in the present invention, when determining the shift amount (wave number inching width) Δν ′ (cm ⁻¹ ) of the wave number at the time of interpolation, the above-mentioned condition (d is an integer multiple of Δν ′) ), And the desired number (n) in one channel
And the data point density per channel is improved by n times. That is, instead of the above-mentioned conditional expression, when it is not an integer multiple, an arbitrary value n satisfying the following conditional expression is adopted.

N ≧ [d / Δν ′] + 1 where the operation [a] is a maximum integer not exceeding a.
, The data obtained from the first time to the [d / Δν ′] time exist in the same channel, and [d / Δν ′]
In the measurement from the Δν ′] + 1-th measurement, the data protrudes to the right channel (k + 1), and at the same time, the data protruding from the left (k−1) side enters the channel (k). As a result, n data exists at irregular intervals in one channel as a result.

As a result, as shown in FIG. 5 (for the case of n = 6), the intervals between adjacent measurement data in each time are equally spaced by a distance equal to the channel width d. Δν ^-1 between the second measurement data and
Only the wave number is increasing. Then, each wave number obtained by sequentially increasing the wave number of channel (k) from n = 1 is n
Exist in the same channel from 1 to 4, but n =
5 and 6 respectively protrude to the right channel (k + 1). However, measurement data obtained by increasing the wave number starting from the channel (k-1) adjacent to the left enters the same channel (k) as described above. Is n (6)
Data exists.

The data interpolator 20, which performs a predetermined process based on the above-described operation principle, stores the spectral data acquired each time in the buffer 20a, and if n times of data are acquired, temporarily stores the data. To buffer 2
The data is read from 0a, and the read data are rearranged in the order of wave numbers to generate one spectrum data. As shown in FIG. 3, n is set to 1 first (ST).
1). Further, as an initial setting, the driver 10 is operated to set the diffraction grating in the polychromator 6 to the initial position.

Next, the spectrum data detected by the multi-channel detector 7 is obtained and stored in the buffer 20a as the n-th data (ST2). Then, it is determined whether or not n has reached a preset value (max) (ST).
3) If it is less than max, n is incremented, and the driver 10 is operated to make the center wave number of the multi-channel detector 7 smaller than the (n-1) -th center wave number by Δ.
The diffraction grating is rotated by a predetermined angle so as to increase by ν ′ ⁻¹ (ST5). In the process of step 5, the rotation angle (reference angle) is actually obtained in advance and set in the driver 10. Then, a rotation command signal (trigger pulse) is output from the data interpolation unit 20, and the driver 10 receiving the command signal operates so as to rotate by the reference angle.

When the spectral data has been acquired a predetermined number of times, the data for each of the n wave numbers stored in the buffer 20a is read out, and rearranged in the order of the wave numbers. Is generated (ST6). The physical distance d of each channel of the multi-channel detector at each time is known,
Since the value of the number of waves received in each channel can be calculated from the diffraction grating or the like to be used, and the center wave number of the n-th data is a value obtained by adding n × Δν ′ ⁻¹ to each wave number, the buffer 20a The wave number of each data stored in is obtained by a simple calculation. Therefore, rearrangement is performed according to the wave number.

The equally-spaced data generating section 21 receives a large number of data constituting one spectrum given from the data interpolating section 20 and, based on the data, receives the original data to be received by the multi-channel detector 7 after passing through the slit. A spectrum generation unit 21a for generating a spectrum and, based on the spectrum reproduced (generated) by the spectrum generation unit 21a, the data point density condition is satisfied and sampling is performed so that the wave numbers are equally spaced, and deconvolution is performed. And a deconvolution data extracting unit 21b for extracting data for use.

The function of the equally-spaced data generator 21 is as shown in FIG. First, the spectrum data provided from the data interpolation unit 20 is divided into small sections at equal intervals based on the wave number (ST11).
That is, since the entire spectrum range of the spectrum data acquired by the data interpolation unit 30 is known, the spectrum is divided so that the wave number intervals become equal. The number of divisions is, for example, about the number of channels (number of elements) of the multi-channel detector 7 to be used (in the present invention, it may be larger or smaller than the number of channels). As described above, since the physical step width of each channel and the increase in the wave number of the spectrum data detected there are not proportional, when dividing based on the wave number, the number of data existing by the divided small section is reduced. There is a difference but no problem.

Next, a polynomial fit is performed for each of the divided small sections to obtain a function type (ST12). In other words, all the data points existing in the small section are extracted in order from the leading small section, and a polynomial fit is performed, and a curve that passes through each data point as much as possible (actually, the corresponding curve is defined Function type). Also, at this time, in order to ensure continuity for each adjacent small section, polynomial fitting is performed in a state where several neighboring points existing in both adjacent small sections are added. When performing the polynomial fit, the degree of the polynomial is set so as not to exceed n, which is the average number of data points in a small section. The above steps 11 and 12 are the processing of the spectrum generation unit 21a.

When a spectrum represented by a continuous curve (some of which is partially discontinuous at the boundary of a small section) is generated in this manner, it is converted into a deconvolution data extraction unit 21b at the next stage. To extract necessary spectrum data (ST13). That is, since the function type is determined over the entire spectrum range by the processing up to step 12, the sampling interval is set again over the entire wave number section so that the finally required data point density can be obtained. The data value is recalculated at regular intervals based on the wave number. Specifically, when the wave number to be extracted is determined, a small section in which the wave number exists is detected, and the wave number is substituted into a function type that defines the small section to obtain a data point at the wave number.

When the wave number to be obtained exists over a plurality of small sections (for example, at the boundary of the small section), each data point is obtained by using the function type of each small section, and the data points are averaged. The data point of the wave number (ST
14).

The above steps 13 and 14 are the processing of the deconvolution data extraction unit 21b. And
As a result, a predetermined number of data points (spectral data) obtained by sampling the wave numbers at equal intervals are extracted, and the extraction result is sent to the deconvolution execution unit 22 in the next stage.

In the above-described embodiment, when the data is divided into small sections by the processing of the spectrum generating section 21a, the divisions are made at equal intervals without overlapping each other. The fixed regions (for example, half each) of the small section may overlap each other. In such a case, the probability that the wavelength to be extracted exists over a plurality of small sections is high.
The problem is dealt with by executing the processing of No. 4.

The deconvolution execution unit 22 performs a predetermined arithmetic processing based on the given spectral data (both based on the measurement sample and the slit function).
Reproduces the original (not affected by the slit) spectrum. In this deconvolution process, for example, the least squares estimation solution is solved by the Gauss-Seidel method, and the step width for each iteration is determined by optimizing the steepest descent method.

As described above, in this example, even if a multi-channel detector is used, spectral data sampled at a required data point density and at equal intervals by the wave number is generated, so that the above-described deconvolution is executed. For each specific processing for performing the processing, the same processing as used in a normal single detector can be applied as it is, and a description thereof will be omitted.

Next, an embodiment of the method according to the present invention will be described with reference to the above embodiment. First, a sample to be measured is supplied into the sample cell 4, and light emitted from the light source 1 is applied to the sample cell 4 (sample in the cell) (ST21,
22). Then, Raman scattered light is generated from the sample, and a part of the Raman scattered light is acquired by the multi-channel detector 7 via the spectroscope (polychromator) 6 to obtain spectral data of each wave number. Then, the computer system 9 is operated to acquire a large number of data points while rotating the diffraction grating, and based on the acquired data points, to obtain a measurement spectrum (a discrete spectrum having a desired data point density) obtained by sampling at equal intervals based on the wave number. (Consisting of typical data points) (ST23).

On the other hand, a scattering substance is supplied into the sample cell 4 and the light emitted from the sub light source 11 is irradiated on the sample cell 4 (scattering substance in the cell) by switching the optical system.
The auxiliary light source 11 used at this time selects and uses the bright line of the Ne lamp close to the center wave number of the sample spectrum (S
T24, 25). Then, Raman scattered light is generated from the scattered substance, and a part of the Raman scattered light is acquired by a multi-channel detector 7 via a spectroscope (polychromator) 6 to obtain spectral data of each wave number. Then, the computer system 9 is operated to acquire a large number of data points while rotating the diffraction grating, and based on the acquired data points, to obtain a spectrum (a discrete number of data points having a desired data point density) obtained by sampling at equal intervals based on the wave number. ), That is, a slit function is obtained (ST26).

In order to minimize the influence of spatial sensitivity unevenness of the multi-channel detector 7, the spectra of the sample and the slit function in steps 23 and 26 are measured at the center of the detector.

Next, spectrum data based on the measured spectra obtained in steps 23 and 26, respectively,
The slit function is provided to the deconvolution execution unit 22, where the deconvolution is performed, and the measurement spectrum blunted by the slit is returned to the original state to improve the wavelength resolution (ST27). Step 21
23 to 23 and steps 24 to 26 may be performed first.

Next, the following experiment was conducted in order to verify the effects of the above-described embodiment. Raman spectrometer with multi-channel detector (1800 diffraction gratings / mm, NR-1
Using 800 (manufactured by JASCO Corporation), carbon tetrachloride was measured under the conditions of an entrance slit width of 100 μm, a slit height of 2 mm, an integration time of 30 seconds in a single monochromator mode, and an integration of 4 times. At this time, the excitation wavelength is Ar ⁺
Ion laser (514.5 nm.

As a result, a measurement result as shown in FIG. 8A was obtained. Note that, in the figure, the Raman peak 460c
around m ^-1 (437.5 cm ^{-1 to} 48 including the above peaks)
2.5 cm ^-1 wave number range), and the number of data points is 6.
6 points. The width of one channel of the detector is 24μ.
In m, since the reverse line dispersion of the spectrometer at a wave number position of the Raman shift 460 cm ^-1 is a 28.8cm ^-1, a 0.69 cm ^-1 wavenumber width per channel in the central portion of the detector . Further, the spectral bandwidth becomes 2.88 cm ^-1 .

Further, as a result of measuring the emission line spectrum of Ne near 675 cm ^-1 by using the same apparatus, the result shown in FIG. This is the slit function.
In each of the figures, the data points (indicated by black circles) that are measured are drawn with lines connecting them for easy viewing (the same applies hereinafter).

Then, 20 (n = 20) spectrum measurements are performed while shifting the center wave number of the spectrum shown in FIG. 4A to the higher wave number side in steps of 0.1 cm ⁻¹ , and these are measured by the data interpolation unit 20. Rearranged into one. Thus, a result as shown in FIG. 9A was obtained. As a result, the total number of data points is 1320 (66 × 20) points,
Data point density improved. The black circles in the enlarged portion in the figure indicate sampling points, but some disturbances occur due to the limitation of mechanical accuracy of the wave number sweep of the spectroscope.

The spectrum shown in FIG. 9A is divided into small sections (wave number width: 1.0 cm ⁻¹ ), and each small section is overlapped by 0.5 cm ⁻¹ before and after each other to obtain a fifth-order polynomial. Fitting was performed, and sampling was performed again at regular intervals of 0.1 cm ^-1 over the entire wave number section. Thereby, a result as shown in FIG. 10A was obtained. Therefore, the number of data points here is 451.

By performing the same processing as above on the slit function (FIG. 8B),
(B), the result as shown in FIG. 10 (B) was obtained. Finally, the data point density of FIG. 10 is about seven times that of FIG.

Then, each data (pseudo data sequence) indicated by a black circle in FIGS. 10A and 10B is given to the deconvolution execution unit 22, and the spectrum of the equally spaced and high data point density (FIG. 10A )) To the slit function (FIG. 10)
(B)). Then Figure 11
The result as shown in (A) was obtained. Here, the number of repetitions was fixed at 50 times. Although the generation of a pseudo peak P in the negative direction is observed, the peak C ³⁵ of three isotopes of carbon tetrachloride is observed.
Cl ₂ ³⁷ Cl ₂ (455.1 cm ⁻¹ ), C ³⁵ Cl ₃ ³⁷ C
1 (458.4 cm ^-1 ), C ³⁵ Cl ₄ (461.5 cm
^-1 ) is well separated.

On the other hand, for comparison, deconvolution was performed based on the data shown in FIGS. 8A and 8B (other processing conditions were the same). As a result, FIG.
(B) shows that the product of the present invention has improved resolution.

Looking at the final result obtained in FIG. 11A, a pseudo peak P in the negative direction is generated. Moreover, this negative peak was not significantly improved by reducing the number of deconvolution iterations. This is presumably because the identity of the optical system between the measurement of the Ne emission line and the measurement of the sample was not maintained, and the line width of the slit function became wider than it actually was, resulting in an over-deconvolution state.

Therefore, the equally-spaced data generator 2 shown in FIG.
1 can be dealt with by modifying the function of the data extraction unit 21b (an apparatus having this modified function is a modified embodiment of the present invention). That is, the spectrum generation unit 21
The shape of the slit function is calculated by a polynomial fitting at a, and sampling is performed at predetermined regular intervals by the data extraction unit 21b. When recalculating the regular sampling data at the final stage, the sampling interval of the sample spectrum data is calculated. The sampling of the slit function is performed slightly wider than the sampling.

With such a configuration, the shape of the spectrum does not change and the slit function can be apparently narrowed. Therefore, if the two data obtained through the above processing are given to the deconvolution execution unit 22 and the deconvolution is executed, good results can be expected.

In order to demonstrate the effect of this modified example, deconvolution was performed on the sample spectrum of FIG. 10A with a slit function sampled at an interval of 1.4 times the sampling interval. Then, a result as shown in FIG. 12 is obtained, and it can be seen that a large negative peak in the negative direction is considerably improved as is apparent from the figure.

In each of the above embodiments, an example in which the present invention is applied to a Raman scattering measuring apparatus has been described. However, the present invention is not limited to this, and can be applied to various spectroscopic measuring apparatuses using a multi-channel detector. Of course. Furthermore, the target of deconvolution processing is not limited to spectral data, but may be image data. However, in this case, an image from a point light source such as scattering of a bead sphere is used as a device function instead of a slit function.

[0075]

As described above, in the deconvolution processing method and apparatus according to the present invention, even if the data point density that can be acquired at one time is low, the data point density can be reduced by acquiring data a plurality of times. Can be higher. And finally, since deconvolution is performed based on a pseudo data sequence obtained by extracting data matching desired conditions from a spectrum generated based on a large number of data points thus obtained by calculation, The deconvolution processing can be performed with high accuracy, and the processing for increasing the data point density can be performed without much consideration of the channel width of the detector, so that the required number of data can be obtained reliably and easily. Since the deconvolution can be performed with high accuracy, the wave number resolution is improved, so that a device such as a spectroscope to be used as a result can be reduced in size. Becomes).

[Brief description of the drawings]

FIG. 1 is a diagram showing an example of a deconvolution processing device according to the present invention and a spectrometer in which the deconvolution processing device is mounted.

FIG. 2 is a block diagram showing an embodiment of a deconvolution processing apparatus.

FIG. 3 is a flowchart illustrating a processing function of a data interpolation unit.

FIG. 4 is a diagram illustrating an operation of a data interpolation unit.

FIG. 5 is a diagram illustrating an operation of a data interpolation unit.

FIG. 6 is a flowchart illustrating a processing function of an equally-spaced data processing unit.

FIG. 7 is a flowchart illustrating an example of a deconvolution processing method according to the present invention.

FIG. 8 is a diagram showing the results of an experiment performed to demonstrate the effects of the present invention.

FIG. 9 is a diagram showing the results of an experiment performed to demonstrate the effects of the present invention.

FIG. 10 is a diagram showing the results of an experiment performed to demonstrate the effects of the present invention.

FIG. 11 is a diagram showing the results of an experiment performed to demonstrate the effects of the present invention.

FIG. 12 is a diagram showing the results of an experiment performed to demonstrate the effects of the present invention.

FIG. 13 is a diagram illustrating deconvolution.

FIG. 14 is a diagram illustrating a conventional problem.

[Explanation of symbols]

Reference Signs List 6 polychromator (spectrometer) 6a entrance slit 7 multi-channel detector 9 computer system (deconvolution processing device) 10 driver for diffraction grating (movement control means) 20 data interpolation unit 21 evenly spaced data processing unit 21a spectrum generation unit 21b decon Revolution data extraction unit 22 Deconvolution execution unit

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-3-226632 (JP, A) JP-A-1-197617 (JP, A) JP-A-1-259226 (JP, A) JP-A-59-1987 67429 (JP, A) JP-A-6-331440 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G01J 3/02-3/52 G01N 21/27 G06F 17/10

Claims (4)

(57) [Claims] 1. When a slit function is separately measured with respect to measurement spectrum data obtained from a multi-channel detector and deconvolution processing is performed using the same, a center wave number of the spectrum data detected by the multi-channel detector is used. Multiple times while shifting the data, a large number of acquired data are rearranged in the order of wave numbers to generate a data sequence with an improved data point density, and the generated data sequence is divided into a plurality of small sections. A polynomial fitting is performed for each of the small sections to obtain a function type, and a pseudo data sequence is obtained by substituting a wave number having a desired data point density and an equal wave number interval into the function type. , For the measured spectrum based on the sample and the spectrum for obtaining the slit function Pseudo data string for random data string and slit function for measuring spectra respectively extracted, then deconvolution processing method to perform deconvolution using the pseudo data sequence the extraction. 2. The apparatus according to claim 1, wherein a sampling interval for extracting the pseudo data sequence for the slit function is set wider than a sampling interval used for extracting a pseudo data sequence for target spectrum data. The deconvolution processing method described. 3. A deconvolution processing device mounted on a spectrometer having a spectroscope with a multi-channel detector, wherein at least one of a diffraction grating in the spectrometer and the multi-channel detector is a predetermined amount. A movement control means for moving the multi-channel detector at a time, a control command is issued to the movement control means, and the center wave number of the spectrum data detected by the multi-channel detector is changed by a predetermined amount to output the multi-channel detector. Receiving, a data interpolating means for generating a data string having a data point density equal to or less than the channel width of the multi-channel detector, based on the data string generated by the data interpolating means,
Equal interval data generating means having a desired data point density required for performing deconvolution, and generating a pseudo data sequence having equal wave number intervals, based on the pseudo data sequence generated by the equal interval data generating device Means for performing deconvolution by dividing the given data sequence into small sections, and performing polynomial fitting based on data existing in the small sections. Using a spectrum generating means for generating a spectrum for each of the small sections, and information on a function type subjected to a polynomial fitting that defines the spectrum generated by the spectrum generating means. Deconvolution data extraction means for extracting Convolution processing apparatus. 4. The sampling interval for the slit function is changed by changing the sampling interval at the time of data extraction by the deconvolution data extracting means to reduce the apparent width of the slit function. 4. The deconvolution processing device according to claim 3, wherein the interval can be wider than the interval.