Method for improving recognition accuracy of spectrum peak of laser probe by using form fuzzy judgment

Abstract

The invention belongs to the related technical field of laser probe element analysis, and discloses a method for improving the recognition accuracy of a laser probe spectrum peak through morphological fuzzy judgment, and the method comprises the following steps: S1, collecting an LIBS spectrum to obtain plasma spectrum data; s2, setting the length of a spectrum band, dividing the spectrum band, and calculating the spectrum intensity difference between all adjacent points on two sides of the central wavelength in all the spectrum bands (before and after) to obtain an intensity difference value matrix on the left side and the right side; s3, judging whether all elements in the strength difference matrixes on the left side (right side) are larger (smaller) than zero or not, and respectively obtaining difference state bit matrixes on the left side and the right side; s4, a fuzzy criterion is established, the sum of the difference state bits on the left side and the right side is solved, symmetric spectrum peaks and asymmetric spectrum peaks are recognized through the fuzzy criterion, and then single-burr spectrum peaks, second-burr spectrum peaks and multi-burr spectrum peaks in the asymmetric spectrum peaks are further recognized; and S5, outputting the wavelengths and the spectral intensities of all the identified spectral peaks.

Description

Technical Field

[0001] The present invention belongs to the technical field related to laser probe spectrum analysis, and more specifically, relates to a method for improving the accuracy of laser probe spectrum peak recognition by utilizing morphological fuzzy judgment. Background Art

[0002] Laser probe technology is laser-induced breakdown spectroscopy (LIBS), or LIBS for short. It is a modern detection technology that uses high-energy-density lasers to ablate the sample surface and utilizes the plasma emission spectrum generated to achieve elemental analysis.

[0003] When using laser probe technology for elemental analysis, it is generally necessary to first identify the peaks of the sample spectrum, then select a specific number of peak wavelengths from the identified peaks as analysis lines, and then extract the spectral intensity characteristics of the selected analysis lines from the original data, and establish an analysis model to achieve qualitative or quantitative analysis. There are two traditional peak-finding methods: manual and automatic. The former has high accuracy, but requires manual inspection and selection one by one, and the efficiency is relatively low; while the latter can use a specific algorithm to achieve automatic identification of all peaks in the spectrum, with high identification efficiency, and can generally be used for qualitative analysis in laser probe technology and analysis line selection in multivariate regression quantitative analysis. Traditional automatic peak identification methods include wavelet transform method, second-order derivative method, and image feature line selection method. For example, in 2014, Jilin University published a paper on peak-finding based on continuous wavelet transform, namely "A LIBS spectrum automatic peak-finding method based on continuous wavelet transform" (Spectroscopy and Spectral Analysis. Vol. 34, No. 7, pp1969-1972). The study published the identification results of a total of 300 pixels in the spectral range of 382.2747 to 385.3477 nm in the Fe3O4 sample, but the accuracy and quality of its spectral peak identification (such as its spectral peak intensity and signal-to-noise ratio) still need to be further improved.

[0004] Based on this, the present invention aims to provide a method for further improving the accuracy of identifying various types of spectral peaks in laser probe spectroscopy according to the spectral peak morphology. Summary of the invention

[0005] In view of the shortcomings of the existing laser probe spectral analysis technology, the problem to be solved by the present invention is to achieve accurate and rapid identification of various spectral peaks through fuzzy judgment of spectral peak morphology, thereby improving the quality of the analysis line in the laser probe technology.

[0006] To achieve the above object, the present invention provides a method for improving the accuracy of laser probe spectrum peak recognition by using morphological fuzzy judgment, the method comprising the following steps:

[0007] S1, perform LIBS spectrum acquisition to obtain plasma spectrum data;

[0008] S2, set the length of the spectrum segment, divide the spectrum segment, and calculate the spectral intensity difference between all adjacent points on both sides of the central wavelength in all spectrum segments, and obtain the intensity difference matrix on the left and right sides of all spectrum segments;

[0009] S3, using the intensity difference matrices on the left and right sides of all spectrum segments to obtain the difference state bit matrices on the left and right sides of all spectrum segments respectively;

[0010] S4, summing up all elements in the state bit matrix on the left and right sides of each spectrum segment to obtain the sum of the state bits of the corresponding spectrum segment, and the sum of the state bits of all spectrum segments constitutes the state bit sum matrix of the entire spectrum; establishing a fuzzy criterion, first using the fuzzy criterion to identify the symmetrical spectrum peak and the asymmetrical spectrum peak, and then further identifying the single spur spectrum peak, the double spur spectrum peak and the multi-spur spectrum peak in the asymmetrical spectrum peak;

[0011] S5, output the wavelengths and spectral intensities of all identified spectral peaks.

[0012] Furthermore, step S2 specifically includes the following sub-steps:

[0013] S21, taking all wavelengths of the full spectrum as the central wavelength, setting the spectrum length, and specifying the spectrum range for each central wavelength;

[0014] S22, calculating the difference of all adjacent points on the left side of the central wavelength in the spectrum segment, subtracting the spectrum intensity of the first point on the left side from the spectrum intensity of the central wavelength to obtain the first intensity difference on the left side, then subtracting the spectrum intensity of the second point on the left side from the spectrum intensity of the first point on the left side to obtain the second intensity difference on the left side, and so on until the left boundary of the spectrum segment, and storing all the obtained intensity differences on the left side into the left difference matrix;

[0015] Calculate the difference of all adjacent points on the right side of the central wavelength in the spectrum band, subtract the spectrum intensity of the central wavelength from the spectrum intensity of the first point on the right side of the central wavelength to obtain the first intensity difference on the right side, then subtract the spectrum intensity of the first point on the right side from the spectrum intensity of the second point on the right side to obtain the second intensity difference on the right side, and so on until the right boundary of the spectrum band, and then store all the intensity differences on the right side into the right difference matrix;

[0016] Furthermore, step S3 specifically includes the following sub-steps:

[0017] Each element in the left difference matrix of each spectrum segment is judged. If the element is greater than zero, the state bit of the element is set to 1, otherwise it is set to 0. Finally, the left difference state bit matrix corresponding to all spectrum segments is obtained;

[0018] Each element in the right difference matrix of each spectral segment is judged. If the element is less than zero, the state bit of the element is set to 1, otherwise it is set to 0. Finally, the right difference state bit matrix corresponding to all spectral segments is obtained.

[0019] Furthermore, step S4 specifically includes the following sub-steps:

[0020] S41, summing up all elements in the state bit matrices on the left and right sides of each spectrum segment to obtain the sum of the state bits of the corresponding spectrum segment, and the sum of the state bits of all spectrum segments constitutes a state bit sum matrix of the entire spectrum;

[0021] S42, using the spectrum peak recognition criterion to sequentially judge each element in the state bit sum matrix of the entire spectrum, so as to realize fuzzy recognition of all spectrum peaks in the entire spectrum;

[0022] The spectrum peak identification criterion is specifically as follows: if the element is equal to the number of pixels in the spectrum band to which it corresponds minus 1, it is determined that there is and only one symmetrical spectrum peak in the spectrum band; if the element is equal to the number of pixels in the spectrum band minus 2, and the maximum spectrum intensity in the spectrum band is not greater than the intensity of the central wavelength, it is determined that there is a single glitch spectrum peak with a span of one pixel in the spectrum band; if the element is equal to the number of pixels in the spectrum band minus 3, and the maximum spectrum intensity in the spectrum band is not greater than the intensity of the central wavelength, it is determined that there is a two-glitch spectrum peak with a span of one pixel in the spectrum band, or there is a single glitch spectrum peak with a span of two pixels;

[0023] S43, outputting the wavelengths and spectral intensities of all identified spectral peaks.

[0024] In general, the present invention has the following advantages and effects:

[0025] 1. The method proposed in the present invention can realize accurate, rapid and automatic identification of laser probe spectrum peaks, thereby improving the analysis efficiency and accuracy of laser probe technology;

[0026] 2. The method proposed in the present invention can identify various types of asymmetric spectrum peaks, such as single burr spectrum peak, double burr spectrum peak and multi-burr spectrum peak, through one state bit interpretation, and there is no strict restriction on the shape and position of the burr, thus achieving the effect of fuzzy identification of laser probe spectrum peak;

[0027] 3. Compared with the traditional wavelet transform method, the method proposed in the present invention improves the quality of the identified spectral peaks, such as spectral intensity, signal-to-noise ratio and position accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a flow chart of a method for improving the accuracy of laser probe spectrum peak recognition by using morphological fuzzy judgment provided by the present invention;

[0029] Figure 2 is the average spectrum of two igneous rock samples;

[0030] Figure 3 This is a schematic diagram of common spectral peak categories of optical probes;

[0031] Figure 4 Taking the local spectrum section of 385.0-415.0 nm in the average spectrum as an example, a schematic diagram of the comparison of spectrum peaks identified by the method of the present invention and the traditional wavelet transform method is shown. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is now further described in detail in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention. In addition, the technical features involved in each embodiment of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0033] The method of the present invention mainly comprises the following specific steps:

[0034] Step 1: Obtain the original spectrum and average spectrum. Use the laser probe system to collect the laser probe spectrum and obtain the original spectrum data of the sample to be tested, denoted as N mⅹn , where m and n are the number of rows and columns in the original number excluding the wavelength column, and then the average spectrum is calculated based on this, recorded as M mⅹ2 ;

[0035] Step 2: Obtain the intensity difference matrix.

[0036] Step 2.1 Divide the spectrum into segments. Set the length of the spectrum segment to N seg , and each wavelength in the spectrum is taken as the central wavelength. Then the full spectrum of m pixels is divided into m spectrum segments, and each spectrum segment is recorded as the matrix Seg(N seg ⅹ1), the length of each spectrum segment is N seg pixels, then there are N pixels on both sides of the central wavelength. side =(N seg -1) / 2 pixels;

[0037] Step 2.2 Calculate the intensity difference matrix I on the left and right sides of the central wavelength left (mⅹN side ) and I right (mⅹN side ). Take a random spectral band in the spectrum as an example, subtract the spectral intensity of the first point on the left from the spectral intensity of its central wavelength, and then subtract the spectral intensity of the second point on the left from the spectral intensity of the first point on the left, and so on, until the left edge of the spectral band, and then get all the intensity differences of the spectral band and store them in a single left difference matrix Ileft (1ⅹN side ), and similarly, the left intensity difference matrix I of all other spectral segments can be obtained left (mⅹN side ) and the intensity difference matrix I on the right right (mⅹN side ). The intensity difference between the left and right boundary points is approximated as -1.

[0038] Step 3: Obtain the difference state bit matrix.

[0039] Step 3.1 First determine the left intensity difference matrix I left (mⅹN side ) is greater than 0, if so, the state bit of the element is 1, otherwise it is 0, and finally the left difference state bit matrix S of all spectrum segments is obtained. left (mⅹN side );

[0040] Step 3.2 Then determine whether each element of the right intensity difference matrix is ​​less than 0. If so, the state bit of the element is 1, otherwise it is 0. Finally, the right difference state bit matrix S of all spectrum segments is obtained. right (mⅹN side );

[0041] Step 4: Spectral peak fuzzy judgment.

[0042] Step 4.1 Merge the state bit matrices on the left and right sides horizontally and sum them to get the state bit and matrix S of the full spectrum sum (mⅹ1);

[0043] Step 4.2 Establish the spectral peak fuzzy judgment criteria:

[0044] a. When the i-th element S in the sum matrix sum (i) = N seg When -1, the spectrum segment has only one symmetrical peak;

[0045] b. When the i-th element S in the sum matrix sum (i) = N seg -2, and the maximum value of all intensity values ​​in the spectral band is less than the intensity of the central wavelength, then there is a single burr peak with a span of 1 pixel in the spectral band;

[0046] c. When the i-th element S in the sum matrix sum (i) = N seg -3, and the maximum value of all intensity values ​​in the spectral band is less than the intensity of the central wavelength, then the spectral band has a single glitch peak with a span of two pixels, or two double glitch peaks with a span of one pixel;

[0047] d. Based on this theory, we can deduce the criterion for multiple burr spectrum peaks;

[0048] Step 4.3 utilizes the spectrum peak judgment criteria obtained in step 4.2 to realize automatic fuzzy recognition of all spectrum peaks in the entire spectrum by sequentially judging the state bits and all elements in the matrix.

[0049] Step 5: Output the wavelength and spectral intensity of all identified spectral peaks.

[0050] Example

[0051] The embodiment of the present invention provides a method for improving the accuracy of laser probe spectrum peak recognition by using morphological fuzzy judgment, which mainly includes the following steps:

[0052] Step 1: Collect the original spectrum and average spectrum of the igneous rock sample. In this example, 12 kinds of igneous rock samples are collected for spectrum, each sample has 1, 10 points are collected at different positions of each sample, 10 spectra are collected at each point, and the number of pixels of each spectrum is 4096. Therefore, the original spectrum data matrix N 4096ⅹ1200 , whose average spectrum is M 4094ⅹ2 ,like Figure 1 shown.

[0053] Step 2: Obtain the intensity difference matrix.

[0054] Step 2.1 Assume the spectral length N of the symmetrical peak and the asymmetrical peak seg 9 and 11 pixels respectively.

[0055] Step 2.2 Take the spectrum segment with 268.51nm as the central wavelength as an example. The starting wavelengths of the spectrum segment are 268.33-268.7nm, and the spectral intensities are [304.33, 302.48, 336.15, 362.07, 397.35, 326.36, 316.44, 285.69, 274.47]. The spectral intensity 397.35 of the central wavelength is subtracted from the spectral intensity 362.07 of the first point on the left side of the central wavelength. Then, the spectral intensity 362.07 of the first point on the left is subtracted from the spectral intensity 336.15 of the second point on the left side. This is repeated until the left boundary of the spectrum segment. Then, all the intensity differences of the spectrum segment are obtained and stored in a single left difference matrix I. left (1ⅹ4)=[-1.85,33.67,25.92,35.28]. Similarly, the left and right intensity difference matrices of all other spectral segments can be obtained, which are:

[0056]

[0057] Step 3: Obtain the difference state bit matrix.

[0058] Step 3.1 First determine the left intensity difference matrix I left (1ⅹ4)=Whether the value of each element in [-1.85, 33.67, 25.92, 35.28] is greater than 0, if it is, the state bit of the element is 1, otherwise it is 0, then the left difference state bit matrix of the spectrum segment is [0, 1, 1, 1], and finally the left difference state bit matrix S of all spectrum segments is obtained. left (4094ⅹ4);

[0059] Step 3.2 Then determine whether the element of the right intensity difference matrix is ​​less than 0. If so, the state bit of the element is 1, otherwise it is 0. Then the right difference state bit matrix of the spectrum segment is [1,1,1,1]. Finally, the right difference state bit matrix S of all spectrum segments is obtained. right (4094ⅹ4), among which,

[0060]

[0061] Step 4: Fuzzy judgment of spectrum peaks.

[0062] Step 4.1: Merge the state bit matrices on the left and right sides of the spectrum segment horizontally to obtain the matrix [0, 1, 1, 1, 1, 1, 1]. The horizontal summation gives the sum of the state bits of the spectrum segment to be 7. Similarly, the state bits and matrix S of the entire spectrum can be obtained for the remaining spectrum segments. sum (4094ⅹ1)=[3,2,2,…,6];

[0063] Step 4.2 Establish fuzzy criteria:

[0064] a. When the i-th element S in the sum matrix sum When (i) = 8, the spectrum segment has only one symmetrical peak;

[0065] b. When the i-th element S in the sum matrix sum When (i) = 7, and the maximum value of all intensities in the spectral band is less than the intensity of the central wavelength, there is a single burr peak with a span of 1 pixel in the spectral band;

[0066] c. When the i-th element S in the sum matrix sum (i) = 6, and the maximum value of all intensity values ​​in the spectral band is less than the intensity of the central wavelength, then the spectral band has a single glitch peak with a span of two pixels, or two double glitch peaks with a span of one pixel;

[0067] d. Based on this theory, we can deduce the criterion for multiple burr spectrum peaks;

[0068] Step 4.3 uses the fuzzy criterion obtained in step 4.2. Since the spectrum segment satisfies S sum(i) = 9-2 = 7, and the maximum values ​​of all intensity values ​​in the spectrum except the central wavelength are less than the intensity of the central wavelength, as shown in its spectrum intensity matrix [304.33, 302.48, 336.15, 362.07, 397.35, 326.36, 316.44, 285.69, 274.47]. Therefore, there is a single burr spectrum peak with a span of 1 pixel in this spectrum; then, by sequentially judging the remaining state bits and all elements in the matrix, automatic fuzzy recognition of all spectrum peaks in the entire spectrum is achieved.

[0069] Step 5: Output the wavelengths and spectral intensities of all identified spectral peaks, as shown in Table 1.

[0070] Table 1 Wavelengths and spectral intensities of all spectral peaks identified by the method of the present invention

[0071]

[0072] Based on the above embodiments, the present invention provides a method for improving the accuracy of laser probe spectrum peak recognition by using morphological fuzzy judgment (hereinafter referred to as fuzzy judgment method) and a traditional spectrum peak recognition method based on wavelet transform (hereinafter referred to as wavelet transform method). All spectrum peaks recognized by them are as follows: Figure 4 The peak quality comparison results are shown in Table 2:

[0073] Table 2 Comparison of the number and quality of spectral peaks identified by the method of the present invention and the traditional wavelet transform method

[0074]

[0075] Depend on Figure 4 It can be seen that both methods can automatically identify the vast majority of significant spectral peaks in the entire spectrum, and as can be seen from Table 2, the number of identical spectral peaks identified by the two methods is 86, accounting for approximately 58.1% of the total number of 148, and the remaining 62 are their own unique spectral peaks, that is, inconsistent spectral peaks identified by the two methods. After calculation of the three commonly used spectral quality indicators of the average spectral intensity, average signal-to-noise ratio, and average signal-to-background ratio of the 62 unique spectral peaks, the average spectral intensity, average signal-to-noise ratio, and average signal-to-background ratio of the unique spectral peaks of the fuzzy judgment method proposed in the present invention are 2400.6, 181.6, and 1.5, respectively, which are significantly higher than 2046.1, 157.3, and 1.1 of the wavelet transform method. Therefore, its spectral peak quality is better than the traditional wavelet transform method. In addition, by Figure 4By comparing the local details of the identified spectral peaks in the figure, it can be found that the spectral peaks identified by the traditional wavelet transform method are mostly located at non-spectral peak positions, as shown by the five-pointed star spectral peak positions indicated by multiple arrows near wavelengths such as 385.0nm, 405.0nm and 412.5nm in the figure, while this situation does not exist in the method of the present invention. In summary, the spectral peak identification accuracy of the method of the present invention is significantly superior to that of the traditional wavelet transform method.

[0076] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for improving the accuracy of laser probe spectrum peak recognition by using morphological fuzzy judgment, characterized in that: The method comprises the following steps: S1, perform LIBS spectrum acquisition to obtain plasma spectrum data; S2, set the length of the spectrum segment, divide the spectrum segment, and calculate the spectral intensity difference between all adjacent points on both sides of the central wavelength in all spectrum segments, and obtain the intensity difference matrix on the left and right sides of all spectrum segments; S3, using the intensity difference matrices on the left and right sides of all spectrum segments to obtain the difference state bit matrices on the left and right sides of all spectrum segments respectively; S4, summing up all elements in the state bit matrix on the left and right sides of each spectrum segment to obtain the sum of the state bits of the corresponding spectrum segment, and the sum of the state bits of all spectrum segments constitutes the state bit sum matrix of the entire spectrum; establishing a fuzzy criterion, first using the fuzzy criterion to identify the symmetrical spectrum peak and the asymmetrical spectrum peak, and then further identifying the single spur spectrum peak, the double spur spectrum peak and the multi-spur spectrum peak in the asymmetrical spectrum peak; S5, output the wavelengths and spectral intensities of all identified spectral peaks.

2. The method of improving the accuracy of laser probe spectrum peak recognition by using morphological fuzzy judgment according to claim 1 is characterized in that: Step S2 specifically includes the following sub-steps: S21, taking all wavelengths of the full spectrum as the central wavelength, setting the spectrum length, and specifying the spectrum range for each central wavelength; S22, calculating the difference of all adjacent points on the left side of the central wavelength in the spectrum segment, subtracting the spectrum intensity of the first point on the left side from the spectrum intensity of the central wavelength to obtain the first intensity difference on the left side, then subtracting the spectrum intensity of the second point on the left side from the spectrum intensity of the first point on the left side to obtain the second intensity difference on the left side, and so on until the left boundary of the spectrum segment, and storing all the obtained intensity differences on the left side into the left difference matrix; Calculate the difference of all adjacent points on the right side of the central wavelength in the spectrum band, subtract the spectral intensity of the central wavelength from the spectral intensity of the first point on the right side of the central wavelength to obtain the first intensity difference on the right side, then subtract the spectral intensity of the first point on the right side from the spectral intensity of the second point on the right side to obtain the second intensity difference on the right side, and repeat this process until the right boundary of the spectrum band, and then store all the intensity differences on the right side into the right difference matrix.

3. The method of improving the accuracy of laser probe spectrum peak recognition by using morphological fuzzy judgment according to claim 1 is characterized in that: Step S3 specifically includes the following sub-steps: Each element in the left difference matrix of each spectrum segment is judged. If the element is greater than zero, the state bit of the element is set to 1, otherwise it is set to 0. Finally, the left difference state bit matrix corresponding to all spectrum segments is obtained; Each element in the right difference matrix of each spectral segment is judged. If the element is less than zero, the state bit of the element is set to 1, otherwise it is set to 0. Finally, the right difference state bit matrix corresponding to all spectral segments is obtained.

4. The method of improving the accuracy of laser probe spectrum peak recognition by using morphological fuzzy judgment according to claim 1 is characterized in that: Step S4 specifically includes the following sub-steps: S41, summing up all elements in the state bit matrices on the left and right sides of each spectrum segment to obtain the sum of the state bits of the corresponding spectrum segment, and the sum of the state bits of all spectrum segments constitutes a state bit sum matrix of the entire spectrum; S42, using the spectrum peak recognition criterion to sequentially judge each element in the state bit sum matrix of the entire spectrum, so as to realize fuzzy recognition of all spectrum peaks in the entire spectrum; The spectrum peak identification criterion is specifically as follows: if the element is equal to the number of pixels in the spectrum band to which it corresponds minus 1, it is determined that there is and only one symmetrical spectrum peak in the spectrum band; if the element is equal to the number of pixels in the spectrum band minus 2, and the maximum spectrum intensity in the spectrum band is not greater than the intensity of the central wavelength, it is determined that there is a single glitch spectrum peak with a span of one pixel in the spectrum band; if the element is equal to the number of pixels in the spectrum band minus 3, and the maximum spectrum intensity in the spectrum band is not greater than the intensity of the central wavelength, it is determined that there is a two-glitch spectrum peak with a span of one pixel in the spectrum band, or there is a single glitch spectrum peak with a span of two pixels; S43, outputting the wavelengths and spectral intensities of all identified spectral peaks.

Images