WO2021052240A1 - Laser probe classification method and device capable of automatically selecting spectral lines on basis of image features - Google Patents

Abstract

The present invention relates to the related technical field of laser probe component analysis, and provides a laser probe classification method and device capable of automatically selecting spectral lines on the basis of image features. The method comprises the following steps: (1) acquiring a plasma spectrum of a sample, and imaging the plasma spectrum to obtain a spectral image; (2) extracting image features and image feature coordinates of the spectral image, and converting the extracted image feature coordinates into actual wavelengths of analytical lines according to linear correspondences between the image features and the actual wavelengths of the analytical lines; and (3) extracting spectral intensity of the analytical lines, constructing a classification model by combining the obtained intensity of the analytical lines with a classification algorithm, and then classifying, by means of the classification model, a product to be classified. The present invention improves the classification efficiency and the classification precision, has a high automation degree, and can effectively avoid human factors while improving the line selection efficiency.

Description

Laser probe classification method and device for automatically selecting spectral lines based on image features 【Technical Field】

The invention belongs to the related technical field of laser probe component analysis, and more specifically, relates to a laser probe classification method and device for automatically selecting spectral lines based on image features.

【Background technique】

Laser probes are laser-induced breakdown spectroscopy (LIBS), which is an elemental analysis technology that uses plasma spectra generated by laser focusing on the sample surface to analyze the composition and content of substances. Compared with traditional chemical analysis techniques such as XRD (X-ray diffusion), XRF (X Ray Fluorescence), ICP-MS (Inductively coupled plasma-Mass Spectrometry), laser probe technology has simple sample preparation and can achieve rapid in-situ Advantages such as detection. LIBS technology has been widely used in pesticide residue detection, geological survey, rock and ore detection and other fields, and research related to its analysis methods has also developed rapidly.

In quantitative or qualitative analysis based on LIBS technology, the choice of analysis line has a crucial influence on the analysis structure. In the LIBS technology, the current commonly used method of line selection is still the traditional manual line selection method. The manual line selection method is generally based on multiple factors such as spectral intensity, background intensity, signal-to-background ratio, signal-to-noise ratio, and whether there is self-absorption. Comprehensive judgment to select the analysis line, therefore, manual line selection has a higher accuracy rate. At present, relevant researchers in this field have done some research, such as the paper "Quantitative analysis of Lead Zirconate Titanate (PZT) ceramics by laser-induced breakdown spectroscopy (LIBS) in combination with multivariate calibration" (Rafael, Microchemical Journal 130 (2017) ):21-26.) discloses an element multivariate analysis method based on manual line selection and PLS algorithm, and quantitatively analyzes the elements such as lead, zirconium, titanium in ceramics, but the shortcomings of this method are also very obvious. For example, manual line selection needs to judge the target spectrum one by one, so the line selection efficiency is low; and this method relies heavily on line selection experience, and the analysis results are obviously affected by humans. The existing automatic line selection method has also developed to a certain extent, such as the paper "In situ classification of rocks using stand-off laser induced breakdown spectroscopy with a compact spectrometer" (WTLi et al., Journal of Analytical Atomic Spectrometry 33.3 (2018): 461-467.) also proposed an automatic line selection method based on wavelet transform algorithm, and combined with LDA algorithm for rock classification. However, in the LIBS technology, the automatic route selection method is not yet mature, and there is no method that can be universally applied and has a standardized operation process, and there are relatively few classification methods based on automatic route selection.

[Summary of the invention]

In view of the above defects or improvement needs of the prior art, the present invention provides a laser probe classification method and device that automatically selects spectral lines based on image features, which solves the problem of low line selection efficiency and easy line selection results in the existing classification methods. Affected by human factors, the analysis process is time-consuming and the analysis performance is poor. The analysis method can automatically perform the line selection process before classification, and effectively avoid human factors while improving the line selection efficiency, thereby improving the classification efficiency and classification accuracy.

To achieve the above objective, according to one aspect of the present invention, a laser probe classification method that automatically selects spectral lines based on image features is provided. The laser probe classification method includes the following steps:

(1) Collect the plasma spectrum of the sample, and perform image processing on the plasma spectrum to obtain a spectral image;

(2) Perform image feature and image feature coordinate extraction on the spectral image, and convert the extracted image feature coordinate into the actual wavelength of the analysis line according to the linear correspondence between the image feature and the actual wavelength of the analysis line;

(3) Extract the spectral intensity of the analysis line, and combine the obtained intensity of the analysis line and the classification algorithm to construct a classification model, and then use the classification model to classify the products to be classified.

Further, in step (1), firstly, the plasma spectrum is displayed graphically, and the size of the spectral image is set; then, the spectral line area of the plasma spectrum displayed in the graphical display is intercepted, and the intercepted spectral line area is stored Is an image, from which a spectral image is obtained.

Further, the image feature is a corner feature, and the corner feature is an acceleration section test feature, a minimum feature value feature, or a Harris feature.

Further, in step (2), first, set the coefficient of determination Q=Q1 for the corner features of the image to obtain all corner features in the spectral image to obtain the corner set CornerSet1; then, set Q=Q2, and Q2> Q1, to obtain the corner point features at the trough position and the peak position in the spectral image to obtain the corner point set CornerSet2; then, calculate the difference between the corner point set CornerSet1 and the corner point set CornerSet2 to obtain the corner point set CornerSet4. The point set CornerSet4 is the set of corner points corresponding to the analysis line to be sought.

Further, the actual spectral wavelength value and the coordinate value of the spectral image feature of the two different points in the plasma spectrum are determined according to the correspondence between the historical spectrum and the spectral image; the actual spectral wavelength value and the spectral image characteristic coordinate value are compared according to the obtained coordinates of the two points. The coordinate values of the image features are linearly fitted to obtain a linear relationship expression, and then the extracted image feature coordinates are converted into the actual wavelength of the analysis line based on the linear relationship expression.

Further, the two points are respectively located at the front end and the back end of the spectral band.

Further, the coordinates of the two points are (x ₁ , y ₁ ) and (x ₂ , y ₂ ) respectively, and the corresponding linear relationship expression is y=kx+b, where k=(y ₂ -y ₁ ) /(x ₂ -x ₁ ), b=y ₁ -kx ₁ .

Further, the classification algorithm is any one of the following algorithms: linear discriminant analysis algorithm, support vector machine classification algorithm, neural network classification algorithm, and K nearest neighbor algorithm.

Further, in step (3), the spectral intensity of the analysis line corresponding to the product to be classified is input into the classification model, and the classification model classifies the product to be classified according to the received spectral intensity data.

According to another aspect of the present invention, there is provided a laser probe classification device that automatically selects spectral lines based on image features. The laser probe classification and analysis device uses the above-mentioned laser probe that automatically selects spectral lines based on image features. According to the classification method to classify the products to be classified.

In general, compared with the prior art through the above technical solutions conceived by the present invention, the laser probe classification method and device for automatically selecting spectral lines based on image features provided by the present invention mainly has the following beneficial effects:

1. The classification method is based on automatic line selection classification based on image features. The entire classification process does not require manual intervention, and the classification result is not affected by human factors. Compared with the existing manual line selection classification method, the method of the present invention Higher efficiency, stronger applicability, and more stable analysis results.

2. This method recognizes the position of the corner point and the corresponding analysis line wavelength through image features before classification, and effectively extracts the analysis line with greater relative intensity through the detection of the top angle of the spectrum. The stronger the relative intensity of the spectral line, the more It is conducive to qualitative or quantitative analysis. The existing line selection method generally first performs algorithm transformation on the spectrum, and then selects the line by calculating the signal-to-background ratio or signal-to-noise ratio of the spectrum. In this process, the spectral background or noise There is often a large error in the determination of, which affects line selection and classification performance. The classification method provided by the present invention does not have the above problems.

3. The image features can be corner features or other image features, where the corner features can be accelerated segment test features (FAST features), minimum feature value features (MinEigen features), Harris features; the classification algorithm can be commonly used Machine learning algorithms such as linear discriminant analysis algorithm, support vector machine algorithm, the above image features and classification algorithms can be combined with each other, so the method of automatic line selection and classification based on image features has better practicability and flexibility.

【Explanation of the drawings】

Fig. 1 is a schematic flow chart of a laser probe classification method based on image features that automatically selects spectral lines according to the present invention;

2 is a schematic diagram of a laser probe classification device that automatically selects spectral lines based on image features provided by the present invention;

Fig. 3 is a schematic diagram of the principle of Harris corner feature detection of spectral image involved in the laser probe classification method based on image feature automatic selection of spectral lines in Fig. 1;

4 is a schematic diagram of Harris corner screening involved in the laser probe classification method based on image feature automatic selection of spectral lines in FIG. 1;

FIG. 5 is a schematic diagram of the Harris corner point detection process involved in the laser probe classification method based on image feature automatic selection of spectral lines in FIG. 1; FIG.

(A), (b), (c), and (d) in FIG. 6 are respectively the schematic diagrams of Harris corner screening results obtained by using the laser probe classification method of automatically selecting spectral lines based on image features in FIG. 1;

Figure 7 is a schematic diagram of the analysis line position of manual line selection;

8 is a schematic diagram of the correspondence between the coordinates of the image corner points and the actual wavelength of the analysis line involved in the laser probe classification method of automatically selecting spectral lines based on image features in FIG. 1;

9 is a schematic diagram of the linear relationship between the coordinates of the image corner points and the actual wavelength of the analysis line involved in the laser probe classification method based on the image feature automatically selecting the spectral line in FIG. 1;

(A) and (b) in FIG. 10 are schematic diagrams of analysis results obtained by using the existing manual line selection classification method and the laser probe classification method of automatically selecting spectral lines based on image features in FIG. 1, respectively.

In all the drawings, the same reference numerals are used to denote the same elements or structures, among which: 1-spectrometer, 2-laser, 3-control unit, 4-microprocessor, 5-collecting head, 6-focusing mirror , 7-sample, 8-connector, 9-fiber.

【detailed description】

In order to make the objectives, technical solutions, and advantages of the present invention clearer, the following further describes the present invention in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, but not to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

Please refer to Figure 1, Figure 2, Figure 3 and Figure 4, the present invention provides a laser probe classification method that automatically selects spectral lines based on image features. The laser probe classification method first needs to convert the LIBS spectrum into a spectral image, and then Extract image features from the spectral image, then establish a linear correspondence between the image features and the wavelength of the analysis line and calculate the actual wavelength of the analysis line. Finally, extract the intensity of the analysis line and combine it with the classification algorithm to realize the identification and classification of the sample .

The laser probe classification method mainly includes the following steps:

S1, using a laser probe sorting device to collect the plasma spectrum of the sample.

Specifically, the sample 7 is placed in the sample placement area of the laser probe classification device, and the control system and the optical path system of the laser probe classification device are combined to realize the spectrum collection of the plasma. Wherein, the laser probe classification device for automatically selecting spectral lines based on image features provided by the present invention, the laser probe classification device includes a spectrometer 1, a laser 2, a control unit 3, a microprocessor 4, a collecting head 5, and a focusing mirror 6. , Connecting the line 8 and the optical fiber 9, the laser 2 and the focusing lens 6 are arranged along the horizontal direction. The control unit 3 is connected to the microprocessor 4, the control unit 3 is respectively connected to the spectrometer 1 and the laser 2 through the connecting wire 8, and the collection head 5 is connected to the optical fiber 9 through the optical fiber 9. The spectrometer 1. The collecting head 5 is located above the focusing mirror 6 and forms a certain angle with the horizontal plane. After the collecting head 5, the plasma spectrum is transmitted to the spectrometer 1 via the optical fiber 9.

The laser 2 is used to generate high energy density laser, the focusing mirror 6 is used to focus the high energy density laser generated by the laser 2, and the control unit 3 is used to coordinate and control the operation of the laser 2 and the spectrometer 1 , And collect other sensor signals of the system. The collection head 5 is used to collect the plasma spectrum generated by the laser beam irradiated on the surface of the sample 7, and the microprocessor 4 is used to control the entire laser probe system and store the generated Spectral data.

The high-energy-density laser beam generated by the laser 2 is focused by the focusing mirror 6 and then irradiated on the surface of the sample 7. The sample 7 is ablated by the high-energy-density laser beam to generate plasma, and the plasma emission spectrum After being collected by the collecting head 5, it is transmitted to the spectrometer 1 via the optical fiber 9. Wherein, the entire spectrum collection process is realized through coordinated control of the control unit 3 and the microprocessor 4.

S2: Perform image processing on the plasma spectrum to obtain a spectrum image.

Specifically, the plasma spectrum is composed of a wavelength column and an intensity value column. When drawing a spectral image, the wavelength is the horizontal axis and the spectral intensity is the vertical axis; and in this embodiment, the horizontal and vertical axes need to be removed to intercept the spectrum. Display the area, and take the intercepted area as the spectral image. The resolution of the spectral image depends on the preset value.

In one embodiment, the plasma spectrum data is displayed graphically, and then the size of the spectral image is set, and the length of the image is L and the width is H, and then the spectral line region of the plasma surface spectrum displayed graphically is intercepted, This area is the entire spectral area excluding the abscissa and ordinate axis of the spectrum. Finally, the obtained L*H image area is stored as a PNG format image, which is a spectral image.

S3: Perform image feature and image feature coordinate extraction on the spectral image, and identify the coordinate of the image feature in the spectral image.

Specifically, the image features can be corner features or other image texture features or shape features, etc., where the corner features can be accelerated segment test features (FAST features), minimum feature value features (MinEigen features) or Harris features, etc., please refer to As shown in Fig. 5 and Fig. 6, step S3 specifically includes the following steps:

S31, taking the image corner feature as an example, the parameter Q is the coefficient of determination of the image corner feature, the smaller the Q is, the more image corners can be detected in the spectral image, and vice versa. In this step, first set Q=Q1, and Q1 is a small value to obtain all corner point features in the spectral image, to obtain a corner point set CornerSet1, which has m corner point features in total.

S32: Set Q=Q2 and Q2>Q1 to obtain corner features at the trough position and the peak position at the low-intensity peak in the spectral image, to obtain a corner set CornerSet2, which has n corner features in total.

S33: Calculate the difference set between the corner point set CornerSet1 and the corner point set CornerSet2 to obtain the corner point set CornerSet4. The corner point set CornerSet4 corresponds to the corner point set of the desired analysis line, and the corner point set CornerSet4 shares the corner points. Features (mn).

S4: According to the linear correspondence between the image feature and the actual wavelength of the analysis line, the extracted image feature coordinates are converted into the actual wavelength of the analysis line.

Specifically, referring to FIG. 8 and FIG. 9, step S4 specifically includes the following steps:

S41. Determine the actual spectral wavelength values of two different points in the plasma spectrum and the coordinate values of the spectral image features according to the correspondence between the historical spectrum and the spectral image, where these two points should be located closer to the front and back of the spectral band. position.

S42: Perform a linear fit to the actual spectral wavelength value and the coordinate value of the spectral image feature according to the obtained coordinates of the two points to obtain a linear relationship expression, and then convert the extracted image feature coordinates into analysis based on the linear relationship expression The actual wavelength of the line.

In one embodiment, the coordinates of the extracted (mn) image features are converted into actual analysis line wavelengths to obtain (mn) analysis line wavelengths. The conversion basis is the linear correspondence between the image features and the actual wavelengths of the analysis line. ; First, according to the corresponding relationship between the historical spectrum and the spectral image, determine the actual spectral wavelength values of the two different points in the spectrum and the coordinate values of the corner points of the spectral image, such as A(x1, y1), B(x2, y2). The two points should be located closer to the front and back of the spectral band, as shown in Figure 8. After that, according to _{the coordinates of the two points A (x 1} , y ₁ ) and B (x ₂ , y ₂ ), the actual spectral wavelength value and the coordinate value of the corner feature of the spectral image are linearly fitted to obtain the linear relationship expression The formula y=kx+b, where k=(y ₂ -y ₁ )/(x ₂ -x ₁ ), b=y ₁ -kx ₁ .

S5, extract the spectral intensity of the analysis line.

Specifically, assuming that the number of samples in the experiment is a and the number of spectra collected for each sample is b, the total number of spectra is N _S =a*b, and the dimensions of the extracted analysis line intensity matrix are (mn) rows and N _S Column.

S6: Combine the obtained spectrum intensity of the analysis line with the classification algorithm to construct a classification model, and then use the classification model to classify the sample.

Specifically, please refer to Figures 7 and 10, the classification algorithm is any one of the following algorithms: linear discriminant analysis algorithm, support vector machine classification algorithm, neural network classification algorithm and K nearest neighbor algorithm; corresponding to the product to be classified The spectral intensity of the analysis line is input to the classification model, and then the classification model classifies the products to be classified.

Example 1

The laser probe classification method for automatically selecting spectral lines based on image features provided by the first embodiment of the present invention mainly includes the following steps:

Step one, sample preparation and spectrum collection. This example uses 24 kinds of igneous rock samples, which are natural stones without any polishing or other treatments; in this example, the maximum single laser pulse energy is 6.3mJ, the frequency is 10Hz, the wavelength is 1064.310nm, and the focal length of the focusing lens It is 25mm, the band of the spectrometer is 268nm～430nm, and the detector of the spectrometer is 4094 pixels. In addition, the spectrum collection method collects 4 points for each rock sample, and each point collects 25 spectra, so a total of 100 spectra are collected for each sample.

Step 2: Perform image processing on the plasma spectrum to obtain a spectrum image. First, the average spectrum of all the spectra in this embodiment is calculated, and the average spectrum is displayed graphically. In this embodiment, the length L and width H of the spectral image are set to 740 and 600, respectively, and the image storage format is PNG format. In addition, the imaging of the spectrum is mainly carried out through the library functions of Matlab. First, the average spectrum is displayed graphically through the plot (xWvData, yItyData) function, where xWvData is the wavelength data of the spectrum, yItyData is the intensity data of the spectrum, and then through getframe The () function intercepts the spectral region of the image with the set resolution, and finally saves the intercepted spectral image through the imwrite() function.

Step three, image corner feature extraction. The image feature used in this embodiment is Harris corner feature. In this step, first set the value Q1 to 0.005 to obtain all the corner features and their coordinates of the spectral image, the corner feature set is CornerSet1, and the feature number is 324. Then, set Q2 to 0.21 to obtain the corner points of the trough position and the low-intensity peak position, the corner point set is CornerSet2, and the number of corner points n is 153. Finally, calculate the difference set of CornerSet1 and CornerSet2 to get CornerSet4, where the value of (m–n) is 171, that is, the corner points in the range of 0.005<Q<0.21, and the coordinates of the corner points in the set CornerSet4 are the analysis lines Location, as shown in Figure 7.

Step 4: Convert the corner coordinates to the actual wavelength of the analysis line. In this step, first, according to the corresponding relationship between the plasma spectrum and the spectral image, query the spectral wavelength values of two different points in the spectrum and the coordinate values of the corner points of the spectral image. In this embodiment, point A (42.51, 279.55) and B (685.37, 422.67). Among them, 42.51 and 685.37 are the abscissa values of corner points A and B, and 279.55 and 422.67 are the actual wavelength values of the analysis line corresponding to the corner points. Then, according to the coordinate values of points A and B, linear fitting is performed on the characteristic coordinate values of the corner points and the actual wavelength values to obtain the linear relationship expression y=0.226x+270.1190.

Step five, extract the spectral intensity of the analysis line. In this embodiment, the number of samples is a=24, and the number of spectra collected by each sample is b=100, then the total number of spectra is N _S =a*b=2400, and the dimension of the extracted analysis line intensity matrix is 171 Rows and 2400 columns.

Step 6: Combine the classification algorithm to establish a classification model for classification. In this embodiment, the classification algorithm used is the linear discriminant analysis (LDA) algorithm. When combined with this algorithm, the ratio of the number of spectra in the training set to the number of spectra in the test set is 8:2, and the data dimension of the training set is 171× 1920, the data dimension of the test set is 171×480. In this step, the above data is trained and tested to obtain a classification model.

The comparison between the method (IFALS-LDA) provided by the present invention and the manual line selection classification method (MLS-LDA) to classify 24 rock samples is shown in Figure 10. The analysis lines of the manual line selection classification method are shown in Table 1. . It can be seen from Figure 10 that the overall average classification accuracy of the MLS-LDA classification is 94.38%, while the overall average classification accuracy of the IFALS-LDA classification is 98.54%. After adopting the method of the present invention, the overall classification accuracy rate is effectively improved, and the increase range is 4.16%. In order to further verify the method of the present invention, the classification model was subjected to 10-fold cross-validation after the classification, wherein the cross-validation accuracy rates of MLS-LDA and IFALS-LDA were 95% and 98.18%, respectively.

Table 1 List of analysis lines selected by manual line selection method

In addition, in terms of classification efficiency, the MLS-LDA method requires 2760s in the entire classification process, while the IFALS-LDA method is a classification method based on automatic line selection, and the time required is only about 4.34s. The above shows that the IFALS-LDA classification method can greatly reduce the classification time of LIBS technology and improve the classification efficiency. Table 2 shows the comparison of comprehensive classification performance indexes of MLS-LDA and IFALS-LDA.

Table 2 Comprehensive classification performance indicators of MLS-LDA and IFALS-LDA

Observation index	MLS-LDA	IFALS-LDA
Forecast accuracy rate (%)	94.38	98.54
Cross-validation accuracy rate (%)	95.00	98.18
Running time (s)	2760.00	4.36
Number of analysis lines	46	171
Operation mode	Manual line selection	Automatic line selection

Those skilled in the art can easily understand that the above descriptions are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modification, equivalent replacement and improvement, etc. made within the spirit and principle of the present invention, All should be included in the protection scope of the present invention.

Claims (10)

1. A laser probe classification method for automatically selecting spectral lines based on image features is characterized in that the laser probe classification method includes the following steps:

   (1) Collect the plasma spectrum of the sample, and perform image processing on the plasma spectrum to obtain a spectral image;

   (2) Perform image feature and image feature coordinate extraction on the spectral image, and convert the extracted image feature coordinate into the actual wavelength of the analysis line according to the linear correspondence between the image feature and the actual wavelength of the analysis line;

   (3) Extract the spectral intensity of the analysis line, and combine the obtained intensity of the analysis line and the classification algorithm to construct a classification model, and then use the classification model to classify the products to be classified.
2. The laser probe classification method for automatically selecting spectral lines based on image features according to claim 1, characterized in that: in step (1), first, the plasma spectrum is imaged and displayed, and the size of the spectrum image is set; , Intercept the spectral line area of the plasma spectrum displayed in the image, and store the intercepted spectral line area as an image, thereby obtaining the spectral image.
3. The laser probe classification method for automatically selecting spectral lines based on image features according to claim 1, wherein the image feature is a corner point feature, and the corner point feature is an acceleration section test feature, a minimum feature value feature, or Harris characteristics.
4. The laser probe classification method for automatically selecting spectral lines based on image features according to claim 3, characterized in that: in step (2), first, the determination coefficient Q=Q1 of the image corner feature is set to obtain the spectral image Then, set Q=Q2 and Q2>Q1 to obtain the corner features at the trough position and the peak position in the spectral image to obtain the corner set CornerSet2; then, calculate the corner point set CornerSet1. The difference of the corner set CornerSet1 and the corner set CornerSet2 to obtain the corner set CornerSet4, the corner set CornerSet4 is the corner set corresponding to the desired analysis line.
5. The laser probe classification method for automatically selecting spectral lines based on image features according to claim 4, wherein the actual spectral wavelength values of two different points in the plasma spectrum are determined according to the correspondence between historical spectra and spectral images. The coordinate value of the spectral image feature; according to the obtained coordinates of the two points, the actual spectral wavelength value and the coordinate value of the spectral image feature are linearly fitted to obtain a linear relationship expression, and then the image to be extracted based on the linear relationship expression The characteristic coordinates are converted to the actual wavelength of the analysis line.
6. The laser probe classification method for automatically selecting spectral lines based on image features according to claim 5, wherein the two points are respectively located at the front and back ends of the spectral band.
7. The laser probe classification method for automatically selecting spectral lines based on image features according to claim 5, characterized in that: the coordinates of the two points are (x ₁ , y ₁ ) and (x ₂ , y ₂ ) respectively, then corresponding The linear relationship expression of is y=kx+b, where k=(y ₂ -y ₁ )/(x ₂ -x ₁ ), b=y ₁ -kx ₁ .
8. The laser probe classification method for automatically selecting spectral lines based on image features according to any one of claims 1-7, wherein the classification algorithm is any one of the following algorithms: linear discriminant analysis algorithm, support vector Machine classification algorithm, neural network classification algorithm and K nearest neighbor algorithm.
9. The laser probe classification method for automatically selecting spectral lines based on image features according to any one of claims 1-7, characterized in that: in step (3), the spectral intensity of the analysis line corresponding to the product to be classified is input into the A classification model, which classifies the products to be classified according to the received spectral intensity data.
10. A laser probe classification device that automatically selects spectral lines based on image features, characterized in that: the laser probe classification device adopts the laser probe that automatically selects spectral lines based on image features according to any one of claims 1-9. According to the classification method to classify the products to be classified.

Images