An AI spectrum identification method

Abstract

The application discloses an AI spectrum graph identification method, and belongs to the technical field of graph identification, and comprises the following steps: step one, identifying a pattern, converting information contained in a data carrier into spectrum graph information; step two, judging a peak type, judging whether the spectrum graph is a multi-peak type graph or a single-peak type graph according to the collected spectrum graph information; step three, performing integration, performing baseline integration and tailing peak processing on the single-peak type graph, and performing cutting integration processing on the double-peak type graph; step four, comparing the integral results, comparing the areas of the integral results, and outputting the final integral result; the problems that the spectrum graph identification efficiency is low in the prior art and the identification accuracy is poor in artificial intelligence are solved.

Description

Technical Field

[0001] This invention belongs to the field of spectrum recognition technology, specifically relating to an AI spectrum recognition method. Background Technology

[0002] Agricultural products entering the market must undergo inspection and testing to ensure food safety. National standards also select chromatography instruments as the detection tool. During operation, chromatography instruments generate analytical spectra, which are crucial raw data required by national standards, necessitating accurate integration and quantitative analysis. However, automatic spectrum integration by chromatography instruments cannot meet the needs of inspection and testing work, thus requiring manual operation. The large volume of pesticide and veterinary drug residue samples and the numerous compound tests result in low work efficiency and a high demand for specialized personnel, placing significant manpower pressure on the inspection and testing industry. Inspection and testing personnel often work overtime, and prolonged work can easily lead to fatigue-related errors. This can result in erroneous reports and legal risks for inspection and testing institutions.

[0003] Currently, instrument manufacturers are mostly researching and expanding instrument workstations, and progress in the professional automatic integration of spectra is relatively slow. Only a few companies are using large-scale models to explore more intelligent spectrum integration, but large-scale models suffer from data ambiguity, while national and even international standards for testing and inspection require absolute accuracy, making the application of large-scale models in this direction relatively limited. Summary of the Invention

[0004] To address the problems of low efficiency in manual spectrum recognition and poor accuracy in artificial intelligence recognition in existing technologies, this invention proposes an AI spectrum recognition method, comprising the following steps: Step 1: Identify the pattern and convert the information contained in the data carrier into spectral information; Step 2: Determine the peak type. Based on the collected spectrum information, determine whether the spectrum is a single-peak or multi-peak spectrum. Step 3: Perform integration. For unimodal graphs, perform baseline integration and tailing peak processing. For bimodal graphs, perform segmentation integration. Step 4: Compare the integrated results by area and output the final integrated result.

[0005] Preferably, in step one, the data carrier includes PDF, tables, and internal instrument data.

[0006] A preferred approach is to determine the peak shape in step two as follows: First, set the starting point of the peak shape graph as the integration starting point A, the ending point of the peak shape graph as the integration ending point B, the highest point of the peak shape graph as point C, the midpoint G between points A and C as the half-peak width point, point E as the highest point of the first adjacent peak, point H as the highest point of the second adjacent peak, point I as the highest point of the third adjacent peak, point D as the concave point between the two peaks, and point F as the calculation point when finding the baseline integration. Then, check whether the above points appear in sequence, and determine the peak shape based on the points found.

[0007] Preferably, in step two, if only point C is found when searching for the highest point, the spectrum is classified as a single-peaked spectrum; if points C and E are found, the spectrum is classified as a double-peaked spectrum; if points C, E, and I are found, the spectrum is classified as a multi-peaked spectrum.

[0008] Ideally, in step two, if the spectrum is determined to be a multi-peak spectrum, an error should be reported directly, and the process should return to step one.

[0009] Preferably, in step three, if the spectrum is determined to be a single-peaked spectrum, then the single-peaked spectrum is integrated; if the spectrum is determined to be a double-peaked spectrum, then the double-peaked spectrum is segmented and integrated.

[0010] Preferably, in step three, when integrating the single-peak graph, it is first necessary to find the starting point A and the ending point B of the integration; if the integration curve does not change, it is determined that there is no peak here; if both points A and B can be found, then the peak graph is integrated along the baseline; if only one of points A and B can be found, then the peak graph is processed to produce a tailing peak.

[0011] In step three, the baseline integration is performed as follows: First, find point F such that line segment AF is perpendicular to the Y-axis and line segment DF is perpendicular to the X-axis; then calculate the area of ​​the figure ACDF as the integration result.

[0012] A preferred approach to processing the tailing peak is as follows: Find an inflection point on the peak shape, designated as point B, and determine a point F such that line segment AF is perpendicular to the Y-axis and line segment DF is perpendicular to the X-axis; then perform a downward baseline integration; if there is a height difference between point A and point B, follow the principle of vertical extension from the high point to the low point; the inflection point is found by taking the second derivative of the peak shape, identifying the point where the derivative changes.

[0013] In step three, the bimodal graph is segmented and integrated as follows: If a trailing peak or a preceding clutter peak appears in the bimodal graph, the trailing peak and the preceding clutter peak need to be removed; the inflection point B is found by calculation, and a point F is determined such that line segment AF is perpendicular to the Y-axis and line segment DF is perpendicular to the X-axis; then, the downward baseline integration is performed.

[0014] Ideally, in step four, the integral result obtained in step three is compared with the standard data recorded in the PDF. If the integral is larger than the standard data, CE is output; if the integral is smaller than the standard data, DF is output.

[0015] Compared with the prior art, the technical solution of the present invention has the following advantages / benefits: 1. After identifying the spectrum, by marking multiple points on the peak pattern, different types of peak patterns can be quickly distinguished, and different integration calculations can be performed according to different peak patterns. This not only saves calculation time but also improves the accuracy of integration.

[0016] 2. Because the present invention has strict specifications for the integration points and performs baseline correction for possible abnormal situations, it can avoid the calculation errors caused by the inaccurate determination of the integration curve in the prior art when performing integration calculations.

[0017] 3. Existing spectral integration methods cannot effectively deal with the presence of tailing peaks and leading clutter peaks in spectral curves. They lack the ability to segment and integrate, requiring manual segmentation and integration during testing and inspection. This relies on experience and lacks mathematical logic support, and the integration inflection point depends on human judgment. This invention automatically calculates the curve inflection point, accurately removes tailing peaks and leading clutter peaks, and obtains more accurate integration data. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating an AI spectrum recognition method according to the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention are described clearly and completely below. Obviously, the described embodiments are only a part of the embodiments of this invention, not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention. Therefore, the detailed description of the embodiments of this invention provided below is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the invention.

[0021] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it need not be further defined and explained in subsequent figures.

[0022] Example: National standards stipulate that all agricultural products entering the market must undergo inspection and testing to ensure food safety. The national standard also selects chromatography as the detection tool. During operation, chromatography instruments generate analytical spectra, which are crucial raw data required by the national standard, necessitating accurate integration and quantitative analysis. However, automatic spectrum integration by chromatography instruments cannot meet the needs of inspection and testing work, thus requiring manual operation. The large volume of pesticide and veterinary drug residue samples, the numerous compound tests, and the low work efficiency necessitate a large number of professional personnel, placing significant manpower pressure on the inspection and testing industry. Inspection and testing personnel often work overtime, and prolonged work can easily lead to fatigue-related errors. This can result in erroneous reports and legal risks for inspection and testing institutions.

[0023] To address the above problems, this invention proposes an AI-based spectral recognition method, such as... Figure 1 As shown, it includes the following steps: Step 1: Identify the pattern and convert the information contained in the data carrier into spectral information; the data carrier can include PDF, tables, internal instrument data, etc., as long as it can record spectral information, it can be considered a data carrier.

[0024] Step 2: Determine the peak type. Based on the collected spectrum information, determine whether the spectrum is a single-peak or multi-peak spectrum. Step 3: Perform integration. For unimodal graphs, perform baseline integration and tailing peak processing. For bimodal graphs, perform segmentation integration. Step 4: Compare the integrated results by area and output the final integrated result.

[0025] Through our long-term performance testing and investigation of existing instruments and working habits in the testing and inspection market, we have found that PDF is currently the primary format for generating spectral reports. However, the quality of PDFs varies between different instruments. Therefore, to address the diverse PDF formats, we currently employ both OCR visual recognition and document recognition methods to ensure the most complete extraction of various information from the PDF.

[0026] In step two, we first set the starting point of the peak plot to the integration starting point A, the ending point of the peak plot to the integration ending point B, the highest point of the peak plot to point C, the point G between points A and C to be the half-peak width point, the point E to be the highest point of the first adjacent peak, the point H to be the highest point of the second adjacent peak, the point I to be the highest point of the third adjacent peak, the point D to be the concave point between the two peaks, and the point F to be the calculated point when finding the baseline integration.

[0027] In step two, if only point C is found, the spectrum is classified as a single-peaked spectrum; if points C and E are found, the spectrum is classified as a double-peaked spectrum; if points C, E, and I are found, the spectrum is classified as a multi-peaked spectrum.

[0028] In step two, if the spectrum is determined to be a multi-peak spectrum, an error will be reported directly.

[0029] In step three, if you want to integrate a single-peak graph, you first need to find the starting point A and the ending point B of the integration. If the integration curve does not change, it is determined that there is no peak here. If both points A and B can be found, then the peak graph is integrated along the baseline. If only one of points A and B can be found, then the peak graph is processed to produce a tailing peak.

[0030] We refer to peaks with the same Y-value at the start and end points of integration, or peaks with a height difference between the start and end points of integration, as regular peaks. In this case, we only need to use an integration tool to integrate the peak.

[0031] If an irregular peak shape is encountered, the peak shape needs to be processed before integration.

[0032] Existing integration methods encounter the following problems when integrating irregular peak shapes: manual integration can lead to problems such as excessive detection time and errors due to worker fatigue, while model integration can easily result in inaccurate judgment of the peak shape to be integrated and inability to handle complex peak shapes, leading to data ambiguity.

[0033] Therefore, in step three, the baseline integration is as follows: First, find point F such that line segment AF is perpendicular to the Y-axis and line segment DF is perpendicular to the X-axis; then calculate the area of ​​the figure ACDF as the integration result.

[0034] The specific handling of tailing peaks in complex peak shapes is as follows: Find an inflection point on the peak shape, designated as point B, and determine a point F such that line segment AF is perpendicular to the Y-axis and line segment DF is perpendicular to the X-axis. Then perform a downward baseline integration; if there is a height difference between point A and point B, follow the principle of extending vertically from the high point to the low point.

[0035] The specific steps for segmenting and integrating bimodal graphs in complex peak shapes are as follows: If a trailing peak or a preceding clutter peak appears in the bimodal graph, the trailing peak and the preceding clutter peak need to be removed; the inflection point B is found through calculation, and a point F is determined such that line segment AF is perpendicular to the Y-axis and line segment DF is perpendicular to the X-axis; then, the downward baseline integration is performed.

[0036] Based on the different image features of the spectral curves, the curves are classified into regular and irregular peak shapes, and different integration schemes are formulated for each peak shape. Furthermore, instead of using a large model for calculation during the integration process, a fixed algorithm structure is employed to handle different situations that may be encountered during integration. This comprehensively covers all problems that may arise in spectral integration. It solves the technical problem in existing technologies where automatic integration using large models fails to accurately identify complex spectral curves, leading to inaccurate integration. It also addresses the issues of slow computation speed and the risk of calculation errors due to fatigue during long periods of manual integration.

[0037] Finally, in step four, the integral result obtained in step three is compared with the standard data recorded in the PDF. If it is larger than the area, CE is output; if it is smaller than the area, DF is output, thus completing the entire spectrum recognition process.

[0038] The above are merely preferred embodiments of the present invention. It should be noted that the above preferred embodiments should not be considered as limitations on the present invention, and the scope of protection of the present invention should be determined by the scope defined in the claims. For those skilled in the art, several improvements and modifications can be made without departing from the spirit and scope of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An AI spectrum recognition method, characterized in that, Includes the following steps: Step 1: Identify the pattern and convert the information contained in the data carrier into spectral information; Step 2: Determine the peak type. Based on the collected spectrum information, determine whether the spectrum is a single-peak or multi-peak spectrum. Step 3: Perform integration. For unimodal graphs, perform baseline integration and tailing peak processing. For bimodal graphs, perform segmentation integration. Step 4: Compare the integral results by area and output the final integral result.

2. The AI ​​spectrogram recognition method according to claim 1, characterized in that, In step two, the process of determining the peak shape is as follows: First, set the starting point of the peak shape graph as the integration starting point A, the ending point of the peak shape graph as the integration ending point B, the highest point of the peak shape graph as point C, the midpoint G between points A and C as the half-peak width point, point E as the highest point of the first adjacent peak, point H as the highest point of the second adjacent peak, point I as the highest point of the third adjacent peak, point D as the concave point between the two peaks, and point F as the calculation point when finding the baseline integration; then, check whether the above points appear in sequence, and determine the peak shape based on the points found.

3. The AI ​​spectrogram recognition method according to claim 2, characterized in that, In step two, if only point C is found when searching for the highest point, the spectrum is classified as a single-peaked spectrum; if points C and E are found, the spectrum is classified as a double-peaked spectrum; if points C, E, and I are found, the spectrum is classified as a multi-peaked spectrum.

4. The AI ​​spectrogram recognition method according to claim 3, characterized in that, In step two, if the spectrum is determined to be a multi-peak spectrum, an error is reported directly, and the process returns to step one.

5. The AI ​​spectrum recognition method according to claim 3, characterized in that, In step three, if the spectrum is determined to be a single-peaked spectrum, then the single-peaked spectrum is integrated; if the spectrum is determined to be a double-peaked spectrum, then the double-peaked spectrum is segmented and integrated.

6. The AI ​​spectrum recognition method according to claim 5, characterized in that, In step three, when integrating the single-peak graph, it is first necessary to find the starting point A and the ending point B of the integration. If the integration curve does not change, it is determined that there is no peak here. If both points A and B can be found, the peak graph is integrated along the baseline. If only one of points A and B can be found, the peak graph is processed to produce a tailing peak.

7. The AI ​​spectrum recognition method according to claim 6, characterized in that, In step three, the baseline integration is as follows: First, find point F such that line segment AF is perpendicular to the Y-axis and line segment DF is perpendicular to the X-axis; then calculate the area of ​​the figure ACDF as the integration result.

8. The AI ​​spectrum recognition method according to claim 6, characterized in that, The tailing peak processing is as follows: Find an inflection point on the peak shape, designated as point B, and determine a point F such that line segment AF is perpendicular to the Y-axis and line segment DF is perpendicular to the X-axis; then perform a downward baseline integration; if there is a height difference between point A and point B, follow the principle of vertical extension from the high point to the low point; the inflection point is found by taking the second derivative of the peak shape, and this point is the inflection point.

9. The AI ​​spectrum recognition method according to claim 5, characterized in that, In step three, the bimodal graph is segmented and integrated as follows: If a trailing peak or a preceding clutter peak appears in the bimodal graph, the trailing peak and the preceding clutter peak need to be removed; the inflection point B is found by calculation, and a point F is determined such that line segment AF is perpendicular to the Y-axis and line segment DF is perpendicular to the X-axis; then, the downward baseline integration is performed.

10. The AI ​​spectrogram recognition method according to claim 1, characterized in that, In step four, the integral result obtained in step three is compared with the standard data recorded in the PDF. If the integral is larger than the standard data, CE is output; if the integral is smaller than the standard data, DF is output.

Images