Coal quality detection method and device based on multispectral fusion and working condition self-correction

Abstract

The invention provides a coal quality detection method and device based on multispectral fusion and working condition self-correction. The method comprises the following steps: acquiring first spectral data and a first image of a coal sample, real-time surface temperature of the coal sample and second spectral data and a second image of an empty belt; the first spectral data and the second spectral data are preprocessed, and coal sample spectral features and empty belt spectral features are extracted based on the processed first spectral data and second spectral data; performing feature extraction on the first image and the second image to obtain coal sample particle features; and inputting the coal sample surface real-time temperature, the coal sample spectral characteristics, the empty belt spectral characteristics and the coal sample particle characteristics into a pre-trained coal quality prediction model to obtain a coal quality detection result. According to the invention, problems in the prior art can be effectively improved, and rapid and accurate detection of coal quality is realized.

Description

Technical Field

[0001] This invention relates to the field of coal quality testing technology, and in particular to a coal quality testing method and apparatus based on multispectral fusion and operating condition self-calibration. Background Technology

[0002] Currently, the mainstream technologies for real-time online coal quality analysis mainly include single detection methods such as near-infrared spectroscopy, X-ray fluorescence spectroscopy, neutron activation analysis, and laser-induced breakdown spectroscopy, as well as several simple combined techniques. Near-infrared spectroscopy is sensitive to organic functional groups in coal (such as CH bonds and OH bonds), and is suitable for predicting moisture, volatile matter, and some calorific value indicators, but its detection capability for inorganic components is limited. X-ray fluorescence spectroscopy has high sensitivity to inorganic elements in coal (such as Si, Al, Fe, Ca, and S), and is a key method for directly determining ash content, ash composition, and sulfur content. While neutron activation analysis can detect most elements, its application in industrial settings is limited due to the complexity and high cost of the equipment, as well as radioactivity management issues. Laser-induced breakdown spectroscopy shows potential in elemental detection, but still faces challenges such as matrix effects and quantitative stability.

[0003] Current common combined solutions are mostly limited to the simple fusion of data such as near-infrared spectroscopy, X-ray fluorescence spectroscopy, and temperature monitoring, and the construction of coal quality prediction models using machine learning or deep learning algorithms, attempting to achieve simultaneous online analysis of multiple indicators. However, when dealing with real-world industrial scenarios such as diverse coal types, complex operating conditions, and significant physical disturbances, these methods often suffer from significant deficiencies in model robustness, accuracy, and long-term stability, and have not yet formed a systematic, adaptive, integrated solution. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a coal quality detection method and device based on multispectral fusion and working condition self-correction, which can effectively improve the problems existing in the prior art and realize rapid and accurate detection of coal quality.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides a coal quality detection method based on multispectral fusion and working condition self-correction, comprising: acquiring first spectral data, a first image, and real-time surface temperature of a coal sample, as well as second spectral data and a second image of an empty conveyor belt; preprocessing the first and second spectral data, and extracting spectral features of the coal sample and the empty conveyor belt based on the processed first and second spectral data; extracting features from the first and second images to obtain coal sample particle features; and inputting the real-time surface temperature of the coal sample, the spectral features of the coal sample, the spectral features of the empty conveyor belt, and the coal sample particle features into a pre-trained coal quality prediction model to obtain coal quality detection results.

[0006] Optionally, the first spectral data includes: a first near-infrared spectrum and a first X-ray fluorescence spectrum; the second spectral data includes: a second near-infrared spectrum and a second X-ray fluorescence spectrum; the acquisition of the first spectral data, a first image, and the real-time surface temperature of the coal sample, as well as the second spectral data and a second image of the empty conveyor belt, includes: detecting the first distance information from the target location to the coal conveyor belt using a laser rangefinder; if the first distance information is less than a preset value, it is determined that a coal sample has passed through the conveyor belt, and the first near-infrared spectrum, the first X-ray fluorescence spectrum, the first image, and the real-time surface temperature of the coal sample are acquired; after the coal sample is collected, the second near-infrared spectrum, the second X-ray fluorescence spectrum, and the second image of the empty conveyor belt are acquired.

[0007] Optionally, the first and second spectral data are preprocessed, including: processing the first and second near-infrared spectra respectively to obtain the second derivative spectrum of the coal sample and the second derivative spectrum of the empty conveyor belt; wherein, the near-infrared spectrum is one-dimensional data, the first derivative spectrum is the spectrum obtained by taking the first derivative of the near-infrared spectrum, and the second derivative spectrum is the spectrum obtained by taking the derivative of the first derivative spectrum; processing the first and second X-ray fluorescence spectra respectively to obtain the characteristic peak spectrum of coal, the background spectrum of coal, the background spectrum of the empty conveyor belt, and the characteristic peak spectrum of the empty conveyor belt.

[0008] Optionally, based on the processed first and second spectral data, spectral features of the coal sample and the empty conveyor belt are extracted, including: extracting functional group absorption intensity features of the coal sample based on the second derivative spectrum of the coal sample, and extracting functional group absorption intensity features of the empty conveyor belt based on the second derivative spectrum of the empty conveyor belt; calculating elemental intensity features of the coal based on the characteristic peak spectrum of the coal sample, and calculating background spectral intensity features of the coal based on the background spectrum of the coal sample; calculating elemental intensity features of the empty conveyor belt based on the characteristic peak spectrum of the empty conveyor belt, and calculating background spectral intensity features of the empty conveyor belt based on the background spectrum of the empty conveyor belt; and calculating coal correlation features based on the functional group absorption intensity features and elemental intensity features of the coal sample.

[0009] Optionally, the first near-infrared spectrum and the second near-infrared spectrum are processed to obtain the second derivative spectrum of the coal sample and the second derivative spectrum of the empty conveyor belt, including: filtering the first near-infrared spectrum and the second near-infrared spectrum to obtain the first smoothed near-infrared spectrum and the second smoothed near-infrared spectrum; calculating the first total absorbance of the first smoothed near-infrared spectrum and the second total absorbance of the second smoothed near-infrared spectrum, and removing abnormal spectra in the first smoothed near-infrared spectrum based on the first total absorbance to obtain the first target spectrum, and removing abnormal spectra in the second smoothed near-infrared spectrum based on the second total absorbance to obtain the second target spectrum; performing scattering correction on the first target spectrum and the second target spectrum to obtain the first corrected spectrum and the second corrected spectrum; performing derivative processing on the first corrected spectrum and the second corrected spectrum to obtain the second derivative spectrum of the coal sample and the second derivative spectrum of the empty conveyor belt, and calculating the peak area of ​​a preset wavelength range based on the second derivative spectrum of the coal sample and the second derivative spectrum of the empty conveyor belt.

[0010] Optionally, the first X-ray fluorescence spectrum and the second X-ray fluorescence spectrum are processed separately to obtain the coal characteristic peak spectrum, the coal background spectrum, the empty conveyor belt background spectrum, and the empty conveyor belt characteristic peak spectrum, including: filtering the first X-ray fluorescence spectrum and the second X-ray fluorescence spectrum to obtain the first X-ray fluorescence smoothed spectrum and the second X-ray fluorescence smoothed spectrum; performing drift correction on the first X-ray fluorescence smoothed spectrum and the second X-ray fluorescence smoothed spectrum based on the standard coal sample spectrum to obtain the first fluorescence corrected spectrum and the second fluorescence corrected spectrum; extracting the coal characteristic peak spectrum and the coal background spectrum based on the first fluorescence corrected spectrum; and extracting the empty conveyor belt background spectrum and the empty conveyor belt characteristic peak spectrum based on the second fluorescence corrected spectrum.

[0011] Optionally, feature extraction is performed on the first image and the second image to obtain coal sample particle features, including: performing median filtering on the second image to obtain a reference empty conveyor belt background image; performing difference between the first image and the reference empty conveyor belt background image to obtain a difference image; performing adaptive histogram equalization on the difference image to obtain an enhanced image; performing global threshold segmentation on the enhanced image to obtain a binary image; performing opening and closing operations on the binary image to obtain a processed binary image; marking the particle regions in the processed binary image, and obtaining coal sample particle features based on each particle region.

[0012] Secondly, the present invention provides a coal quality testing device based on multispectral fusion and working condition self-correction, comprising: a data acquisition module for acquiring first spectral data, a first image, and real-time surface temperature of a coal sample, as well as second spectral data and a second image of an empty conveyor belt; a first feature extraction module for preprocessing the first and second spectral data, and extracting spectral features of the coal sample and the empty conveyor belt based on the processed first and second spectral data; a second feature extraction module for extracting features from the first and second images to obtain coal sample particle features; and a coal quality testing module for inputting the real-time surface temperature of the coal sample, the spectral features of the coal sample, the spectral features of the empty conveyor belt, and the particle features of the coal sample into a pre-trained coal quality prediction model to obtain coal quality testing results.

[0013] Thirdly, the present invention provides an electronic device including a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the steps of the method provided in any of the first aspects above.

[0014] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, performs the steps of the method provided in any of the first aspects above.

[0015] This invention brings the following beneficial effects: The coal quality detection method and apparatus based on multispectral fusion and working condition self-correction provided by the present invention first acquires the first spectral data, first image, and real-time surface temperature of the coal sample, as well as the second spectral data and second image of the empty conveyor belt; then, the first and second spectral data are preprocessed, and the spectral features of the coal sample and the empty conveyor belt are extracted based on the processed first and second spectral data; next, feature extraction is performed on the first and second images to obtain the coal sample particle features; finally, the real-time surface temperature of the coal sample, the spectral features of the coal sample, the spectral features of the empty conveyor belt, and the coal sample particle features are input into a pre-trained coal quality prediction model to obtain the coal quality detection results. The above method can acquire the real-time surface temperature of the coal sample and identify the working condition based on the real-time surface temperature; by acquiring the first spectral data of the coal sample and the second spectral data of the empty conveyor belt and extracting features, the spectral features of the empty conveyor belt can be used to correct the spectral features of the coal sample during coal quality testing, thereby improving the accuracy of coal sample spectral feature detection; by acquiring the first image of the coal sample and the second image of the empty conveyor belt and extracting features to obtain the particle features of the coal sample, and combining the real-time surface temperature of the coal sample, the spectral features of the coal sample, the spectral features of the empty conveyor belt, and the particle features of the coal sample for coal quality testing, it is possible to simultaneously detect indicators such as coal ash content, moisture, volatile matter, and sulfur content through deep fusion of multi-source information and adaptive correction of working conditions, effectively improving the problems existing in the prior art and improving the speed and accuracy of coal quality testing.

[0016] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a coal quality detection method based on multispectral fusion and operating condition self-calibration provided in an embodiment of the present invention; Figure 2 A near-infrared spectrum diagram of a coal sample provided in an embodiment of the present invention; Figure 3 An example X-ray fluorescence spectrum of a coal sample provided in an embodiment of the present invention; Figure 4 An example X-ray fluorescence spectrum of an empty belt is provided as an embodiment of the present invention; Figure 5 A schematic diagram of a coal quality detection method based on multispectral fusion and working condition self-correction provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Currently, most common coal quality testing solutions are limited to simply fusing data from near-infrared spectroscopy, X-ray fluorescence spectroscopy, and temperature monitoring, and then using machine learning or deep learning algorithms to build coal quality prediction models, attempting to achieve simultaneous online analysis of multiple indicators. However, these methods have the following drawbacks, adversely affecting the accuracy and reliability of coal quality analysis.

[0022] (1) Temperature interference: In actual industrial settings, temperature changes can significantly affect the results of spectral measurements, thereby reducing the accuracy of the measurements. Large differences in ambient temperature and the increasing temperature of equipment over time can lead to baseline drift and peak shift in near-infrared (NIR) spectra, as well as varying degrees of increase or decrease in intensity in different energy ranges of X-ray fluorescence (XRF) spectra, thus causing errors in on-site measurements.

[0023] (2) Particle size effect: Changes in the size and distribution of coal powder particles alter light scattering characteristics, affecting both NIR and XRF signal intensity and introducing significant measurement errors. The size and distribution of coal powder particles directly affect light scattering characteristics, which in turn affects the signal intensity of near-infrared and X-ray fluorescence spectra, resulting in large errors in the measurement results.

[0024] (3) Regional (coal quality) differences: Coal from different regions varies greatly in terms of organic matter type and mineral composition. Existing single models are difficult to adapt to these changes and need to be recalibrated frequently, which increases the cost and difficulty of use.

[0025] (4) Large sample size requirement: Most of the current fusion technologies use deep learning, which is based on full-spectrum data. During the learning process, a large amount of full-spectrum data needs to be processed, resulting in a very large number of features and learning parameters. Therefore, a large number of samples are needed for training, which increases the workload of data collection and processing.

[0026] (5) High environmental requirements: In the actual production process, due to the continuity of production, the equipment cannot be cleaned and maintained in a timely manner, which will cause the window to gradually become dirty. The spectra of pollutants will be superimposed on the spectra of coal samples, affecting the accuracy of measurement results. Therefore, it is necessary to process the pollution data reasonably.

[0027] Based on this, the coal quality detection method and device based on multispectral fusion and working condition self-correction provided by the embodiments of the present invention can effectively improve the problems existing in the prior art and realize rapid and accurate detection of coal quality.

[0028] To facilitate understanding of this embodiment, a detailed description of a coal quality detection method based on multispectral fusion and operating condition self-correction disclosed in this invention will be provided first. This method can be executed by electronic devices, such as smartphones, computers, and tablets. See also... Figure 1 The flowchart shown illustrates a coal quality detection method based on multispectral fusion and operating condition self-correction, indicating that the method mainly includes the following steps S101 to S104: Step S101: Collect the first spectral data, first image, and real-time surface temperature of the coal sample, as well as the second spectral data and second image of the empty conveyor belt.

[0029] In one embodiment, the first spectral data includes: a first near-infrared spectrum and a first X-ray fluorescence spectrum; the second spectral data includes: a second near-infrared spectrum and a second X-ray fluorescence spectrum.

[0030] In the implementation of this invention, when collecting the first spectral data, first image, and real-time temperature of the coal sample surface, as well as the second spectral data and second image of the empty conveyor belt, the following methods may be used, including but not limited to: First, the first distance information from the target position to the coal conveyor belt is detected by a laser rangefinder; if the first distance information is less than a preset value, it is determined that a coal sample has passed through the coal conveyor belt, and the first near-infrared spectrum, first X-ray fluorescence spectrum, first image, and real-time temperature of the coal sample surface are collected; then, after the coal sample collection is completed, the second near-infrared spectrum, second X-ray fluorescence spectrum, and second image of the empty conveyor belt are collected.

[0031] In practice, the distance measured by the laser rangefinder is used to determine whether a coal sample has passed through. If the first distance information from the target location to the coal conveyor belt is less than the preset value, it is determined that a coal sample has passed through. The X-ray fluorescence instrument and the near-infrared instrument are then activated to perform spectral scanning. The first near-infrared spectrum, the first X-ray fluorescence spectrum, the real-time surface temperature of the coal sample (measured by non-contact infrared thermometry), and the first image (i.e., the apparent image, which is acquired by an industrial camera and used to assess particle size and distribution) of the coal sample are collected simultaneously on the coal conveyor belt or the sampling path.

[0032] The coal sample collection process is completed based on the distance measured by the laser rangefinder. After a waiting period (generally around 10 seconds), the second near-infrared spectrum, second X-ray fluorescence spectrum, and second image of the empty conveyor belt are acquired. The conveyor belt needs to maintain a consistent coal sample height for at least 10 seconds; therefore, each detection requires at least approximately 20 seconds. In this embodiment of the invention, by simultaneously acquiring multiple data points on the coal conveyor belt or sampling path, a comprehensive understanding of the coal sample's characteristics and operating conditions can be obtained, providing rich data support for subsequent analysis.

[0033] Step S102: Preprocess the first spectral data and the second spectral data, and extract the spectral features of the coal sample and the empty conveyor belt based on the processed first spectral data and the second spectral data.

[0034] In one implementation, the acquired first and second spectral data are further processed to improve data quality and usability. Specifically, the first and second spectral data are processed using the same steps to ensure consistency and comparability of the processing results. These steps include, but are not limited to, spectral smoothing, outlier removal, scattering correction, and derivative processing. Furthermore, feature extraction is performed on the processed first spectral data to obtain coal sample spectral features, and feature extraction is performed on the processed second spectral data to obtain empty conveyor belt spectral features.

[0035] Step S103: Extract features from the first and second images to obtain the particle features of the coal sample.

[0036] In one embodiment, the first image is corrected using the second image, and then features are extracted from the corrected image to obtain the particle area. A p Particle count N P Average particle size d avg Flatness F PThe characteristics of coal sample particles are analyzed. Specifically, the second image is first filtered to obtain a baseline empty conveyor belt background image. Then, the first image and the baseline empty conveyor belt background image are differentially analyzed. Finally, the differential image is enhanced, subjected to global thresholding, and opened and closed operations to obtain the corrected image.

[0037] Step S104: Input the real-time surface temperature of the coal sample, the spectral characteristics of the coal sample, the spectral characteristics of the empty belt, and the particle characteristics of the coal sample into the pre-trained coal quality prediction model to obtain the coal quality detection results.

[0038] In one embodiment, the spectral characteristics of the coal sample include at least: functional group absorption intensity characteristics, elemental intensity characteristics, background spectral intensity characteristics, and correlation characteristics; the spectral characteristics of the empty conveyor belt include at least: functional group absorption intensity characteristics, elemental intensity characteristics, and background spectral intensity characteristics.

[0039] In this embodiment of the invention, a feature vector containing various chemical and physical information can be constructed in advance through the extraction of the aforementioned features, providing rich data support for coal quality modeling. Machine learning algorithms, such as Partial Least Squares Regression (PLS), Support Vector Machine (SVM), Random Forest (RF), or Neural Network (NN), are then used to train and learn the extracted features, establishing an accurate coal quality prediction model to achieve rapid and accurate prediction of coal quality. After the model is established, it undergoes cross-validation and testing to evaluate its accuracy and generalization ability. Cross-validation and testing ensure the model's performance on unknown data, improving its reliability and practicality. Finally, the established coal quality detection model is deployed to an online detection system to analyze and predict coal quality in real time, providing real-time coal quality information for the production process and improving production efficiency and product quality.

[0040] Based on this, in this embodiment of the invention, the real-time surface temperature of the coal sample, the spectral characteristics of the coal sample, the spectral characteristics of the empty conveyor belt, and the particle characteristics of the coal sample can be input into a pre-trained coal quality prediction model to obtain coal quality detection results. The coal quality detection results include at least the following coal quality indicators: ash content, sulfur content, moisture content, and volatile matter.

[0041] The coal quality detection method based on multispectral fusion and working condition self-correction provided in this invention can acquire the real-time surface temperature of coal samples and identify working conditions based on the real-time surface temperature. By acquiring the first spectral data of the coal sample and the second spectral data of the empty conveyor belt and extracting features, the spectral features of the empty conveyor belt can be used to correct the spectral features of the coal sample during coal quality detection, thereby improving the accuracy of coal sample spectral feature detection. By acquiring the first image of the coal sample and the second image of the empty conveyor belt and extracting features to obtain the particle features of the coal sample, and combining the real-time surface temperature of the coal sample, the spectral features of the coal sample, the spectral features of the empty conveyor belt, and the particle features of the coal sample for coal quality detection, it is possible to simultaneously detect indicators such as coal ash, moisture, volatile matter, and sulfur through deep fusion of multi-source information and adaptive working condition correction, effectively improving the problems existing in the prior art and improving the speed and accuracy of coal quality detection.

[0042] In one implementation, for the aforementioned step S102, i.e., when preprocessing the first spectral data and the second spectral data, the following methods may be used, including but not limited to: (1) The first near-infrared spectrum and the second near-infrared spectrum were processed respectively to obtain the second derivative spectrum of the coal sample and the second derivative spectrum of the empty belt.

[0043] In practice, the first and second near-infrared spectra are first filtered to obtain the first and second near-infrared smoothed spectra.

[0044] For details, see Figure 2 The diagram shows a near-infrared spectrum of a coal sample. Savitzky-Golay filtering is used to filter the first near-infrared spectrum to obtain a smoothed first near-infrared spectrum. Similarly, Savitzky-Golay filtering is used to filter the second near-infrared spectrum to obtain a smoothed second near-infrared spectrum. In this embodiment of the invention, the Savitzky-Golay filtering method can effectively remove noise from the spectrum, resulting in a smooth spectrum.

[0045] Then, the first total absorbance of the first near-infrared smoothed spectrum and the second total absorbance of the second near-infrared smoothed spectrum are calculated respectively. Based on the first total absorbance, abnormal spectra in the first near-infrared smoothed spectrum are removed to obtain the first target spectrum. Based on the second total absorbance, abnormal spectra in the second near-infrared smoothed spectrum are removed to obtain the second target spectrum.

[0046] Specifically, abnormal spectrum removal involves eliminating abnormal spectra caused by insufficient distance or other reasons. For the first near-infrared smoothed spectrum of a coal sample, the first total absorbance of the first near-infrared smoothed spectrum is calculated, and samples with abnormally low absorbance are removed using a percentage method; the remaining spectra are considered normal. Assuming there are N samples, each corresponding to a near-infrared spectral curve, and each spectrum containing absorbance measurements within a wavelength range (e.g., 900-2500 nm), for each sample, all absorbance values ​​within the entire effective wavelength range are summed (or approximated by integration) to obtain the first total absorbance of that sample. Then, the first total absorbance values ​​of all samples are organized into a one-dimensional list, and this sequence is sorted from smallest to largest to obtain an ordered sequence. Next, the proportion of abnormally low absorbance samples to be removed is determined, for example, 5%, and the number of samples to be removed is calculated. K =N×5%, remove the original samples corresponding to the first K minimum values ​​from the sorted sequence, and the remaining ones are the normal spectrum, i.e. the first target spectrum.

[0047] For the second near-infrared smoothed spectrum of the empty belt, the first total absorbance of the first near-infrared smoothed spectrum is first calculated, and samples with abnormally low absorbance are eliminated using a percentage method. The remaining samples are considered normal spectra, i.e., the second target spectrum. The specific elimination process is the same as the abnormal spectrum elimination process for the first near-infrared smoothed spectrum, and will not be repeated here.

[0048] In this embodiment of the invention, after removing abnormal spectra, the average total intensity of the normal spectrum is compared with the average over a historical period. If the average total intensity of the normal spectrum is less than a certain threshold (which needs to be determined based on actual production conditions), the overall intensity of the coal sample is considered low, the test fails, and no result is issued. Otherwise, the normal spectrum is retained for further processing. In practice, the IQR method can generally be used to determine the threshold for total intensity, ensuring that the coal sample intensity is normal in most cases. Specifically, the average total intensity over a historical period is obtained, and the quartiles are calculated: Q1 = 25th percentile (lower quartile), Q3 = 75th percentile (upper quartile); then IQR = Q3 is calculated. Q1, and determine the threshold for the total average intensity as: T=Q1 k×IQR, where k can be 1.5.

[0049] In this embodiment of the invention, abnormal spectra can be effectively identified and eliminated by calculating the total absorbance and using a percentage method to remove samples with abnormally low absorbance. Eliminating abnormal spectra improves the quality of spectral data and avoids the impact of abnormal data on subsequent analysis results.

[0050] Next, scattering corrections were performed on the first target spectrum and the second target spectrum respectively to obtain the first corrected spectrum and the second corrected spectrum.

[0051] Specifically, multivariate scattering correction corrects for baseline shifts and tilts caused by changes in physical conditions, mitigating distance variations to some extent. Furthermore, it can also overcome some differences caused by varying particle sizes. Taking the spectrum of the first target as an example, the scattering correction method is as follows: Step 1: Calculate the average spectrum of the first target spectrum.

[0052] Assume there is n There are 10 samples, and each sample is in P The absorbance value is measured at the nth wavelength point (or wavenumber point, channel), then the nth wavelength point (or wavenumber point, channel) is the absorbance value measured at the nth wavelength point (or wavenumber point, channel). i The sample at the th j The absorbance at each wavelength point is The average spectrum is then:

[0053] in, 1 P spectral vector.

[0054] Step 2: For each spectrum Perform univariate linear regression: ,in, The intercept is... The slope This is the residual.

[0055] Solve and : , .

[0056] in, For the first i The sample at the th j Absorbance at each wavelength point The average spectrum at the 1st j Absorbance at each wavelength point Average spectrum The average absorbance at all wavelengths For the spectrum The average absorbance at all wavelengths.

[0057] Step 3: Correct each spectrum:

[0058] Step 4: Average all the corrected spectral spheres to obtain the final sum spectrum, i.e., the first corrected spectrum.

[0059] It should be noted that the scattering correction process for the second target spectrum is the same as the scattering correction process described above, and will not be repeated here.

[0060] In this embodiment of the invention, the multivariate scattering correction method can effectively correct baseline shifts and tilts caused by changes in physical conditions, reducing the impact of distance variations and different particle sizes on the spectrum.

[0061] Finally, derivative processing was performed on the first and second corrected spectra to obtain the second derivative spectra of the coal sample and the second derivative spectra of the empty conveyor belt. The peak area of ​​the preset wavelength range was then calculated based on the second derivative spectra of the coal sample and the second derivative spectra of the empty conveyor belt.

[0062] In this study, the near-infrared spectrum is one-dimensional data. The first derivative spectrum is obtained by taking the first derivative of the near-infrared spectrum, and the second derivative spectrum is obtained by taking the derivative of the first derivative spectrum. Specifically, the first derivative of the first correction spectrum is first processed to obtain the first derivative spectrum of the coal sample; then, the derivative of the first derivative spectrum of the coal sample is taken to obtain the second derivative spectrum of the coal sample. Similarly, the first derivative of the second correction spectrum is processed to obtain the first derivative spectrum of the empty conveyor belt; then, the derivative of the first derivative spectrum of the empty conveyor belt is taken to obtain the second derivative spectrum of the empty conveyor belt. Furthermore, the peak areas of the corresponding wavelength ranges are extracted from the second derivative spectra of the coal sample and the second derivative spectra of the empty conveyor belt, and the peak areas are normalized.

[0063] (2) The first X-ray fluorescence spectrum and the second X-ray fluorescence spectrum were processed respectively to obtain the characteristic peak spectrum of coal, the background spectrum of coal, the background spectrum of empty conveyor belt and the characteristic peak spectrum of empty conveyor belt.

[0064] In practice, the first and second X-ray fluorescence spectra are first filtered to obtain smoothed X-ray fluorescence spectra and smoothed X-ray fluorescence spectra respectively.

[0065] Specifically, Savitzky-Golay filtering was used to filter the first X-ray fluorescence spectrum and the second X-ray fluorescence spectrum respectively to obtain the smoothed first X-ray fluorescence spectrum and the smoothed second X-ray fluorescence spectrum.

[0066] Then, based on the standard coal sample spectrum, the first X-ray fluorescence smoothed spectrum and the second X-ray fluorescence smoothed spectrum were drift corrected to obtain the first fluorescence corrected spectrum and the second fluorescence corrected spectrum.

[0067] Specifically, the first and second smoothed X-ray fluorescence spectra are drift-corrected and determined using the standard coal sample spectrum. If drift occurs, it is corrected to obtain the first and second corrected fluorescence spectra. The standard coal sample spectrum is the coal sample spectrum used when building the equipment model. The X-ray fluorescence spectra of the standard coal sample (i.e., coal sample prepared according to certification standards) are pre-acquired, and the X-ray fluorescence spectra of the standard coal sample are pre-processed (e.g., background subtraction, noise smoothing, characteristic peak identification, and peak area calculation) to obtain the standard coal sample spectrum.

[0068] Finally, based on the first fluorescence correction spectrum, the characteristic peak spectrum and background spectrum of coal were extracted, and based on the second fluorescence correction spectrum, the background spectrum and characteristic peak spectrum of empty conveyor belt were extracted.

[0069] For details, see Figure 3 An example X-ray fluorescence spectrum of a coal sample is shown, and Figure 4 The image shown is an example of an empty belt X-ray fluorescence spectrum. Fluorescence spectra typically include true fluorescence characteristic peaks (generated by specific fluorescent substances in the sample, such as polycyclic aromatic hydrocarbons, humic acids, etc., appearing as sharp or broad peaks) and background signals (originating from Rayleigh / Raman scattering, instrument dark current, solvent Raman peaks, stray light, or nonspecific broadband fluorescence, appearing as slowly changing baseline drift).

[0070] In this embodiment of the invention, the first fluorescence correction spectrum is subjected to background subtraction to obtain the coal characteristic peak spectrum and the coal background spectrum, and the second fluorescence correction spectrum is subjected to background subtraction to obtain the empty conveyor belt background spectrum and the empty conveyor belt characteristic peak spectrum.

[0071] Taking the first fluorescence correction spectrum of a coal sample as an example, observe the spectrum and identify several wavelength ranges that clearly do not contain characteristic peaks. Select a series of anchor points within these regions and record their wavelengths and corresponding intensity values. Then, use a smoothing function (such as cubic spline, polynomial, or locally weighted regression LOESS) to fit a continuous, smooth curve through the anchor points, which is the estimated coal background spectrum. Next, subtract the corresponding background estimate from the original first fluorescence correction spectrum at each wavelength point to obtain the coal characteristic peak spectrum. Further, after obtaining the coal characteristic peak spectrum and the coal background spectrum, overlapping peaks can be processed, and the area intensity of each characteristic peak can be calculated. It should be noted that the processing method for the second fluorescence correction spectrum is the same and will not be repeated here.

[0072] In one embodiment, for the aforementioned step S102, i.e., when extracting the spectral features of the coal sample and the spectral features of the empty conveyor belt based on the processed first and second spectral data, the following methods may be used, including but not limited to: First, the functional group absorption intensity characteristics of coal samples were extracted based on the second derivative spectrum of coal samples, and the functional group absorption intensity characteristics of empty conveyor belts were extracted based on the second derivative spectrum of empty conveyor belts.

[0073] In specific implementation, the absorption intensity characteristics of specific functional groups are extracted from the second derivative spectrum of the coal sample, including but not limited to the following characteristics: R_OH1, R_OH2, R_CH1, R_CH2, R_CH3, and S_1200_1400. Specifically, R_OH1 is the first-order overtone absorption intensity of OH at 1450±20 nm (main moisture absorption band); R_OH2 is the combination overtone absorption intensity of OH at 1940±30 nm (secondary moisture absorption band); R_CH1 is the second-order overtone absorption intensity of CH at 1200±30 nm (aliphatic hydrocarbon characteristic); R_CH2 is the first-order overtone absorption intensity of CH at 1700±30 nm (aromatic hydrocarbon characteristic); R_CH3 is the combination overtone absorption intensity of CH at 2300±40 nm; and S_1200_1400 is the spectral slope in the 1200-1400 nm range.

[0074] Meanwhile, the same functional group absorption intensity feature, Re_CH1, was extracted from the second derivative spectrum of the empty belt to characterize the pollution status.

[0075] In this embodiment of the invention, the extracted absorption intensity characteristics of specific functional groups can effectively characterize the distribution and relative abundance of organic functional groups in coal. For example, the absorption peak intensity of OH stretching vibration at approximately 1940 nm and 1450 nm directly reflects the moisture content in coal; while the CH bond absorption characteristics near 1200 nm, 1700 nm, and 2300 nm are closely related to the organic matter composition of coal and are key indicators for assessing volatile matter and predicting calorific value. Simultaneously, extracting the spectral slope of these characteristic bands can further reflect the chemical structural characteristics and maturity differences of coal, providing sensitive spectral fingerprint information for distinguishing coal types and identifying coal quality changes. This analytical method transforms complex spectral signals into characteristic parameters with clear chemical meaning, laying a reliable foundation for subsequent quantitative modeling and online real-time analysis.

[0076] Then, the elemental intensity characteristics of coal are calculated based on the characteristic peak spectra of coal, and the background spectral intensity characteristics of coal are calculated based on the background spectra of coal.

[0077] In practical implementation, the elemental intensity characteristics of coal are calculated based on the characteristic spectra of coal, including but not limited to: I Si I Al I Ca I Fe I Ti I SElemental characteristics. Specifically, based on atomic spectral databases (such as NIST), the main sensitive spectral line wavelengths of key elements closely related to coal quality are determined; then, for each target element, the characteristic wavelengths are... A narrow wavelength window (e.g., ±0.1 nm or ±0.2 nm) is set as the analysis channel for this element. Simultaneously, a neighboring region without emission peaks is selected near this window as the local background. For each analysis window, the average intensity of the background regions on both sides of the window is calculated. Extract the original intensity at the characteristic wavelength (or the maximum value within the window). And calculate the net intensity as the intensity characteristic of the element: Next, the intensity characteristics of the element are normalized to obtain the final intensity characteristics of the element.

[0078] Furthermore, the background spectral intensity I is calculated based on the background spectrum of coal. Back Specifically, for each selected background window, the spectral intensity values ​​of all data points within that band are extracted, and their average intensity is calculated. The average intensity of all background windows is then averaged again to obtain the background spectral intensity I. Back .

[0079] In this embodiment of the invention, by extracting the elemental intensities and Compton scattering peak intensities from the characteristic spectra of coal, the chemical composition information of coal can be obtained. These components are important components of coal ash and are crucial for coal quality analysis.

[0080] Next, the elemental intensity characteristics of the empty belt are calculated based on the characteristic peak spectra of the empty belt, and the background spectral intensity characteristics of the empty belt are calculated based on the background spectra of the empty belt.

[0081] In practical implementation, the elemental intensity characteristics I of the empty belt are extracted from the characteristic peak spectrum of the empty belt. eSi And extracting the background spectral intensity features of the empty belt from the background spectrum of the empty belt. eBack The background intensity and elemental intensity of the empty belt can be used as a reference, reflecting the degree of contamination of the window and the degree of environmental change. For example, the dirtier the belt, the greater the Si elemental intensity. Temperature also affects the overall intensity and can be used for subsequent data correction and analysis.

[0082] Finally, based on the functional group absorption intensity characteristics and elemental intensity characteristics of the coal samples, the correlation characteristics of the coal are calculated.

[0083] In practice, the functional group absorption intensity characteristics and elemental intensity characteristics of coal samples are used to calculate coal correlation characteristics, including but not limited to: (1) Characteristics related to moisture resistance to ash interference: F_OH_Si=R_OH1 / (I Si +ε); F_OH_Ai=R_OH1 / (I Ai +ε).

[0084] (2) Volatile matter resistance to moisture interference: F_C_H=R_CH2 / (R_OH1+R_OH2+ε).

[0085] (3) Coal quality characteristics: F_S=I S / (I Fe +I Ca +ε); F_SiAl_CaFe=(I Si +I Al ) / (I Ca +I Fe +ε).

[0086] In one embodiment, for the aforementioned step S103, i.e., when extracting features from the first image and the second image to obtain the coal sample particle features, the following methods, including but not limited to, can be used, mainly including the following steps 1 to 6: Step 1: Perform median filtering on the second image to obtain the baseline empty conveyor belt background image. In specific implementation, pixel-level median filtering is performed on the second image (i.e., the empty conveyor belt image) acquired during the empty conveyor belt acquisition period to obtain the baseline empty conveyor belt background image. Specifically, firstly, a sliding window of a preset size (e.g., 3×3) is used as the filtering window, with the center of the window aligned with the pixel position of the second image to be processed; then, starting from the upper left corner of the second image, the center of the window is aligned with each pixel in turn. For example, when the center of the window is located at the (2,2)th pixel (i.e., the 2nd row and 2nd column, with the index starting from 1), the values ​​of the 9 pixels covered by the window are extracted, and the 9 pixel values ​​are sorted from largest to smallest. The median value of the sorted sequence is used as the new value of the center pixel; next, the median value obtained above is written to the corresponding position of a new image of the same size as the original image. The above steps are repeated for each pixel in the second image in the order from left to right and from top to bottom until the entire image is processed.

[0087] In this embodiment of the invention, by performing median filtering on the second image, transient interference (such as sporadic coal dust and noise) in the image can be eliminated, stable background features can be preserved, and the image quality can be improved.

[0088] Step 2: Difference the first image and the reference empty conveyor belt background image to obtain a difference image. In practice, the first image of the current coal sample is differenced from the reference empty conveyor belt background image obtained in Step 1. Specifically, firstly, the sizes of the first image and the reference empty conveyor belt background image are adjusted to be the same, and the two images are precisely aligned by feature matching or manual methods; then, for each pair of corresponding pixels (i.e., pixels at the same position) in the first image and the reference empty conveyor belt background image, the difference between the first image and the reference empty conveyor belt background image is calculated to obtain the difference image.

[0089] In this embodiment of the invention, the difference between the first image of the coal sample and the background image of the reference empty belt can be highlighted through differential operation, which facilitates subsequent analysis.

[0090] Step 3: Perform adaptive histogram equalization on the difference image to obtain the enhanced image. In practice, adaptive histogram equalization or contrast-limited adaptive histogram equalization is performed on the difference image. Specifically, the entire difference image is divided into several non-overlapping (or partially overlapping) small regions, i.e., local sub-blocks, and each local sub-block is processed separately. For each local sub-block, the gray-level distribution of all pixels within it is statistically analyzed, i.e., a gray-level histogram of that local region is constructed. Then, using the gray-level histogram of that local sub-block, its cumulative distribution function is calculated, and based on the cumulative distribution function, the original gray-level value is mapped to a new gray-level value, making the gray-level distribution within the local sub-block as uniform as possible, thereby improving local contrast. Next, for brightness jumps between adjacent local sub-blocks, for any pixel in the image, with its location as the center, the four nearest local sub-blocks (such as the four neighboring blocks in bilinear interpolation) around it are found, and the mapping value of that pixel after equalization in each local sub-block is calculated. Then, a weighted average is performed based on the distance to obtain the final output value, thus obtaining the enhanced image.

[0091] To avoid amplifying image noise, this embodiment of the invention can also employ contrast-limited adaptive histogram equalization. When calculating the grayscale histogram of each local sub-block, a clipping threshold is set, and the height of the histogram bars exceeding the clipping threshold is truncated. The number of clipped pixels is evenly redistributed to other grayscale levels to prevent some grayscale levels from being overstretched. Finally, equalization is performed based on the adjusted grayscale histogram.

[0092] In this embodiment of the invention, adaptive histogram equalization is performed on the difference image, and the contrast is adjusted according to the local characteristics of the image, thereby enhancing the contrast of the image and making the details of the image more prominent.

[0093] Step 4: Perform global thresholding on the enhanced image to obtain a binary image. In practice, Otsu's method is used for global thresholding of the enhanced image. Otsu's method is an image segmentation technique that, by calculating a global threshold, can segment the image into foreground and background parts, facilitating subsequent image analysis.

[0094] Specifically, firstly, the number of pixels at each gray level (0-255) in the entire enhanced image is counted, forming a histogram containing 256 gray levels. This histogram reflects the overall gray-level distribution of the image and typically exhibits a bimodal characteristic: one peak corresponds to the background (such as empty conveyor belts or areas without coal), and the other peak corresponds to the foreground (such as coal sample areas). Then, each possible gray value is tried sequentially as a candidate threshold (from 0 to 255). For each candidate threshold, the image pixels are divided into two categories (background: pixels with gray values ​​≤ T; foreground: pixels with gray values ​​> T). The image is divided into 256 candidate thresholds. The pixel ratio and average gray value of the two types of images are calculated. The inter-class variance under the current candidate threshold is calculated based on the pixel ratio and average gray value. Then, after traversing all 256 candidate thresholds, the candidate threshold that maximizes the inter-class variance is selected as the final segmentation threshold. Finally, the enhanced image is binarized using the final selected segmentation threshold. All pixels with gray values ​​less than or equal to the segmentation threshold are set to 0 (black) to represent the background. All pixels with gray values ​​greater than the segmentation threshold are set to 255 (white) to represent the foreground. The final output is a binary image.

[0095] Step 5: Perform opening and closing operations on the binary image to obtain the processed binary image. In practice, a 3x3 or 5x5 circular kernel is selected to perform opening and closing operations on the binary image. Specifically, the opening operation includes: first, performing an erosion operation on the binary image, sliding the circular kernel pixel by pixel on the binary image. For each position, the corresponding position in the output image is set to white only if all pixels within the kernel's coverage area are white; otherwise, it is set to black, thus completely eliminating small white noise (smaller than the kernel size) and causing the coal sample edges to shrink inward. Next, the same circular kernel is used again to perform a dilation operation on the eroded image. As long as there is at least one white pixel within the kernel's coverage area, the output position is set to white, thus restoring the main body of the coal sample to close to its original size, but the previously removed small noise will not be restored, and the edges become smoother.

[0096] The closing operation includes: first, using the same circular kernel to dilate the original binary image, causing the white area to expand outward. If the size of the small black holes inside the coal sample is smaller than the kernel diameter, they may be covered by white, and adjacent coal blocks may be connected. Then, using the same circular kernel to erode the dilated image, the overall white area shrinks back to close to its original size, but the previously filled holes are no longer restored, and the broken parts remain connected if they are bridged.

[0097] In this embodiment of the invention, by selecting circular kernels of different sizes to perform opening and closing operations, noise and small interferences in the image can be effectively removed while maintaining the main features of the image.

[0098] Step 6: Label the particle regions in the processed binary image and obtain the coal sample particle features based on each particle region. In practice, the processed binary image is labeled to identify each independent particle region. Specifically, after binary image processing, the image is labeled to identify each independent particle region and the particle area is calculated. A p Particle count N P Average particle size d avg Flatness F P Particle characteristics of coal samples.

[0099] Specifically, firstly, connected component analysis is used to traverse all white regions in the image. When an unlabeled white pixel is encountered, a region growing operation is performed, grouping that pixel and all white pixels connected by 8-neighborhoods into the same region and assigning it a unique label number. Then, the total number of distinct labels appearing in the labeled image (i.e., the maximum label value) is counted, which is the number of particles N. P Next, for each label i ( i =1,2,…, N P ), Traverse the entire marked image, and count all pixels whose values ​​are equal to i The number of pixels is the number of pixels in the first position. i Particle area of ​​each particle A pi Assuming the particles are approximately spherical, their equivalent particle size (diameter) can be deduced from their area: Then, the arithmetic mean of the equivalent particle sizes of all particles is calculated to obtain the average particle size: Finally, the flatness of the particles is calculated based on their area. .in, For the first i The perimeter of each particle. The flatness ranges between (0,1). The closer to 1, the closer the particle is to an ideal circle. The smaller the value, the more irregular the shape, and the rougher or thinner the edges.

[0100] The method provided in this invention suppresses major physical interferences at the source through real-time temperature correction and particle size sensing. Through deep fusion of XRF and NIR, it achieves chemical complementarity between organic and inorganic information, improving the prediction accuracy of indicators such as ash, sulfur, moisture, and volatile matter by 30%-50% compared to single-spectral techniques. For a dataset containing 1500 samples, particle sizes range from 0.5mm to 3mm, with coarse particles of 3-6mm, and a temperature range of 10-50 degrees Celsius. Coal quality ranges: ash: 5%-50%, moisture: 2%-20%, volatile matter: 15%-40%, calorific value: 3000-6500 kcal / kg. This invention compares the prediction results of the following three schemes: Scheme A: X-ray fluorescence technology; Scheme B: X-ray fluorescence technology + temperature correction; Scheme C: X-ray fluorescence + near-infrared + temperature + particle size sensing. The accuracy comparison is shown in Table 1.

[0101] Table 1 Comparison of prediction accuracy among the three schemes

[0102] The method provided in this embodiment of the invention includes the following features: 1) Considering the relatively serious environmental pollution and high dust levels in coal mining, a correction process for empty conveyor belt spectroscopy is added, and empty conveyor belt spectroscopy is collected and features are extracted using the gaps in the coal sample spectroscopy; 2) Near-infrared spectroscopy and X-ray fluorescence spectroscopy are collected simultaneously and complement each other, with near-infrared spectroscopy supplementing the influence of moisture in coal and the influence of organic matter on different coal types; 3) Considering the influence of different working conditions, temperature is introduced into the model; 4) Considering the influence of particle size and coal plane flatness on scattering effects, particle size features are introduced; 5) Key features are extracted, and multicollinearity features are introduced, which reduces the sample size requirements of complex models and avoids the problem of overfitting in complex models, reducing the difficulty of debugging and modeling. Maintenance personnel only need to maintain the basic operation of the hardware. Once the model is not used, it will automatically prompt, for example, if the temperature exceeds the model range, the particle size exceeds the existing model range, or the window contamination reaches a certain level.

[0103] In summary, this invention provides an online rapid coal quality detection method with deep fusion of multi-source information and adaptive correction capability under various operating conditions. It can simultaneously detect coal ash, moisture, volatile matter, and sulfur, improving the speed and accuracy of coal quality detection. The model incorporates multiple operating condition variables during the training phase, coupled with an online adaptive learning mechanism, significantly enhancing its generalization ability and long-term operational stability in the face of coal quality variations in different mining areas, and reducing the frequency of recalibration. A single system and a single measurement simultaneously output all key industrial analysis indicators for coal, truly meeting the needs of online rapid detection and providing real-time data support for intelligent coal blending, process control, and fair trading.

[0104] Regarding the coal quality testing method based on multispectral fusion and operating condition self-calibration provided in the foregoing embodiments, this invention also provides a coal quality testing device based on multispectral fusion and operating condition self-calibration, see [link to relevant documentation]. Figure 5 The diagram shows a structural schematic of a coal quality detection method based on multispectral fusion and operating condition self-correction, illustrating that the method mainly includes the following parts: The data acquisition module 501 is used to acquire the first spectral data, the first image, and the real-time surface temperature of the coal sample, as well as the second spectral data and the second image of the empty conveyor belt.

[0105] The first feature extraction module 502 is used to preprocess the first spectral data and the second spectral data, and extract the spectral features of the coal sample and the empty belt based on the processed first spectral data and the second spectral data.

[0106] The second feature extraction module 503 is used to extract features from the first image and the second image to obtain the particle features of the coal sample.

[0107] The coal quality detection module 504 is used to input the real-time surface temperature of the coal sample, the spectral characteristics of the coal sample, the spectral characteristics of the empty belt, and the particle characteristics of the coal sample into a pre-trained coal quality prediction model to obtain the coal quality detection results.

[0108] The coal quality testing device based on multispectral fusion and working condition self-correction provided in this invention can acquire the real-time surface temperature of coal samples and identify the working condition based on the real-time surface temperature. By acquiring the first spectral data of the coal sample and the second spectral data of the empty conveyor belt and extracting features, the spectral features of the empty conveyor belt can be used to correct the spectral features of the coal sample during coal quality testing, thereby improving the accuracy of coal sample spectral feature detection. By acquiring the first image of the coal sample and the second image of the empty conveyor belt and extracting features to obtain the particle features of the coal sample, and combining the real-time surface temperature of the coal sample, the spectral features of the coal sample, the spectral features of the empty conveyor belt, and the particle features of the coal sample for coal quality testing, it can simultaneously detect indicators such as coal ash content, moisture, volatile matter, and sulfur content through deep fusion of multi-source information and adaptive working condition correction, effectively improving the problems existing in the prior art and improving the speed and accuracy of coal quality testing.

[0109] It should be noted that the device provided in this embodiment of the invention has the same implementation principle and technical effects as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the aforementioned method embodiment. The specific numerical values ​​provided in this embodiment are merely exemplary and are not intended to limit the scope of the invention.

[0110] This invention also provides an electronic device, specifically, the electronic device includes a processor and a storage device; the storage device stores a computer program, and the computer program, when run by the processor, executes the method described in any of the above embodiments.

[0111] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The electronic device 100 includes: a processor 60, a memory 61, a bus 62, and a communication interface 63. The processor 60, the communication interface 63, and the memory 61 are connected through the bus 62. The processor 60 is used to execute executable modules, such as computer programs, stored in the memory 61.

[0112] The memory 61 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 63 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0113] Bus 62 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 6 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0114] The memory 61 is used to store programs. After receiving an execution instruction, the processor 60 executes the program. The method executed by the device for defining the flow process disclosed in any of the foregoing embodiments of the present invention can be applied to the processor 60 or implemented by the processor 60.

[0115] Processor 60 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 60 or by instructions in software form. Processor 60 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 61. Processor 60 reads the information in memory 61 and, in conjunction with its hardware, completes the steps of the above method.

[0116] The computer program product of the readable storage medium provided in the embodiments of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the foregoing method embodiments. For specific implementation, please refer to the foregoing method embodiments, which will not be repeated here.

[0117] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0118] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A coal quality detection method based on multispectral fusion and operating condition self-calibration, characterized in that, include: The first spectral data, first image, and real-time surface temperature of the coal sample were collected, along with the second spectral data and second image of the empty conveyor belt. The first spectral data and the second spectral data are preprocessed, and the spectral features of the coal sample and the empty conveyor belt are extracted based on the processed first spectral data and second spectral data. Feature extraction is performed on the first image and the second image to obtain the particle characteristics of the coal sample. The real-time surface temperature of the coal sample, the spectral characteristics of the coal sample, the spectral characteristics of the empty belt, and the particle characteristics of the coal sample are input into a pre-trained coal quality prediction model to obtain the coal quality detection results.

2. The method according to claim 1, characterized in that, The first spectral data includes: a first near-infrared spectrum and a first X-ray fluorescence spectrum; the second spectral data includes: a second near-infrared spectrum and a second X-ray fluorescence spectrum; The first spectral data, first image, and real-time surface temperature of the coal sample were collected, along with the second spectral data and second image of the empty conveyor belt, including: The first distance information from the target location to the coal conveyor belt is detected by a laser rangefinder; If the first distance information is less than a preset value, it is determined that a coal sample has passed through the coal conveyor belt, and the first near-infrared spectrum, first X-ray fluorescence spectrum, first image, and real-time surface temperature of the coal sample are collected. After the coal samples were collected, the second near-infrared spectrum, the second X-ray fluorescence spectrum, and the second image of the empty conveyor belt were acquired.

3. The method according to claim 2, characterized in that, Preprocessing of the first spectral data and the second spectral data includes: The first near-infrared spectrum and the second near-infrared spectrum are processed respectively to obtain the second derivative spectrum of the coal sample and the second derivative spectrum of the empty conveyor belt; wherein, the near-infrared spectrum is one-dimensional data, the first derivative spectrum is the spectrum obtained by taking the first derivative of the near-infrared spectrum, and the second derivative spectrum is the spectrum obtained by taking the derivative of the first derivative spectrum. The first X-ray fluorescence spectrum and the second X-ray fluorescence spectrum were processed respectively to obtain the characteristic peak spectrum of coal, the background spectrum of coal, the background spectrum of empty conveyor belt, and the characteristic peak spectrum of empty conveyor belt.

4. The method according to claim 3, characterized in that, Based on the processed first and second spectral data, spectral features of coal samples and empty conveyor belts were extracted, including: The functional group absorption intensity characteristics of the coal sample are extracted based on the second derivative spectrum of the coal sample, and the functional group absorption intensity characteristics of the empty conveyor belt are extracted based on the second derivative spectrum of the empty conveyor belt. The elemental intensity characteristics of coal are calculated based on the characteristic peak spectra of the coal, and the background spectral intensity characteristics of coal are calculated based on the background spectra of the coal. The elemental intensity characteristics of the empty belt are calculated based on the characteristic peak spectrum of the empty belt, and the background spectral intensity characteristics of the empty belt are calculated based on the background spectrum of the empty belt. Based on the functional group absorption intensity characteristics and elemental intensity characteristics of the coal sample, the coal correlation characteristics are calculated.

5. The method according to claim 3, characterized in that, The first near-infrared spectrum and the second near-infrared spectrum are processed respectively to obtain the second derivative spectrum of the coal sample and the second derivative spectrum of the empty conveyor belt, including: The first near-infrared spectrum and the second near-infrared spectrum are filtered respectively to obtain the first near-infrared smooth spectrum and the second near-infrared smooth spectrum. The first total absorbance of the first near-infrared smoothed spectrum and the second total absorbance of the second near-infrared smoothed spectrum are calculated respectively. Based on the first total absorbance, abnormal spectra in the first near-infrared smoothed spectrum are removed to obtain the first target spectrum. Based on the second total absorbance, abnormal spectra in the second near-infrared smoothed spectrum are removed to obtain the second target spectrum. Scattering corrections are performed on the first target spectrum and the second target spectrum respectively to obtain a first corrected spectrum and a second corrected spectrum; The first and second corrected spectra are processed by derivatives to obtain the second derivative spectra of the coal sample and the second derivative spectra of the empty conveyor belt. The peak area of ​​the preset wavelength range is calculated based on the second derivative spectra of the coal sample and the second derivative spectra of the empty conveyor belt.

6. The method according to claim 3, characterized in that, The first X-ray fluorescence spectrum and the second X-ray fluorescence spectrum are processed respectively to obtain the characteristic peak spectrum of coal, the background spectrum of coal, the background spectrum of empty conveyor belt, and the characteristic peak spectrum of empty conveyor belt, including: The first X-ray fluorescence spectrum and the second X-ray fluorescence spectrum are filtered respectively to obtain the first X-ray fluorescence smooth spectrum and the second X-ray fluorescence smooth spectrum; Based on the standard coal sample spectrum, the first X-ray fluorescence smoothed spectrum and the second X-ray fluorescence smoothed spectrum are drift corrected to obtain the first fluorescence corrected spectrum and the second fluorescence corrected spectrum; Based on the first fluorescence-corrected spectrum, the characteristic peak spectrum of coal and the background spectrum of coal are extracted, and based on the second fluorescence-corrected spectrum, the background spectrum of empty conveyor belt and the characteristic peak spectrum of empty conveyor belt are extracted.

7. The method according to claim 1, characterized in that, Feature extraction is performed on the first image and the second image to obtain the particle features of the coal sample, including: The second image is subjected to median filtering to obtain the reference empty belt background image; The first image and the reference empty belt background image are differentially divided to obtain a differential image; The difference image is subjected to adaptive histogram equalization to obtain an enhanced image; The enhanced image is subjected to global thresholding to obtain a binary image; The binary image is subjected to opening and closing operations to obtain the processed binary image; The particle regions in the processed binary image are labeled, and the particle features of the coal sample are obtained based on each particle region.

8. A coal quality detection device based on multispectral fusion and operating condition self-calibration, characterized in that, include: The data acquisition module is used to acquire the first spectral data, first image, and real-time surface temperature of the coal sample, as well as the second spectral data and second image of the empty conveyor belt. The first feature extraction module is used to preprocess the first spectral data and the second spectral data, and extract the spectral features of the coal sample and the empty conveyor belt based on the processed first spectral data and second spectral data. The second feature extraction module is used to extract features from the first image and the second image to obtain the particle features of the coal sample; The coal quality detection module is used to input the real-time surface temperature of the coal sample, the spectral characteristics of the coal sample, the spectral characteristics of the empty belt, and the particle characteristics of the coal sample into a pre-trained coal quality prediction model to obtain the coal quality detection results.

9. An electronic device, characterized in that, The method includes a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program thereon, characterized in that, The computer program is executed by the processor to perform the steps of the method described in any one of claims 1 to 7.

Images