Hazardous chemicals identification method and device based on multispectral fusion

Abstract

The present invention provides a method and device for hazardous chemical identification based on multi-spectral fusion, relating to the technical field of hazardous chemical identification. The method comprises: extracting features from a first spectrum and a second spectrum of an object to be detected, respectively, to obtain a first feature of the first spectrum and a second feature of the second spectrum; determining a first matching degree of the first feature and a second matching degree of the second feature based on the object characteristics of the object to be detected; fusing the first and second features based on the first and second matching degrees to obtain spectral features of the object to be detected; and inputting the spectral features into a pre-trained hazardous chemical identification model to obtain a hazardous chemical identification result for the object to be detected. The present invention can effectively improve the accuracy of spectral features and the detection precision of hazardous chemicals.

Description

Technical Field

[0001] The present invention relates to the technical field of hazardous chemical identification, and in particular to a hazardous chemical identification method and device based on multi-spectral fusion. Background Art

[0002] Hazardous chemicals, such as strong acids, organic solvents, and cyanide, are chemical substances or mixtures that pose physical, health, or environmental hazards. These substances can pose a serious threat to human health, public safety, or the ecological environment. These chemicals are typically strictly regulated, and their production, storage, transportation, and use must comply with specific standards and regulations.

[0003] While international trade in hazardous chemicals meets industrial needs and facilitates technological exchange, it can also pose significant risks. Cross-border movement can exacerbate environmental leaks and transportation accidents, such as toxic spills that contaminate water sources or explosions caused by the transport of flammable materials. Currently, infrared spectroscopy is often used as a rapid screening tool for on-site inspections. This technology can identify the components of a substance by identifying its characteristic spectral peaks, providing a preliminary assessment of whether the cargo contains flammable, explosive, or toxic hazardous chemicals.

[0004] However, infrared spectroscopy detection technology is easily affected by interference from mixtures, leading to misjudgment or missed detection, and it lacks sensitivity to low-concentration substances or complex matrices. In addition, due to the limited coverage of infrared spectroscopy databases, it is difficult to deal with new or modified chemicals. Summary of the Invention

[0005] The embodiments of the present invention provide a method and device for identifying hazardous chemicals based on multi-spectral fusion to solve the problem that infrared spectroscopy detection technology has low detection accuracy and is difficult to deal with new or modified chemicals when performing hazardous chemical detection.

[0006] In a first aspect, an embodiment of the present invention provides a method for identifying hazardous chemicals based on multispectral fusion, comprising:

[0007] Perform feature extraction on the first spectrum and the second spectrum of the object to be detected, respectively, to obtain a first feature of the first spectrum and a second feature of the second spectrum;

[0008] Determining a first matching degree of the first feature and a second matching degree of the second feature according to an object feature of the object to be detected;

[0009] Based on the first matching degree and the second matching degree, the first feature and the second feature are fused to obtain a spectral feature of the object to be detected;

[0010] The spectral features are input into the pre-trained hazardous chemical identification model to obtain the hazardous chemical identification results of the items to be tested.

[0011] In a second aspect, an embodiment of the present invention provides a hazardous chemicals identification device based on multispectral fusion, comprising:

[0012] An extraction module, configured to extract features from the first spectrum and the second spectrum of the object to be detected, respectively, to obtain a first feature of the first spectrum and a second feature of the second spectrum;

[0013] a determination module, configured to determine a first matching degree of the first feature and a second matching degree of the second feature according to the object feature of the object to be detected;

[0014] a fusion module, configured to fuse the first feature and the second feature based on the first matching degree and the second matching degree to obtain a spectral feature of the object to be detected;

[0015] The recognition module is used to input the spectral features into a pre-trained hazardous chemical recognition model to obtain the hazardous chemical recognition results of the items to be tested.

[0016] In an embodiment of the present invention, by synchronously extracting the first spectrum and the second spectrum features of the object to be detected, and evaluating the feature matching degree in combination with the object features, performing feature fusion according to the feature matching degree to obtain spectral features, and inputting the spectral features into a pre-trained hazardous chemical identification model, a hazardous chemical identification result of the object to be detected is obtained. Using object features to obtain the matching degree, and obtaining spectral features based on the characteristics of the matching degree can improve the accuracy of feature characterization. The hazardous chemical identification model combined with deep learning can accurately identify known hazardous chemicals through spectral features, and judge whether there is a danger when the object is not a known hazardous chemical based on the spectral features. Therefore, the present application can effectively improve the accuracy of spectral features and the detection accuracy of hazardous chemicals, and has the ability to identify new or modified chemicals. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flowchart of an implementation method for hazardous chemicals identification based on multispectral fusion provided by an embodiment of the present invention;

[0018] Figure 2 This is a flowchart for implementing step S120 of the method for identifying hazardous chemicals based on multispectral fusion provided by an embodiment of the present invention;

[0019] Figure 3 This is a flowchart for implementing step S130 of the method for identifying hazardous chemicals based on multispectral fusion provided by an embodiment of the present invention;

[0020] Figure 4 This is a flowchart for implementing step S140 of the method for identifying hazardous chemicals based on multispectral fusion provided by an embodiment of the present invention;

[0021] Figure 5This is a flowchart for implementing step S1402 of the method for identifying hazardous chemicals based on multispectral fusion provided by an embodiment of the present invention;

[0022] Figure 6 It is a structural diagram of a hazardous chemicals identification device based on multi-spectral fusion provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0024] See also Figure 1 , which shows a flowchart of the implementation of the hazardous chemicals identification method based on multi-spectral fusion provided by an embodiment of the present invention, and is detailed as follows:

[0025] Step S110 , performing feature extraction on the first spectrum and the second spectrum of the object to be detected, respectively, to obtain a first feature of the first spectrum and a second feature of the second spectrum.

[0026] In some embodiments, a first spectrum and a second spectrum of an object to be detected can be obtained through spectral detection. Specifically, the object to be detected can first be spectrally detected to obtain an initial first spectrum and an initial second spectrum of the object to be detected. Subsequently, the initial first spectrum and the initial second spectrum can be preprocessed to obtain the first spectrum and the second spectrum. The preprocessing can include data denoising and baseline correction.

[0027] Specifically, spectroscopy is a technique that analyzes the interaction between matter and electromagnetic waves to obtain information about a substance's composition. When light of a specific wavelength strikes a surface, the substance absorbs or scatters that wavelength due to molecular vibrations, electronic transitions, and other processes, forming a characteristic spectrum. For example, in hazardous chemical testing, infrared spectroscopy can identify the C-H bond absorption peaks of organic solvents, while Raman spectroscopy can detect molecular vibrational modes. Combining these two methods allows for a more comprehensive analysis of chemical composition.

[0028] It should be noted that in this application, two different detection methods are required to perform spectral detection on the inspected item. This application adopts infrared spectroscopy and Raman spectroscopy. The initial first spectrum and the initial second spectrum are raw data directly obtained using different spectral detection methods. The initial first spectrum is the infrared spectral response obtained by measuring hazardous chemicals using infrared spectroscopy, while the initial second spectrum is the Raman spectral response obtained by measuring hazardous chemicals using Raman spectroscopy. Both contain unprocessed noise and baseline drift, such as ambient light interference in infrared spectra.

[0029] In some embodiments, data denoising is used to eliminate interference from non-target signals in the spectrum through an algorithm. For example, the infrared spectrum can be smoothed using a Savitzky-Golay filter to retain the effective peak shape while removing small fluctuations during instrument acquisition. Baseline correction is used to eliminate nonlinear background offsets in the spectrum. For example, when detecting a strong acid solution, the fluorescence effect will cause the Raman spectrum baseline to warp as a whole. In this case, the iterative least squares method can be used to fit the baseline curve, which can then be subtracted from the original spectrum. In addition, preprocessing can also include spectral smoothing, normalization, wavelength calibration, scattering processing, and other processing methods.

[0030] By synchronously collecting the initial spectral data of the object to be tested through multi-spectral fusion, combined with data denoising and baseline correction preprocessing, it is possible to effectively eliminate noise interference and baseline drift, improve the authenticity and signal-to-noise ratio of spectral features, and thus enhance the accuracy of subsequent feature fusion and model recognition.

[0031] In this way, noise interference can be eliminated through preprocessing, and the signal-to-noise ratio of the first spectrum and the second spectrum can be improved.

[0032] In some embodiments, feature extraction is the process of selecting key discriminative information from the first and second spectra. For example, derivatives are used to enhance the resolution of overlapping peaks in infrared spectra, or principal component analysis is performed on Raman spectra to reduce dimensionality and extract the first three principal components as feature vectors representing the overall spectral variation.

[0033] In some embodiments, the first feature is a quantitative indicator extracted from the first spectrum, including intensity, peak area, or waveform parameters at a specific wavelength. For example, the first feature of the infrared spectrum may include the intensity at 1740 cm -1 The second feature is the quantitative index extracted from the pre-processed second spectrum, including the intensity, peak area, half-peak width and peak position shift of the specific Raman shift. For example, when detecting sulfuric acid, its SO bond vibration at 980 cm -1 A characteristic peak is generated at the position where the peak is located, and the second characteristic can quantify the intensity, peak area ratio and half-peak width of the peak.

[0034] Step S120 : determining a first matching degree of the first feature and a second matching degree of the second feature according to the object feature of the object to be detected.

[0035] In some embodiments, object features refer to the physical properties of the object being detected, including packaging material, packaging form, packaging surface characteristics, and possible interference factors. For example, the object features of a viscous liquid in a brown plastic bottle include the plastic material, opaque packaging, and liquid form.

[0036] In some embodiments, the first match metric measures the compatibility of infrared spectral features with object features. For example, if the object being tested is a powder in a metal can, the high reflectivity of infrared light from the metal surface will result in a decreased signal-to-noise ratio, resulting in a low infrared feature match. The second match metric assesses the correlation between Raman spectral features and object features. For example, a transparent glass container will have minimal interference with the Raman signal, resulting in a high Raman feature match.

[0037] See also Figure 2 The specific processing of the above step S120 may include steps S1201-S1203, the specific contents of which are as follows:

[0038] Step S1201 : Determine packaging information and interference information of the object to be detected based on the object characteristics of the object to be detected.

[0039] In some embodiments, packaging information includes packaging material, packaging shape, and packaging surface characteristics. For example, for powder packaged in a sealed aluminum can, the packaging information would include aluminum can, cylindrical shape, and matte coating. Interference information refers to the interference caused by the packaging or the object itself on the detection signal. For example, a glass bottle may cause ambient light noise in the Raman spectrum due to surface reflection, and liquid sloshing may cause baseline drift in the infrared spectrum.

[0040] Step S1202 : Compare the packaging information and the interference information with the acquisition requirements of the first spectrum to determine a first matching degree of the first feature.

[0041] In some embodiments, the first degree of matching is used to quantify the compatibility between the infrared spectral characteristics and the object characteristics of the object to be detected. The first degree of matching can be calculated based on the degree of influence of the packaging information and interference information on the infrared spectral acquisition requirements. Specifically, by analyzing the matching relationship between the object characteristics and the infrared light penetrability and anti-interference ability, the key parameters are converted into weight indicators, and weighted scoring is performed in combination with the preset threshold to obtain the first degree of matching. For example, if the object to be tested is an opaque plastic bottle, its infrared light penetration is blocked, resulting in attenuation of the effective signal and an increase in the proportion of reflected noise, the system will calculate the first degree of matching based on the degree of influence of the opaque plastic bottle, and finally output a value in the range of 0-1. The higher the value, the more reliable the infrared spectral characteristics.

[0042] In some embodiments, the acquisition requirements refer to the core requirements of different spectral technologies on the physical properties of the measured object in order to obtain effective signals. For example, infrared spectroscopy relies on transmittance and low reflection interference, while Raman spectroscopy requires high scattering efficiency and low fluorescence background.

[0043] Step S1203 : Compare the packaging information and the interference information with the acquisition requirements of the second spectrum to determine a second matching degree of the second feature.

[0044] In some embodiments, the second degree of match is used to evaluate the compatibility of the Raman spectral characteristics with the object characteristics of the object to be detected. The second degree of match can be calculated by the degree to which the packaging information and interference information meet the Raman spectral acquisition requirements. Specifically, by analyzing the correlation between the object characteristics and the Raman light scattering efficiency and the anti-fluorescence interference ability, the key parameters are converted into quantitative indicators, and a comprehensive score is performed in combination with the preset weights to obtain the second degree of match. For example, if the object to be tested is a transparent glass bottle with significant Raman signal intensity and clear characteristic peaks, the system will calculate the matching score based on the scattering efficiency and fluorescence background intensity, and output a value in the range of 0-1. The higher the value, the more suitable the Raman spectral characteristics are for the current detection scenario.

[0045] In some embodiments, the matching degree calculation formula is: matching degree = Σ (parameter score × weight coefficient). Table 1 is an example of some object features and acquisition requirements. Referring to Table 1, the first matching degree and the second matching degree can be calculated through the mapping table of object features and acquisition requirements and the above matching degree calculation formula. The parameter score can be calculated comprehensively based on the score level and score range. For example, if object A is a transparent glass bottle, the score level of its infrared spectrum material transmittance is high, the score range is 0.8-1, and the corresponding specific transmittance is 85%, then the corresponding score is 0.85 points; if object B is a colorless transparent quartz container, the score level of its Raman spectrum fluorescence interference intensity is low, the score range is 0.8-1.0, and the corresponding fluorescence interference intensity is 6%, then when calculating the corresponding score, it is necessary to consider the correspondence between the range of fluorescence interference intensity corresponding to the score level and the score range. Since the score level of fluorescence interference intensity is low, its range is [0-10%], and the corresponding score range is [0.8-1.0], after calculation, it is obtained that when the fluorescence interference intensity is 6%, the corresponding score is 0.88 points.

[0046] Table 1 Mapping table of object features and collection requirements

[0047]

[0048] By dynamically combining object features to analyze packaging information and interference information, and performing matching evaluation based on the acquisition requirements of different spectral technologies, multi-dimensional detection optimization is achieved.

[0049] Step S130 : Based on the first matching degree and the second matching degree, the first feature and the second feature are fused to obtain the spectral feature of the object to be detected.

[0050] In some embodiments, feature fusion involves combining two types of features based on their matching. For example, when detecting a viscous liquid in a brown plastic bottle, the infrared match is low due to the opaque packaging, while the Raman match is high due to effective scattering. The system assigns a higher weight to the Raman feature and linearly fuses the two features, ultimately generating a spectral signature that includes weighted peak intensities and peak position correction information.

[0051] See also Figure 3 The specific processing of the above step S130 may include steps S1301-S1302, the specific contents of which are as follows:

[0052] Step S1301 : Based on the first matching degree, feature screening is performed on the first feature, and based on the second matching degree, feature screening is performed on the second feature.

[0053] In some embodiments, feature screening is to select or weight features based on the degree of matching. For example, when the object to be tested is a metal can, due to the poor light transmittance of the metal can, the infrared matching degree is low, and the system will eliminate the band features affected by reflection interference.

[0054] Step S1302 : performing feature fusion on the filtered first feature and the filtered second feature according to the first matching degree and the second matching degree to obtain the spectral feature of the object to be detected.

[0055] In some embodiments, the first feature after screening refers to the effective infrared feature screened according to the first matching degree, for example, retaining the 1200 cm-1 infrared spectrum that is not interfered by the package reflection. -1 The second feature after screening refers to the effective Raman feature retained according to the second matching degree.

[0056] In some embodiments, feature fusion involves integrating two types of features based on matching, such as an infrared matching degree of 0.2 and a Raman matching degree of 0.9, and fusing the peak intensities of the two features in a ratio of 2:9 to form a composite spectral feature. A spectral feature is a fused multidimensional feature vector, such as one containing weighted ester peak intensity, SO peak intensity, and a combined peak shift parameter, which is used to accurately characterize hazardous chemical components.

[0057] In one implementation, the specific processing method of step S1302 is: based on the first matching degree and the second matching degree, determine the first credibility of the first feature and the second credibility of the second feature; according to the first credibility and the second credibility, perform feature fusion on the screened first feature and the screened second feature to obtain the spectral characteristics of the object to be detected.

[0058] In some embodiments, the first credibility and the second credibility are feature reliability weights calculated based on the matching degree. The first credibility is the quotient of the first matching degree and the sum of the first and second matching degrees, and the second credibility is the quotient of the second matching degree and the sum of the first and second matching degrees. For example, when the infrared matching degree is 0.3 and the Raman matching degree is 0.8, the first credibility corresponding to the infrared is 0.27, and the second credibility corresponding to the Raman is 0.73, indicating that the Raman feature is more reliable.

[0059] In some embodiments, during the feature fusion stage, valid features are first screened based on matching accuracy, such as eliminating invalid wavelengths in the infrared spectrum due to interference from metal packaging reflections. Subsequently, the first and second features of the infrared and Raman spectra are dynamically weighted and fused based on their confidence. Confidence serves as a weighting factor, directly determining the contribution of each feature.

[0060] Dynamic fusion of multispectral features improves the robustness and accuracy of hazardous chemical detection. By filtering and eliminating invalid features affected by interference based on matching, and assigning higher weights to highly compatible features based on their credibility, this approach achieves precise weighted fusion. This approach not only adapts to complex scenarios but also leverages the complementary nature of multispectral features to overcome the limitations of single technologies, ultimately generating spectral signatures that effectively improve the accuracy of hazardous chemical identification.

[0061] In step S140 , the spectral features are input into a pre-trained hazardous chemical identification model to obtain a hazardous chemical identification result of the object to be detected.

[0062] In some embodiments, a pre-trained hazardous chemical identification model refers to a machine learning model trained on a large amount of hazardous chemical spectral data. It can learn the mapping between spectral features and hazardous chemical categories. For example, a convolutional neural network is trained on fused spectral data from thousands of hazardous chemicals, enabling the model to distinguish between ammonium nitrate and sulfuric acid. The hazardous chemical identification result refers to the classification conclusion output by the model, including substance category and hazard level. For example, after inputting the fused spectral features of a liquid, the model outputs a 98% probability of concentrated sulfuric acid, with a hazard level of Class I corrosive.

[0063] It should be noted that the training process of the pre-trained hazardous chemical identification model is as follows: First, spectral enhancement technology is used to simulate diverse interference scenarios. By adding noise and baseline drift, a synthetic dataset covering different packaging materials and environmental interference is constructed. Second, a deep hybrid network architecture is designed, combining a residual structure with an attention mechanism to hierarchically analyze spectral features. A shallow network is used to extract basic waveform features, while a deep network is used to explore the laws of molecular bond coordination and variation. A domain adaptation module is introduced to eliminate device differences. Finally, a two-stage dynamic training strategy is adopted, first transferring general visual feature extraction capabilities, then fine-tuning parameters based on spectral characteristics. At the same time, an uncertainty quantification mechanism is integrated to dynamically assess the risks of unknown substances through probabilistic modeling. This training method enables the model to both identify subtle spectral differences in known hazardous chemicals and infer new compound categories through feature association, achieving stable and reliable generalization performance in complex detection scenarios.

[0064] See also Figure 4 The specific processing of the above step S140 may include steps S1401-S1402, and the specific contents are as follows:

[0065] Step S1401: Input the spectral features into a pre-trained hazardous chemicals identification model to obtain the composition information of the item to be detected.

[0066] In some embodiments, the component information refers to the chemical composition and properties of the object to be detected, including the substance name, concentration, chemical bond characteristics, etc.

[0067] Step S1402: Based on the component information, obtain the hazardous chemical identification result of the item to be tested.

[0068] See also Figure 5 The specific processing of the above step S1402 may include steps S14021-S14022, the specific contents of which are as follows:

[0069] Step S14021, compare the component information with the component information of each known hazardous chemical in the hazardous chemical information database to obtain a comparison result.

[0070] In some embodiments, the hazardous chemicals information database is a database that stores standardized data on known hazardous chemicals, including information such as substance name, CAS number, ingredient ratio, hazard category, etc. For example, the entry for potassium nitrate in the database is marked as "oxidizing solid, hazard category 5.1, concentration ≥10% is considered a hazardous substance." The comparison result refers to the calculated degree of correlation between the component to be tested and the hazardous chemicals in the database. For example, a liquid component has a Raman peak of 783 cm-1 with respect to "acetone" in the database. -1 and infrared peak at 1720 cm -1 Exact match, with a similarity of up to 98%.

[0071] Step S14022: Obtain the hazardous chemical identification result of the item to be tested based on the component information and the comparison result.

[0072] In one implementation, the specific processing method of step S14022 is: based on the comparison result, determine the similarity between the composition information of the item to be detected and the composition information of each known hazardous chemical in the hazardous chemicals information library; if there is a target known hazardous chemical whose similarity with the item to be detected exceeds a preset threshold, then the item to be detected is determined to be the target known hazardous chemical; if there is no target known hazardous chemical whose similarity with the item to be detected exceeds the preset threshold, then based on the composition information of the item to be detected, calculate the hazard level of the item to be detected, and based on the hazard level, obtain the hazardous chemical identification result of the item to be detected.

[0073] In some embodiments, similarity refers to a match value quantified by algorithms such as cosine similarity or spectral peak matching, such as an 85% match between the Raman peaks of a gas and methane. A preset threshold is a critical value used to determine whether a match occurs. Hazard refers to a risk value calculated based on the toxicity, flammability, and other properties of a component. For example, a mixture containing 50% ethanol has a hazard level of 0.8. Hazardous chemical identification results refer to the final classification conclusion, such as "ammonium nitrate, explosive" or "unknown high-hazard liquid."

[0074] By inputting multi-spectral fusion features into the deep learning model, the intelligence level and detection accuracy of hazardous chemical identification have been significantly improved. First, the spectral features are analyzed through convolutional neural networks to accurately extract material composition information. Secondly, two-way comparison is performed in combination with the hazardous chemical information database, and high-precision matching is achieved using algorithms such as cosine similarity, effectively solving the problem of traditional methods having difficulty in identifying new chemicals and modified chemicals.

[0075] In some embodiments, after obtaining the hazardous chemical identification results, a corresponding alarm message may be issued. Specifically, after obtaining the hazardous chemical identification results, the hazard category and hazard level of the item to be detected may be determined based on the hazardous chemical identification results; and based on the hazard category and hazard level, an alarm message corresponding to the item to be detected may be issued.

[0076] Specifically, hazard categories refer to the types of hazards classified by chemical properties, including explosiveness, flammability, and corrosiveness. Hazard levels are hierarchical indicators used to quantify the degree of hazard, with level 1 representing a high hazard and level 2 representing a moderate hazard. Alarm information is a graded response signal triggered by hazard assessment, including audible and visual alarms, emergency instructions, and more. For example, the detection of a level 1 explosive triggers an area-wide evacuation alarm, while a level 2 corrosive substance triggers an isolation zone warning light.

[0077] Adopting a complementary infrared and Raman spectroscopy acquisition strategy, adaptive feature screening is achieved through matching evaluation, which can significantly reduce the interference of mixtures and the impact of packaging materials. A spectral feature fusion model based on deep learning is constructed, combined with a two-way comparison mechanism of the hazardous chemicals information database, breaking through the limitations of traditional database coverage. This not only improves the detection sensitivity of low-concentration substances, but also can identify new chemicals and modified chemicals through feature association reasoning. An object feature-driven credibility assessment system is introduced to achieve dynamic weight allocation of multi-source features, enabling the detection system to maintain high robustness in complex scenarios. Therefore, this application can effectively improve the detection accuracy of hazardous chemicals and has the ability to identify new chemicals and modified chemicals.

[0078] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0079] The following are device embodiments of the present invention. For details not fully described therein, reference may be made to the corresponding method embodiments described above.

[0080] Figure 6 The following is a schematic diagram of the structure of a hazardous chemical identification device based on multispectral fusion provided by an embodiment of the present invention. For ease of explanation, only the parts related to the embodiment of the present invention are shown, which are detailed as follows:

[0081] like Figure 6 As shown, the hazardous chemicals identification device 6 based on multi-spectral fusion includes:

[0082] An extraction module 61 is configured to extract features from the first spectrum and the second spectrum of the object to be detected, respectively, to obtain a first feature of the first spectrum and a second feature of the second spectrum;

[0083] a determination module 62 for determining a first matching degree of the first feature and a second matching degree of the second feature according to the object feature of the object to be detected;

[0084] A fusion module 63 is configured to fuse the first feature and the second feature based on the first matching degree and the second matching degree to obtain a spectral feature of the object to be detected;

[0085] The identification module 64 is used to input the spectral features into a pre-trained hazardous chemical identification model to obtain a hazardous chemical identification result of the object to be detected.

[0086] In one possible implementation, the extraction module 61 is specifically configured to: perform spectral detection on the object to be detected to obtain an initial first spectrum and an initial second spectrum of the object to be detected; and preprocess the initial first spectrum and the initial second spectrum to obtain a first spectrum and a second spectrum; wherein the preprocessing includes data denoising and baseline correction.

[0087] In one possible implementation, the determination module 62 is specifically used to: determine the packaging information and interference information of the item to be detected based on the object characteristics of the item to be detected; compare the packaging information and interference information with the acquisition requirements of the first spectrum to determine a first matching degree of the first feature; compare the packaging information and interference information with the acquisition requirements of the second spectrum to determine a second matching degree of the second feature.

[0088] In one possible implementation, the fusion module 63 is specifically used to: perform feature screening on the first feature based on the first matching degree, and perform feature screening on the second feature based on the second matching degree; perform feature fusion on the screened first feature and the screened second feature according to the first matching degree and the second matching degree to obtain the spectral characteristics of the object to be detected.

[0089] In one possible implementation, the fusion module 63 is further used to: determine the first credibility of the first feature and the second credibility of the second feature based on the first matching degree and the second matching degree; and perform feature fusion on the screened first feature and the screened second feature according to the first credibility and the second credibility to obtain the spectral characteristics of the object to be detected.

[0090] In one possible implementation, the identification module 64 is specifically configured to: input the spectral features into a pre-trained hazardous chemical identification model to obtain component information of the object to be detected; and obtain a hazardous chemical identification result of the object to be detected based on the component information.

[0091] In one possible implementation, the identification module 64 is further used to: compare the component information with the component information of each known hazardous chemical in the hazardous chemical information database to obtain a comparison result; and obtain a hazardous chemical identification result of the item to be tested based on the component information and the comparison result.

[0092] In one possible implementation, the identification module 64 is also used to: determine the similarity between the composition information of the item to be detected and the composition information of each known hazardous chemical in the hazardous chemicals information database based on the comparison results; if there is a target known hazardous chemical whose similarity with the item to be detected exceeds a preset threshold, the item to be detected is determined to be a target known hazardous chemical; if there is no target known hazardous chemical whose similarity with the item to be detected exceeds the preset threshold, calculate the hazard level of the item to be detected based on the composition information of the item to be detected, and obtain a hazardous chemical identification result of the item to be detected based on the hazard level.

[0093] In a possible implementation, the identification module 64 is further configured to: determine the hazard category and hazard level of the item to be detected based on the hazardous chemical identification result; and issue corresponding alarm information for the item to be detected according to the hazard category and hazard level.

[0094] In the above embodiments, the descriptions of each embodiment have their own focus. For parts not described or recorded in detail in one embodiment, please refer to the relevant descriptions of other embodiments. Unless otherwise specified or there is a logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced to each other. The technical features of different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0095] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A method for identifying hazardous chemicals based on multispectral fusion, characterized in that: include: Performing feature extraction on the first spectrum and the second spectrum of the object to be detected, respectively, to obtain a first feature of the first spectrum and a second feature of the second spectrum; determining a first matching degree of the first feature and a second matching degree of the second feature according to the object feature of the object to be detected; Based on the first matching degree, performing feature screening on the first feature, and based on the second matching degree, performing feature screening on the second feature; performing feature fusion on the filtered first feature and the filtered second feature according to the first matching degree and the second matching degree to obtain a spectral feature of the object to be detected; The spectral features are input into a pre-trained hazardous chemical identification model to obtain a hazardous chemical identification result of the object to be detected.

2. The method for identifying hazardous chemicals based on multispectral fusion according to claim 1, characterized in that: The step of performing feature fusion on the filtered first feature and the filtered second feature according to the first matching degree and the second matching degree to obtain the spectral feature of the object to be detected includes: Determining a first credibility of the first feature and a second credibility of the second feature based on the first matching degree and the second matching degree; According to the first credibility and the second credibility, feature fusion is performed on the screened first feature and the screened second feature to obtain the spectral feature of the object to be detected.

3. The method for identifying hazardous chemicals based on multispectral fusion according to claim 1, characterized in that: The determining, based on the object feature of the object to be detected, a first matching degree of the first feature and a second matching degree of the second feature includes: Determining packaging information and interference information of the object to be detected based on the object characteristics of the object to be detected; comparing the packaging information and the interference information with the acquisition requirement of the first spectrum to determine a first matching degree of the first feature; The packaging information and the interference information are compared with the collection requirement of the second spectrum to determine a second matching degree of the second feature.

4. The method for identifying hazardous chemicals based on multispectral fusion according to claim 1, characterized in that: Inputting the spectral features into a pre-trained hazardous chemical identification model to obtain a hazardous chemical identification result of the object to be detected includes: Input the spectral features into a pre-trained hazardous chemical identification model to obtain the composition information of the object to be detected; Based on the component information, a hazardous chemical identification result of the item to be detected is obtained.

5. The method for identifying hazardous chemicals based on multispectral fusion according to claim 4, characterized in that: Obtaining a hazardous chemical identification result of the object to be detected based on the component information includes: Comparing the composition information with the composition information of each known hazardous chemical in the hazardous chemicals information database to obtain a comparison result; Based on the component information and the comparison result, a hazardous chemical identification result of the item to be detected is obtained.

6. The method for identifying hazardous chemicals based on multispectral fusion according to claim 5, characterized in that: Obtaining a hazardous chemical identification result of the object to be detected based on the component information and the comparison result includes: Determining, based on the comparison results, the similarity between the composition information of the object to be tested and the composition information of each known hazardous chemical in the hazardous chemicals information database; If there is a target known hazardous chemical whose similarity with the object to be detected exceeds a preset threshold, the object to be detected is determined to be the target known hazardous chemical; If there is no known target hazardous chemical whose similarity with the item to be detected exceeds a preset threshold, the hazard level of the item to be detected is calculated based on the component information of the item to be detected, and based on the hazard level, a hazardous chemical identification result of the item to be detected is obtained.

7. The method for identifying hazardous chemicals based on multispectral fusion according to claim 1, characterized in that: The method further comprises: Determine the hazard category and hazard level of the item to be detected based on the hazardous chemical identification result; According to the hazard category and the hazard level, corresponding alarm information of the object to be detected is issued.

8. The method for identifying hazardous chemicals based on multispectral fusion according to claim 1, characterized in that: The method further comprises: Performing spectral detection on the object to be detected to obtain an initial first spectrum and an initial second spectrum of the object to be detected; The initial first spectrum and the initial second spectrum are preprocessed to obtain a first spectrum and a second spectrum; wherein the preprocessing includes data denoising processing and baseline correction processing.

9. A hazardous chemicals identification device based on multi-spectral fusion, characterized in that: include: An extraction module, configured to extract features from the first spectrum and the second spectrum of the object to be detected, respectively, to obtain a first feature of the first spectrum and a second feature of the second spectrum; a determination module, configured to determine a first matching degree of the first feature and a second matching degree of the second feature according to the object feature of the object to be detected; a fusion module, configured to perform feature screening on the first feature based on the first matching degree, and perform feature screening on the second feature based on the second matching degree; performing feature fusion on the filtered first feature and the filtered second feature according to the first matching degree and the second matching degree to obtain a spectral feature of the object to be detected; The identification module is used to input the spectral characteristics into a pre-trained hazardous chemical identification model to obtain a hazardous chemical identification result of the object to be detected.

Images