Marine product automatic detection method and system

Abstract

The invention discloses a marine product automatic detection method and system, and relates to the field of image data processing. The method comprises the following steps: acquiring Raman spectrum data and near infrared spectrum data of the surface of a target object; judging whether the physical index of each side surface is qualified or not based on the near infrared spectrum data; when the physical indexes are qualified, the Raman spectrum data and the near infrared spectrum data are sent to a remote processing unit, and the Raman spectrum data and the near infrared spectrum data are fused; and based on the fusion result, detecting and analyzing the product quality of the corresponding side surface to obtain an analysis result. According to the scheme provided by the embodiment of the invention, the comprehensiveness of the marine product surface component detection result can be improved, and the accuracy of the detection result is improved.

Description

Technical Field

[0001] The present application relates to the field of image data processing, and in particular to a method and system for automatic detection of seafood. Background Art

[0002] Surface component testing is a key step in the seafood production process. For example, testing the surface moisture and protein content of marine fish is crucial. Existing techniques typically rely on conventional methods such as visual inspection and chemical reagent analysis for surface component analysis. However, these methods are inefficient, subjective, and potentially damaging to samples, making them difficult to meet the demands of large-scale industrial production. Furthermore, with increasing environmental pollution and increasingly serious food safety issues, existing testing methods lack the ability to detect specific compounds, significantly impacting the comprehensiveness and accuracy of seafood testing results.

[0003] Therefore, how to overcome the above-mentioned technical problems and defects becomes a key issue that needs to be solved. Summary of the Invention

[0004] In order to solve the above technical problems, the present application provides an automatic detection of seafood, which can improve the comprehensiveness of the detection results of seafood surface components and improve the accuracy of the detection results.

[0005] According to one aspect of the present application, a method for automatic detection of seafood is provided, the method comprising: The spectrum analyzer obtains Raman spectrum data and near-infrared spectrum data of the surface of the target object; the surface of the target object includes two sides, a first side surface and a second side surface; For each side, the spectrum analyzer determines whether the physical indicators of the corresponding side are qualified based on the near-infrared spectrum data; When the physical indicators are qualified, the Raman spectrum data and the near-infrared spectrum data are sent to a remote processing unit; A remote processing unit fuses the Raman spectrum data and the near-infrared spectrum data; Based on the fusion results, the product quality of the corresponding aspects is tested and analyzed to obtain the analysis results.

[0006] In the above solution, the method further includes: When the physical indicators are unqualified, the corresponding Raman spectrum data and near-infrared spectrum data are sent to the database for storage, and a notification is sent to the remote processing unit.

[0007] In the above solution, the fusion of the Raman spectrum data and the near-infrared spectrum data includes: Obtaining characteristic peak information of a target object based on the Raman spectrum data; Determining a first matching degree between the surface Raman spectrum of the target object and a standard Raman spectrum corresponding to the target compound based on the characteristic peak information; Based on the first matching degree, determining whether the matching degree of the Raman spectrum data meets the standard, and obtaining a first determination result; When the first judgment result indicates that the characterization is up to standard, performing data layer fusion on the Raman spectrum data and the near-infrared spectrum data; When the first judgment result indicates that the characterization does not meet the standard, feature layer fusion is performed on the Raman spectrum data and the near-infrared spectrum data.

[0008] In the above solution, the data layer fusion of the Raman spectrum data and the near-infrared spectrum data includes: The Raman spectrum data and the near-infrared spectrum data are spliced ​​into a combined spectrum.

[0009] In the above solution, based on the fusion results, the product quality of the corresponding aspects is tested and analyzed to obtain the analysis results, including: Determining the similarity between the combined spectrum and the standard combined spectrum of the target compound to obtain a second matching degree; Based on the first matching degree and the second matching degree, it is determined whether the corresponding side surface contains the target compound.

[0010] In the above solution, the feature layer fusion of the Raman spectrum data and the near-infrared spectrum data includes: extracting key features of the Raman spectral data and the near-infrared spectral data respectively; Using the key features and the quantitative model, the estimated concentration of the target compound of the corresponding aspect is determined.

[0011] In the above solution, based on the fusion results, the product quality of the corresponding aspects is tested and analyzed to obtain the analysis results, including: Based on the first matching degree and the estimated concentration, it is determined whether the corresponding side contains the target compound.

[0012] In the above solution, obtaining Raman spectrum data and near-infrared spectrum data of the surface of the target object includes: Acquire first Raman spectrum data and first near-infrared spectrum data of a first side surface of a target object at a first station; The target object is transferred to the second station by using the first conveyor belt, the target object is flipped from the first side facing upward to the second side facing upward by using the flipping mechanism of the second station, and the flipped target object is conveyed to the second conveyor belt; Using the second conveyor belt to transfer the target object to a third workstation; At the third station, second Raman spectrum data and second near-infrared spectrum data of the second side surface of the target object are acquired.

[0013] According to another aspect of the present application, there is provided a seafood automatic detection system, the system comprising a first conveyor belt, a turnover mechanism, a second conveyor belt, a spectral analysis unit, and a remote processing unit; a first conveyor belt, configured to transport the target object to the flipping mechanism with the first side facing upward; A flipping mechanism, configured to flip the target object from a first side-up state to a second side-up state, and convey the target object to a second conveyor belt; the flipping mechanism is disposed between the first conveyor belt and the second conveyor belt; a second conveyor belt, configured to convey the target object to a downstream device with the second side facing upward; a spectral analysis unit for collecting Raman spectral data and near-infrared spectral data of target objects on the first conveyor belt and the second conveyor belt; and for determining whether the physical indicators of the corresponding side surfaces are qualified based on the near-infrared spectral data, and sending the obtained Raman spectral data and near-infrared spectral data to the remote processing unit if qualified; The processing unit is used to fuse the received Raman spectrum data and near-infrared spectrum data; based on the fusion result, the product quality of the corresponding aspect is detected and analyzed to obtain the analysis result.

[0014] In the above scheme, the flipping mechanism includes a driving motor, a mounting base and two groups of clamping assemblies; the two groups of clamping assemblies are fixedly connected to the mounting base and are relatively arranged on both sides of the mounting base, and the two groups of clamping assemblies are arranged along the transmission direction of the target object; the driving motor is transmission-connected to the mounting base and drives the mounting base to rotate circumferentially along the transmission direction; each clamping assembly includes a fixed baffle and an elastic baffle arranged in parallel, and when the clamping assembly rotates to the side of the first conveyor belt, the fixed baffle is located below the elastic baffle; the first end of the fixed baffle is fixedly connected to the mounting base, the first end of the elastic baffle is rotationally connected to the mounting base, and a tension spring is provided on the side of the first end away from the mounting base, and the two ends of the tension spring are respectively connected to the elastic baffle and the mounting base.

[0015] The automatic seafood detection method and system provided in this application analyzes the surface chemical composition of seafood by utilizing the fusion results of Raman spectroscopy and near-infrared spectroscopy data, and can perform full-dimensional quality analysis from the perspective of physical and chemical properties, thereby improving the comprehensiveness and accuracy of the seafood surface composition detection results; further, by utilizing a fully automatic flow and detection process to realize the detection of both sides of the seafood surface, it is possible to achieve comprehensive detection of the seafood surface on the basis of improving the detection efficiency, thereby improving the integrity of the data source, reducing the risk of missed detection, and improving the accuracy of the detection results.

[0016] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the specification are used to explain the principles of the present application, and do not constitute an improper limitation on the present application.

[0018] Figure 1 A schematic flow chart of a method for automatically detecting seafood provided in an embodiment of the present application; Figure 2 This is a flow chart of S104 in the seafood automatic detection method according to an embodiment of the present application; Figure 3 Schematic diagram of the standard Raman spectrum in the automatic detection method of seafood in the embodiment of the present application; Figure 4 A schematic diagram of the architecture of an automatic seafood detection system provided in an embodiment of the present application; Figure 5 This is a structural diagram of the transmission system in the seafood automatic detection system according to an embodiment of the present application; Figure 6 for Figure 5 A partial cross-sectional view of . DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0020] The embodiment of the present application provides a method for automatically detecting seafood, which is applied to electronic equipment, and can be specifically applied to an automatic seafood detection system, such as Figure 1 As shown, the method may include S101 to S105. S101 to S105 are described in detail below in conjunction with specific embodiments.

[0021] S101: A spectrum analyzer obtains Raman spectrum data and near-infrared spectrum data of a target object surface; the target object surface includes two sides, a first side surface and a second side surface.

[0022] In actual application, the target object is the seafood to be detected, such as marine fish, which can also be called target seafood.

[0023] In actual application, the spectrum analyzer may include a Raman spectrum acquisition module and a near-infrared spectrum acquisition module; when the target object moves to the position of the spectrum analyzer, the upward side of the target object serves as the detection surface, the Raman spectrum acquisition module collects Raman spectrum data of the detection surface, and the near-infrared spectrum acquisition module collects near-infrared spectrum data of the detection surface.

[0024] In one embodiment, acquiring Raman spectrum data and near-infrared spectrum data of the surface of the target object may include: Acquire first Raman spectrum data and first near-infrared spectrum data of a first side surface of a target object at a first station; The target object is transferred to the second station by using the first conveyor belt, the target object is flipped from the first side facing upward to the second side facing upward by using the flipping mechanism of the second station, and the flipped target object is conveyed to the second conveyor belt; Using the second conveyor belt to transfer the target object to a third workstation; At the third station, second Raman spectrum data and second near-infrared spectrum data of the second side surface of the target object are acquired.

[0025] In actual application, the first workstation can be set on the side of the first conveyor belt, and the second workstation can be set on the side of the second conveyor belt; there can be two spectrometers, namely the first spectrometer and the second spectrometer; the first spectrometer is set at the first workstation to obtain first Raman spectrum data and first near-infrared spectrum data; the second spectrometer is set at the second workstation to obtain second Raman spectrum data and second near-infrared spectrum data.

[0026] Here, by automatically flipping the target object and automatically inspecting both sides, a comprehensive inspection of the target product surface can be achieved. Compared with inspecting only one side, the reliability and accuracy of the inspection results are improved. In particular, for local problems such as point defects, double-sided inspection can effectively reduce the missed detection rate.

[0027] S102: For each side surface, the spectrum analyzer determines whether the physical indicators of the corresponding side surface are qualified based on the near-infrared spectrum data.

[0028] In actual application, the near-infrared spectral data can be used to judge whether physical indicators such as moisture and texture are qualified; physical indicators can also be called physical parameters.

[0029] S103: When the physical indicators are qualified, the Raman spectrum data and the near-infrared spectrum data are sent to a remote processing unit.

[0030] In actual application, the remote processing unit may be a processing unit of a remote control center, specifically, a control center of a seafood processing system, more specifically, a processing unit of a control center, such as an industrial computer.

[0031] In actual application, preliminary judgment and analysis of data, namely physical indicator evaluation, is performed on local equipment (spectrum analyzer), and data fusion and analysis are performed by remote processing center to achieve collaborative work of local computing and remote processing. This not only enables rapid completion of preliminary processing of spectral data and rapid decision-making, but also reduces the amount of data transmitted to the control center and reduces data processing pressure, thereby ensuring the smooth and reliable operation of the entire monitoring system.

[0032] In actual application, the physical indicators may include moisture content, etc. When the physical indicators are unqualified, the target object can be determined to be an unqualified product, and the target object or the current batch of products can be recycled or destroyed in downstream equipment.

[0033] In actual application, the spectrum analyzers at the first and third stations can start analysis immediately after obtaining the spectral data; when the physical indicator test results of the spectrum analyzer at the first station are unqualified, an alarm can be issued, and the spectral data will no longer be collected when the product is transferred to the third station, and the unqualified product processing procedures can be carried out in downstream equipment, such as recycling or destruction.

[0034] In one embodiment, the method may further include: When the physical indicators are unqualified, the corresponding Raman spectrum data and near-infrared spectrum data are sent to the database for storage, and a notification is sent to the remote processing unit.

[0035] Here, the data of unqualified products is stored in the database to realize data archiving, which is convenient for subsequent retrieval and viewing; notifications are sent to the remote processing unit so that on-site staff can be informed of abnormal situations in a timely manner; at the same time, the remote processing unit can automatically generate or generate control instructions for unqualified products under the operation of on-site staff, such as transferring the target object to the recycling channel.

[0036] S104: The remote processing unit fuses the Raman spectrum data and the near-infrared spectrum data.

[0037] In one embodiment, if Figure 2 As shown, fusing the Raman spectrum data and the near-infrared spectrum data, namely S104, may include S201 to S205.

[0038] S201: Obtain characteristic peak information of a target object based on the Raman spectrum data.

[0039] In practical applications, characteristic peak information may include characteristic peak position (such as the C=C peak at 1650 cm⁻¹), peak width, and polarization degree.

[0040] S202: Determine a first matching degree between the Raman spectrum of the surface of the target object and a standard Raman spectrum corresponding to the target compound based on the characteristic peak information.

[0041] In practical applications, the first matching degree between the Raman spectrum and the standard Raman spectrum can reflect the similarity of their characteristic peaks; in practical applications, the first matching degree can be determined based on the degree of shift between the characteristic peaks in the target object Raman spectrum and the standard Raman spectrum.

[0042] Based on this, in one embodiment, determining a first matching degree between the target object surface Raman spectrum and a standard Raman spectrum corresponding to the target compound based on the characteristic peak information, that is, S202, may include: Based on the characteristic peak information, determining all target characteristic peaks in the Raman spectrum of the target object surface; Determine the standard characteristic peak corresponding to each target characteristic peak in the standard Raman spectrum corresponding to the target compound; Determine whether the deviation between each target characteristic peak and the corresponding standard characteristic peak is lower than a preset deviation threshold, and if lower than the deviation threshold, use the corresponding target characteristic peak as the deviation peak position; A first matching degree is determined based on the ratio of the deviation peak positions in all target characteristic peaks.

[0043] Here, the standard Raman spectrum may be pre-configured based on Raman spectrum data of seafood containing target compound residues.

[0044] In actual application, the database can be pre-configured with standard Raman spectra corresponding to different compounds, that is, the Raman spectrum when the target compound is contained in seafood. When the target chemical needs to be detected, the corresponding standard Raman spectrum is called from the database for analysis; here, the target chemical can be one of the pre-configured surface chemicals of the product that needs to be tested for safety; for example, the target compound is malachite green, and the standard Raman spectrum of malachite green is as follows: Figure 3 shown.

[0045] For example, the calculation formula of the first matching degree is: ; in, Indicates the first matching degree, Indicates the number of deviation peaks, Indicates the total amount of characteristic peaks.

[0046] S203: Based on the first matching degree, determine whether the matching degree of the Raman spectrum data meets the requirements, and obtain a first determination result.

[0047] In actual application, it can be determined whether the first matching degree reaches a preset matching degree threshold. If it exceeds the threshold, the first judgment result indicates that the matching degree of the Raman spectral data meets the standard. Correspondingly, if it does not reach the threshold, the first judgment result indicates that the matching degree of the Raman spectral data does not meet the standard. Exemplarily, the matching degree threshold is configured to be 90%. When the first matching degree is greater than or equal to 90%, the matching degree of the Raman spectral data meets the standard, that is, at this time, the analysis is mainly based on the Raman spectrum. Correspondingly, when the first matching degree is less than 90%, the matching degree of the Raman spectral data does not meet the standard. At this time, the characteristic layer fusion results of the Raman spectrum and the near-infrared spectrum are analyzed.

[0048] S204: When the first judgment result indicates that the characterization is up to standard, data layer fusion is performed on the Raman spectrum data and the near-infrared spectrum data.

[0049] In actual application, when the first judgment result is that it meets the standard, it means that the Raman spectrum is sufficient. At this time, the Raman spectrum and the near-infrared spectrum can be directly spliced.

[0050] Based on this, in one embodiment, the data layer fusion of the Raman spectrum data and the near-infrared spectrum data, i.e., S204, may include: The Raman spectrum data and the near-infrared spectrum data are spliced ​​into a combined spectrum.

[0051] Based on this, in one embodiment, splicing the Raman spectrum data and the near-infrared spectrum data into a combined spectrum may include: performing spectral range alignment processing on the Raman spectrum data and the near-infrared spectrum data; Normalize the aligned Raman spectral data and near-infrared spectral data; The normalized Raman spectral data and near-infrared spectral data are fused by dimensionality reduction to construct a multidimensional feature vector.

[0052] In practical applications, spectral range alignment processing can unify the wavenumber ranges of Raman and near-infrared spectra to ensure consistent data dimensions. Specifically, the Raman and near-infrared spectra can be beam-intercepted first, and then the intercepted results can be difference-aligned to achieve dimensional unification.

[0053] In practical applications, the standard normal variate transformation (SNV) can be used for normalization processing to eliminate the influence of the difference in illumination intensity during spectral acquisition and make the spectra of different samples comparable.

[0054] For example, the following formula can be used for SNV normalization: ; in, represents the original intensity value of wave number point i, represents the mean of all wave number points, represents the standard deviation of all wavenumber points.

[0055] In practical applications, when performing dimensionality reduction and fusion, the principal components of the two types of spectra can be extracted through the principal component analysis (PCA) method, and then the extracted principal components can be feature merged; specifically, the Raman spectral data and the near-infrared spectral data can be first normalized separately, and then the standardized covariance matrix can be calculated, and then the eigenvalues ​​and eigenvectors of the covariance matrix can be solved, and the eigenvalues ​​can be sorted from large to small to achieve eigenvalue decomposition. Then, the first k principal components are selected, where k is an integer greater than 1. Finally, the principal components corresponding to the Raman spectral data and the principal components corresponding to the near-infrared spectral data are spliced ​​into a multidimensional feature vector.

[0056] S205: When the first judgment result indicates that the characterization does not meet the standard, feature layer fusion is performed on the Raman spectrum data and the near-infrared spectrum data.

[0057] In actual citation, when the judgment result is characterized by failure to meet the standards, it means that the Raman spectrum data is insufficient. At this time, the wide spectrum characteristics of the near-infrared spectrum can be used to perform feature layer fusion of the Raman spectrum and the near-infrared spectrum.

[0058] Based on this, in one embodiment, if the first judgment result is not up to standard, the feature layer fusion of the Raman spectrum data and the near-infrared spectrum data, i.e., S205, may include: extracting key features of the Raman spectral data and the near-infrared spectral data respectively; Using the key features and the quantitative model, the estimated concentration of the target compound of the corresponding aspect is determined.

[0059] In practical applications, the HSIC-VSIO algorithm can be used to extract key features of Raman and near-infrared spectral data. Specifically, characteristic peak information of the Raman spectrum (such as the C=C peak value) and partial least squares (PLS) loading vectors in the near-infrared spectrum can be extracted respectively, and key variables, i.e., key features, can be screened through analysis methods such as the random forest algorithm and canonical correlation analysis (CCA). Here, since redundant spectral segments (such as background signals or irrelevant peaks) are eliminated, the significance of high-information-density regions can be increased, thereby improving the efficiency and accuracy of feature extraction results.

[0060] In practical applications, the quantitative model can adopt a dual-channel deep learning network (CNN) model. In the dual-channel CNN model, one channel is the Raman spectrum input layer, and the other channel is the near-infrared spectrum input layer. That is, the dual-channel CNN model includes a Raman CNN sub-model and a near-infrared CNN sub-model; the dual-channel CNN model uses the attention mechanism to optimize feature interaction, fuses the key features of the Raman spectrum and the near-infrared spectrum, and outputs the prediction results.

[0061] In actual application, when the first judgment result is not up to standard, it means that the Raman spectrum is insufficient. At this time, by using the near-infrared spectrum to supplement the information, for example, using the wide-spectrum absorption characteristics of the near-infrared to enhance the signal, the fusion of the Raman spectrum and the near-infrared spectrum at the characteristic layer can be achieved. In this way, the high spatial resolution of the Raman spectrum and the wide-spectrum coverage characteristics of the near-infrared can be combined to make up for the defects of the Raman spectrum data and ensure the reliability of the detection results.

[0062] S105: Based on the fusion result, the product quality of the corresponding aspect is tested and analyzed to obtain the analysis result.

[0063] In one embodiment, when the first judgment result indicates that the quality of the product on the corresponding side is up to standard, that is, when the Raman spectrum data is sufficient, the step of performing a test and analysis on the product quality on the corresponding side based on the fusion result to obtain the analysis result, that is, S105, may include: Determining the similarity between the combined spectrum and the standard combined spectrum of the target compound to obtain a second matching degree; Based on the first matching degree and the second matching degree, it is determined whether the corresponding side surface contains the target compound.

[0064] In practical applications, the standard combination spectrum can be pre-configured based on the Raman spectrum and near-infrared spectrum combination of seafood containing target compound residues.

[0065] Here, when there are sufficient Raman spectra, the spliced ​​combined spectrum can be matched with the standard combined spectrum of known compounds in the database to calculate the similarity between the two. Specifically, the cosine similarity can be used to represent the similarity between the combined spectrum and the standard combined spectrum at the same wave number point. For example, the cosine similarity can be calculated using the following formula: ; in, Represents cosine similarity, the second matching degree, Indicates the standard spectrum at wave number point The intensity value of Indicates the combined spectrum at wave number point The intensity value of .

[0066] In actual application, based on the first matching degree and the second matching degree, when judging whether the corresponding side contains the target compound, it can be judged whether the first matching degree and the second matching degree meet the preset conditions when the Raman spectrum is sufficient, for example, the first matching degree is greater than the preset first threshold and the second matching degree is greater than the preset second threshold; when the preset conditions are met, it is determined that the target object contains the target compound; illustratively, the corresponding preset conditions when the Raman spectrum is sufficient are that the first matching degree is greater than or equal to 95%, and the second matching degree is greater than or equal to 85%. When this condition is met, it is judged that the target object contains the target chemical malachite green, otherwise, it does not contain it.

[0067] In one embodiment, when the first judgment result indicates that the product quality of the corresponding aspect is not up to standard, that is, when the Raman spectrum data is insufficient, based on the first judgment result, in one embodiment, the fusion result is used to detect and analyze the product quality of the corresponding aspect to obtain the analysis result, that is, S105, which may include: Based on the first matching degree and the estimated concentration, it is determined whether the corresponding side contains the target compound.

[0068] In actual application, based on the first matching degree and the estimated concentration, when judging whether the corresponding side contains the target compound, it can be judged whether the first matching degree and the estimated concentration meet the preset conditions when the Raman spectrum is insufficient, for example, the first matching degree is greater than the preset third threshold and the estimated concentration is greater than the fourth threshold; when the preset conditions are met, it is judged that the target object contains the target compound; illustratively, the corresponding preset condition when the Raman spectrum is insufficient is that the first matching degree is greater than or equal to 75%, and the estimated concentration is greater than or equal to 5.2 μg / kg. When this condition is met, it is determined that the target object contains the target compound malachite green, otherwise it is not contained.

[0069] Here, by dynamically selecting different fusion strategies for scenes with sufficient and insufficient Raman spectra, adaptive optimization of the detection system is achieved, thereby improving the overall matching degree between the detection method described in the embodiment of the present application and various scenes, thereby improving the accuracy and reliability of the detection results in different scenes.

[0070] In actual application, each aspect is analyzed separately, that is, whether each aspect contains the target compound is determined. When either of the two aspects contains the target compound, it is determined that the target object contains the target compound.

[0071] In actual application, both sides are tested, which can improve the comprehensiveness of the analysis data and the accuracy of the test results. At the same time, the quality of seafood can be tested based on the difference in near-infrared spectral data of the two sides.

[0072] Therefore, in one embodiment, the method may further include: Determining a difference in a physical parameter between the first side surface and the second side surface based on the first near-infrared spectrum data and the second near-infrared spectrum data to obtain a degree of difference in the physical parameter; It is determined whether the physical difference is greater than a difference threshold, and if it is greater than the difference threshold, the target object is determined to be a non-compliant product.

[0073] Here, the difference in physical parameters can be the difference in moisture content and texture parameters (such as excited acid freshness); the difference threshold can be pre-set according to the type of seafood to be tested. For example, for marine fish, the difference threshold is configured to 5%. When the physical difference between the two sides is greater than 5%, the batch of marine fish may be marine fish that have been injected with water-retaining agents to falsify freshness. The output judgment result is an illegal product, and an abnormal alarm can be issued.

[0074] Here, by using the difference between the two sides to determine whether seafood has illegally added chemical ingredients, abnormal seafood can be screened out when the physical parameters on both sides meet the standards. This ensures that while using spectral data to detect visible compounds, it can also detect illegal chemical components that are invisible because they are dissolved in the body tissues of seafood, thereby improving the comprehensiveness of seafood testing and ensuring the food safety of seafood.

[0075] In summary, the automatic seafood detection method provided in the embodiment of the present application analyzes the chemical composition of the seafood surface by utilizing the fusion results of Raman spectroscopy and near-infrared spectroscopy data, and can perform full-dimensional quality analysis from the perspective of physical and chemical properties, thereby improving the comprehensiveness and accuracy of the seafood surface composition detection results; further, by utilizing a fully automatic flow and detection process to realize the detection of both sides of the seafood surface, it is possible to achieve comprehensive detection of the seafood surface on the basis of improving the detection efficiency, thereby improving the integrity of the data source, reducing the risk of missed detection, and improving the accuracy of the detection results.

[0076] In order to implement the above method, the embodiment of the present application also provides a seafood automatic detection system, such as Figure 4 As shown, the system includes a transmission unit 401, a spectrum analysis unit 402 and a remote processing unit 403; Figure 5 As shown, the transmission unit 401 includes a first conveyor belt 501, a turning mechanism, and a second conveyor belt 502; The first conveyor belt 501 is used to transport the target object to the flipping mechanism with the first side facing upward; A flipping mechanism, configured to flip the target object from a first side facing upward to a second side facing upward, and convey the target object to the second conveyor belt 502; the flipping mechanism is disposed between the first conveyor belt 501 and the second conveyor belt 502; A second conveyor belt 502 is used to convey the target object to a downstream device with the second side facing upward; a spectral analysis unit for collecting Raman spectral data and near-infrared spectral data of target objects on the first conveyor belt 501 and the second conveyor belt 502; and for determining whether the physical indicators of the corresponding side surface are qualified based on the near-infrared spectral data, and sending the obtained Raman spectral data and near-infrared spectral data to the remote processing unit if qualified; The processing unit is used to fuse the received Raman spectrum data and near-infrared spectrum data; based on the fusion result, the product quality of the corresponding aspect is detected and analyzed to obtain the analysis result.

[0077] In actual application, the first workstation can be set on the side of the first conveyor belt 501, and the second workstation can be set on the side of the second conveyor belt 502; there can be two spectrometers, namely a first spectrometer and a second spectrometer; the first spectrometer is set at the first workstation to obtain first Raman spectrum data and first near-infrared spectrum data of the first side; the second spectrometer is set at the second workstation to obtain second Raman spectrum data and second near-infrared spectrum data of the second side.

[0078] In one embodiment, if Figure 6 As shown, the flip mechanism may include a drive motor 506, a mounting base 503 and two sets of clamping assemblies; the two sets of clamping assemblies are fixedly connected to the mounting base 503 and are relatively arranged on both sides of the mounting base 503, and the two sets of clamping assemblies are arranged along the transmission direction of the target object; the drive motor 506 is transmission-connected to the mounting base 503 and drives the mounting base 503 to rotate circumferentially along the transmission direction; each clamping assembly includes a fixed baffle 505 and an elastic baffle 504 arranged in parallel, and when the clamping assembly rotates to one side of the first conveyor belt 501, the fixed baffle 505 is located on the elastic baffle 504 Below; the first end of the fixed baffle 505 is fixedly connected to the mounting base 503, the first end of the elastic baffle 504 is rotatably connected to the mounting base 503, and a tension spring 507 is provided on the side of the first end away from the mounting base 503, and the two ends of the tension spring 507 are respectively connected to the elastic baffle 504 and the mounting base 503; when the clamping assembly is located on one side of the first conveyor belt 501, the second ends of the two baffles of the clamping assembly constitute the input port of the target object, and when the clamping assembly is located on one side of the second conveyor belt 502, the second ends of the two baffles of the clamping assembly constitute the output port of the target object.

[0079] In actual application, the fixed baffles 505 of the two sets of clamping components can be integrally formed with the mounting base 503, such as Figure 5As shown; the first end of the elastic baffle 504 is rotatably connected to the mounting base 503 via a rotating shaft; the fixed baffle 505 and the elastic baffle 504 are of the same length and are smaller than the length of the seafood to be inspected (i.e., the target object), to ensure that when the seafood enters and leaves the clamping mechanism, there is still a part on the surface of the seafood that is in contact with the conveyor belt, so that it can move under the drive of the conveyor belt.

[0080] In actual application, the two groups of clamping components are respectively the first clamping component and the second clamping component, and the first clamping component and the second clamping component are respectively arranged on both sides of the mounting base 503, and the fixed baffle 505 of the first clamping component and the elastic baffle 504 of the second clamping component are located in the same plane, and the elastic baffle 504 of the first clamping component and the fixed baffle 505 of the second clamping component are located in the same plane; the edges of the two groups of clamping components on the side opposite to the conveyor belt are parallel to the edge of the conveyor belt on that side, and the baffle plane located below is lower than the conveying plane of the conveyor belt on the corresponding side; specifically, when the first clamping component is located on one side of the first conveyor belt 501, the upper plane of the fixed baffle 505 of the first clamping component is lower than the upper surface of the first conveyor belt 501, thereby ensuring the smoothness of the target object when it slides from the first conveyor belt 501 to the first clamping component.

[0081] The working principle of the turnover mechanism is described below.

[0082] When seafood (target object) enters the end of the first conveyor belt 501, the first clamping assembly, acting as an input mechanism, is located on the side facing the first conveyor belt 501, while the second clamping assembly, acting as an output mechanism, is located on the side facing the second conveyor belt 502. Driven by the first conveyor belt 501, the seafood enters the first clamping mechanism baffle. Since the elastic baffle 504 is provided with a tension spring 507, it can accommodate seafood of different sizes (thicknesses) and ensure the clamping effect. When the seafood completely enters the first clamping assembly, the drive motor 506 drives the flipping mechanism to rotate, causing the first clamping mechanism to flip upward and rotate toward the second conveyor belt 502. At this time, the first conveyor belt 501 becomes the output mechanism, while the second clamping mechanism is flipped to face the first conveyor belt 501, becoming the input mechanism. The seafood in the first clamping mechanism enters the second conveyor belt 502 and, driven by the second conveyor belt 502, leaves the first clamping structure. At this time, the second clamping assembly, acting as an input mechanism, begins to receive the seafood conveyed by the first conveyor belt 501 and flips according to the above steps.

[0083] Here, through the continuous flipping operation of the two groups of clamping components, that is, when one group receives seafood from the first conveyor belt 501, the other group transmits seafood to the second conveyor belt 502, not only can a comprehensive inspection of the complete surface of the seafood be achieved, but also continuous and uninterrupted operation can be achieved, thereby improving the inspection and transmission efficiency.

[0084] In actual application, the second end of the fixed baffle 505 (i.e., the section close to the conveyor belt) is provided with a wedge-shaped structure, and the thickness of the wedge-shaped structure decreases from the second end to the first end, forming a limiting slope to prevent the seafood from slipping out during the flipping process; the second end of the elastic baffle 504 is provided with a wedge-shaped structure, and the thickness of the wedge-shaped structure increases from the second end to the first end, forming a sliding slope. When the clamping assembly flips to the side of the second conveyor belt 502, the elastic baffle 504 is located below. At this time, the tension spring 507 tilts downward under the action of the gravity of the seafood, and slides into the second conveyor belt 502 from the sliding slope. At this time, the sliding slope can improve the smoothness of the seafood when sliding to the second conveyor belt 502.

[0085] In one embodiment, the processing unit may further be configured to: When the physical indicators are unqualified, the corresponding Raman spectrum data and near-infrared spectrum data are sent to the database for storage, and a notification is sent to the remote processing unit.

[0086] In one embodiment, when fusing the Raman spectrum data and the near-infrared spectrum data, the processing unit is specifically configured to: Obtaining characteristic peak information of a target object based on the Raman spectrum data; Determining a first matching degree between the surface Raman spectrum of the target object and a standard Raman spectrum corresponding to the target compound based on the characteristic peak information; Based on the first matching degree, determining whether the matching degree of the Raman spectrum data meets the standard, and obtaining a first determination result; When the first judgment result indicates that the characterization is up to standard, performing data layer fusion on the Raman spectrum data and the near-infrared spectrum data; When the first judgment result indicates that the characterization does not meet the standard, feature layer fusion is performed on the Raman spectrum data and the near-infrared spectrum data.

[0087] In one embodiment, when performing data layer fusion on the Raman spectrum data and the near-infrared spectrum data, the processing unit is specifically configured to: The Raman spectrum data and the near-infrared spectrum data are spliced ​​into a combined spectrum.

[0088] In one embodiment, based on the fusion result, the product quality of the corresponding aspect is detected and analyzed. When the analysis result is obtained, the processing unit can be specifically used to: Determining the similarity between the combined spectrum and the standard combined spectrum of the target compound to obtain a second matching degree; Based on the first matching degree and the second matching degree, it is determined whether the corresponding side surface contains the target compound.

[0089] In one embodiment, when performing feature layer fusion on the Raman spectrum data and the near-infrared spectrum data, the processing unit may be specifically configured to: extracting key features of the Raman spectral data and the near-infrared spectral data respectively; Using the key features and the quantitative model, the estimated concentration of the target compound of the corresponding aspect is determined.

[0090] In one embodiment, based on the fusion result, the product quality of the corresponding aspect is detected and analyzed. When the analysis result is obtained, the processing unit can be specifically used to: Based on the first matching degree and the estimated concentration, it is determined whether the corresponding side contains the target compound.

[0091] In summary, the automatic seafood detection system provided in the embodiment of the present application analyzes the chemical composition of the surface of seafood by utilizing the fusion results of Raman spectroscopy and near-infrared spectroscopy data, and can perform full-dimensional quality analysis from the perspective of physical and chemical properties, thereby improving the comprehensiveness and accuracy of the seafood surface composition detection results; further, by utilizing a fully automatic flow and detection process to realize the detection of both sides of the seafood surface, it is possible to achieve comprehensive detection of the seafood surface on the basis of improving the detection efficiency, thereby improving the integrity of the data source, reducing the risk of missed detection, and improving the accuracy of the detection results.

[0092] It should be noted that the automatic seafood detection system provided in the above embodiment is only illustrated by the division of the above-mentioned program modules when performing automatic seafood detection. In actual applications, the above-mentioned processing can be assigned to different program modules as needed, that is, the internal structure of the device can be divided into different program modules to complete all or part of the above-mentioned processing. In addition, the automatic seafood detection system provided in the above embodiment and the automatic seafood detection method embodiment are based on the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0093] It should be noted that: "first", "second", etc. are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0094] In addition, the technical solutions described in the embodiments of the present application can be arbitrarily combined without conflict.

[0095] The above description is merely a preferred embodiment of the present application and is not intended to limit the scope of protection of the present application.

Claims

1. A method for automatic detection of seafood, characterized in that: The method comprises: The spectrum analyzer obtains Raman spectrum data and near-infrared spectrum data of the surface of the target object; the surface of the target object includes two sides, a first side surface and a second side surface; For each side, the spectrum analyzer determines whether the physical indicators of the corresponding side are qualified based on the near-infrared spectrum data; When the physical indicators are qualified, the Raman spectrum data and the near-infrared spectrum data are sent to a remote processing unit; A remote processing unit fuses the Raman spectrum data and the near-infrared spectrum data; Based on the fusion results, the product quality of the corresponding aspects is tested and analyzed to obtain the analysis results.

2. The method according to claim 1, characterized in that The method further comprises: When the physical indicators are unqualified, the corresponding Raman spectrum data and near-infrared spectrum data are sent to the database for storage, and a notification is sent to the remote processing unit.

3. The method according to claim 1, wherein fusing the Raman spectral data and the near-infrared spectral data comprises: Obtaining characteristic peak information of a target object based on the Raman spectrum data; Determining a first matching degree between the surface Raman spectrum of the target object and a standard Raman spectrum corresponding to the target compound based on the characteristic peak information; Based on the first matching degree, determining whether the matching degree of the Raman spectrum data meets the standard, and obtaining a first determination result; When the first judgment result indicates that the characterization is up to standard, performing data layer fusion on the Raman spectrum data and the near-infrared spectrum data; When the first judgment result indicates that the characterization does not meet the standard, feature layer fusion is performed on the Raman spectrum data and the near-infrared spectrum data.

4. The method according to claim 3, characterized in that The performing data layer fusion on the Raman spectrum data and the near-infrared spectrum data includes: The Raman spectrum data and the near-infrared spectrum data are spliced ​​into a combined spectrum.

5. The method according to claim 4, wherein the detecting and analyzing the product quality of the corresponding aspect based on the fusion result to obtain the analysis result comprises: Determining the similarity between the combined spectrum and the standard combined spectrum of the target compound to obtain a second matching degree; Based on the first matching degree and the second matching degree, it is determined whether the corresponding side surface contains the target compound.

6. The method according to claim 3, characterized in that The performing feature layer fusion on the Raman spectrum data and the near-infrared spectrum data includes: extracting key features of the Raman spectral data and the near-infrared spectral data respectively; Using the key features and the quantitative model, the estimated concentration of the target compound of the corresponding aspect is determined.

7. The method according to claim 6, characterized in that Based on the fusion results, the product quality of the corresponding aspects is tested and analyzed to obtain analysis results, including: Based on the first matching degree and the estimated concentration, it is determined whether the corresponding side contains the target compound.

8. The method according to any one of claims 1 to 7, characterized in that The obtaining of Raman spectrum data and near-infrared spectrum data of the surface of the target object includes: Acquire first Raman spectrum data and first near-infrared spectrum data of a first side surface of a target object at a first station; The target object is transferred to the second station by using the first conveyor belt, the target object is flipped from the first side facing upward to the second side facing upward by using the flipping mechanism of the second station, and the flipped target object is conveyed to the second conveyor belt; Using the second conveyor belt to transfer the target object to a third workstation; At the third station, second Raman spectrum data and second near-infrared spectrum data of the second side surface of the target object are acquired.

9. A seafood automatic detection system, characterized in that: The system includes a first conveyor belt, a turnover mechanism, a second conveyor belt, a spectral analysis unit, and a remote processing unit; a first conveyor belt, configured to transport the target object to the flipping mechanism with the first side facing upward; A flipping mechanism, configured to flip the target object from a first side-up state to a second side-up state, and convey the target object to a second conveyor belt; the flipping mechanism is disposed between the first conveyor belt and the second conveyor belt; a second conveyor belt, for conveying the target object to a downstream device with the second side facing upward; a spectral analysis unit for collecting Raman spectral data and near-infrared spectral data of target objects on the first conveyor belt and the second conveyor belt; and for determining whether the physical indicators of the corresponding side surfaces are qualified based on the near-infrared spectral data, and sending the obtained Raman spectral data and near-infrared spectral data to the remote processing unit if qualified; A processing unit, configured to perform data fusion on the received Raman spectrum data and near-infrared spectrum data; Based on the fusion results, the product quality of the corresponding aspects is tested and analyzed to obtain the analysis results.

10. The system according to claim 9, characterized in that The flipping mechanism includes a driving motor, a mounting base and two groups of clamping assemblies; the two groups of clamping assemblies are fixedly connected to the mounting base and are relatively arranged on both sides of the mounting base, and the two groups of clamping assemblies are arranged along the transmission direction of the target object; the driving motor is transmission-connected to the mounting base and drives the mounting base to rotate circumferentially along the transmission direction; each clamping assembly includes a fixed baffle and an elastic baffle arranged in parallel, and when the clamping assembly rotates to one side of the first conveyor belt, the fixed baffle is located below the elastic baffle; the first end of the fixed baffle is fixedly connected to the mounting base, the first end of the elastic baffle is rotationally connected to the mounting base, and a tension spring is provided on the side of the first end away from the mounting base, and the two ends of the tension spring are respectively connected to the elastic baffle and the mounting base.

Images