A food detection method and system based on surface-enhanced Raman spectroscopy and visual recognition

Abstract

This invention relates to a food detection method and system based on surface-enhanced Raman spectroscopy and visual recognition. The method includes: acquiring multiple spectral images of a food sample to be tested, including visible light spectral images; using a trained transformer neural network to identify the size, location, and freshness of defects in the food sample; based on the defect size and freshness index of the food sample, sequentially determining whether to acquire surface-enhanced Raman spectral images and confocal microscopy images of the food sample; cross-validating the recognition results of the transformer neural network based on the visible light spectral images, surface-enhanced Raman spectral images, and confocal microscopy images of the food sample, and outputting the detection results. This invention combines spectral image transformer neural network for preliminary detection of food samples, and then improves the universality and reliability of the detection results through surface-enhanced Raman spectroscopy and confocal microscopy images.

Description

Technical Field

[0001] This invention belongs to the field of spectroscopy and food detection technology, specifically relating to a food detection method and system based on surface-enhanced Raman spectroscopy and visual recognition. Background Technology

[0002] Raman spectroscopy is a type of spectroscopy that characterizes the vibrational energy levels of molecules. It possesses extremely high molecular specificity, but its scattering intensity is relatively weak. The scattering cross section of a Raman spectrum is 10⁻⁶ units per molecule. -30 cm 2 The effective color scattering of fluorescence is 10. -17 ~10 -16 cm 2 Therefore, it is susceptible to interference from fluorescence. Fleischmann et al. obtained Raman scattering signals on rough silver electrode surfaces, and VanDuyna et al. proposed the surface-enhanced Raman spectroscopy (SERS) effect based on Fleischmann's work, that is, the surface effect is related to the roughness of the metal surface. This is a milestone in the development of Raman spectroscopy.

[0003] SERS technology is fast, sensitive, and non-destructive, possessing molecular fingerprint specificity and single-molecule sensitivity. It can provide information about the structure of matter at the molecular level and has gradually become a powerful detection method in chemistry, biology, environment, and food. When target molecules are adsorbed onto certain rough metal surfaces, their Raman scattering intensity is enhanced by 10% compared to conventional methods. 4 ~10 14 The portable Raman spectrometer and corresponding nano-reinforced substrate have enabled rapid detection of various substances in food, and semi-quantitative analysis can be performed. Currently, the detection results are good for food additives such as preservatives, colorings, and cyclamate; and chemical hazards in food such as illegal additives, pesticide residues, and veterinary drug residues. However, different SERS require different specific substrates, leading to variations in SERS performance. Summary of the Invention

[0004] To improve the versatility and reliability of surface-enhanced Raman spectroscopy in food detection, a first aspect of this invention provides a food detection method and system combining surface-enhanced Raman spectroscopy and visual recognition, comprising: acquiring multiple spectral images of a food sample to be detected, including visible light spectral images; using a trained transformer neural network to identify the defect size, defect location, and freshness index of the food sample to be detected; based on the defect size and freshness index of the food sample to be detected, sequentially determining whether to acquire surface-enhanced Raman spectral images and confocal microscope images of the food sample to be detected; cross-validating the recognition results of the transformer neural network based on the visible light spectral images, surface-enhanced Raman spectral images, and confocal microscope images of the food sample to be detected, and outputting the detection results.

[0005] In some embodiments of the present invention, the transformer neural network includes a first transformer neural network and a second transformer neural network. The first transformer neural network is used to identify the size and location of defects in the food sample to be tested; the second transformer neural network is used to identify the freshness index of the food sample to be tested.

[0006] Furthermore, the first transformer neural network is a YOLOv5-based transformer neural network.

[0007] In some embodiments of the present invention, the step of determining whether to acquire a surface-enhanced Raman spectroscopy image and a confocal microscope image of the food sample to be tested based on the defect size and freshness index of the food sample to be tested includes: if the maximum diameter of the defect of the food sample to be tested is less than or equal to a first threshold and the freshness index is lower than a second threshold, then acquire a surface-enhanced Raman spectroscopy image of the food sample to be tested; if the maximum diameter of the defect of the food sample to be tested is less than or equal to a third threshold and the freshness index is lower than a fourth threshold, then acquire a confocal microscope image of the food sample to be tested.

[0008] Furthermore, the acquisition of the surface-enhanced Raman spectral image of the food sample to be tested includes: identifying the marker substance or microorganism of the food sample to be tested, and determining one or more substrate materials required for the surface-enhanced Raman spectroscopy based on it; constructing one or more specific probes based on the one or more substrate materials, and using the one or more specific probes to acquire the surface-enhanced Raman spectral image of the food sample to be tested.

[0009] In the above embodiments, the cross-validation of the recognition results of the transformer neural network based on the visible light spectrum image, surface-enhanced Raman spectrum image, and confocal microscope image of the food sample to be tested, and the output of the detection results, includes: verifying the freshness index of the food sample to be tested by quantitatively calculating the characteristic peaks of the surface-enhanced Raman spectrum image of the microorganisms in the food sample to be tested; and verifying the defect size of the food sample to be tested by the confocal microscope image of the microorganisms in the food sample to be tested.

[0010] A second aspect of the present invention provides a food detection system based on surface-enhanced Raman spectroscopy and visual recognition, comprising: an acquisition module for acquiring multiple spectral images of a food sample to be detected, the spectral images including visible light spectral images; an identification module for identifying the defect size, defect location, and freshness of the food sample to be detected using a trained transformer neural network; a judgment module for sequentially determining whether to acquire surface-enhanced Raman spectral images and confocal microscope images of the food sample to be detected based on the defect size and freshness index; and an output module for cross-validating the identification results of the transformer neural network based on the visible light spectral image, surface-enhanced Raman spectral image, and confocal microscope image of the food sample to be detected, and outputting the detection results.

[0011] A third aspect of the present invention provides an electronic device comprising: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the surface-enhanced Raman spectroscopy and visual recognition food detection method provided in the first aspect of the present invention.

[0012] In a fourth aspect, the present invention provides a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the food detection method of surface-enhanced Raman spectroscopy and visual recognition provided in the first aspect of the present invention.

[0013] The beneficial effects of this invention are:

[0014] This invention combines a spectral image transformer neural network to perform preliminary detection on food samples, determining the size, location, and freshness of defects. Then, surface-enhanced Raman spectroscopy and confocal microscopy images are used to further verify the size, location, and freshness of defects, thereby improving the versatility and reliability of the detection results. Attached Figure Description

[0015] Figure 1This is a schematic diagram of the basic process of a food detection method combining surface-enhanced Raman spectroscopy and visual recognition in some embodiments of the present invention;

[0016] Figure 2 This is a schematic diagram of the basic structure of the transformer neural network in some embodiments of the present invention;

[0017] Figure 3 This is a schematic diagram illustrating the basic structure of the transformer neural network in some embodiments of the present invention;

[0018] Figure 4 This is a schematic diagram of the structure of the convolutional attention module in the transformer neural network in some embodiments of the present invention;

[0019] Figure 5 This is a schematic diagram of the structure of the encoding module in the transformer neural network in some embodiments of the present invention;

[0020] Figure 6 These are confocal microscope images of a fruit in some embodiments of the present invention;

[0021] Figure 7 This is a schematic diagram of the structure of a food detection system combining surface-enhanced Raman spectroscopy and visual recognition in some embodiments of the present invention;

[0022] Figure 8 This is a schematic diagram of the structure of an electronic device in some embodiments of the present invention. Detailed Implementation

[0023] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0024] refer to Figure 1 and Figure 2 In a first aspect, the present invention provides a food detection method and system based on surface-enhanced Raman spectroscopy and visual recognition, comprising: S100. acquiring multiple spectral images of a food sample to be detected, the spectral images including visible light spectral images; S200. using a trained transformer neural network to identify the defect size, defect location, and freshness of the food sample to be detected; S300. based on the defect size and freshness index of the food sample to be detected, sequentially determining whether to acquire surface-enhanced Raman spectral images and confocal microscope images of the food sample to be detected; S400. cross-validating the recognition results of the transformer neural network based on the visible light spectral image, surface-enhanced Raman spectral image, and confocal microscope image of the food sample to be detected, and outputting the detection results.

[0025] It is understandable that chemical reactions such as Maillard reactions, enzymatic browning, and non-enzymatic browning may occur during food processing and storage, leading to changes in food color. Therefore, visual recognition can be used to learn from standard samples and establish color information and a food freshness index. By measuring parameters such as reflectance, transmittance, color difference, lightness, whiteness, and opacity, food color can be characterized, compensating for the limitations of colorimeters when measuring objects with uneven textures or small volumes. In particular, the color change pattern of each food from fresh to spoiled varies; therefore, it is necessary to establish a correlation between color information and the food freshness index for each type of food. For example, pork mainly exhibits a red color change, while citrus fruits exhibit a yellow color change.

[0026] In step S100 of some embodiments of the present invention, multiple spectral images of the food sample to be detected are acquired, including visible light spectral images. Specifically, the target objects for spectral image detection include food additives, minimally processed foods, highly processed foods, and agricultural, livestock, and fishery products.

[0027] 1) Food additives

[0028] Examples of samples: preservatives, sweeteners, thickeners, color protectants, flavorings and fragrances, antioxidants, emulsifiers, freshness preservers, leavening agents, acidulants, bleaching agents, nutrient fortifiers, moisture retainers, quality improvers, enzyme preparations, stabilizers, defoamers, anti-caking agents, desiccants, coating agents, flavor enhancers, fermentation agents, coagulants, fillers, food colorings, and other additives. States: liquid, powder;

[0029] 2) Simple processing type

[0030] Sample examples: Rice and flour products, vegetable products, meat, sugars, starches, edible oils, processed aquatic products, processed fruits, fillings, tea products, frozen foods, and other minimally processed foods. State: Small granules, powder, irregular shapes, liquid;

[0031] 3) Deep processing

[0032] Sample examples: meat products, dairy products, soy products, egg products, condiments, beverages / drinks, alcoholic beverages, convenience foods, snack foods, health foods, canned foods, extracts, cigarettes, pet food, other foods, etc.

[0033] 4) Agriculture, animal husbandry and fishery

[0034] Examples of samples include: grains, fresh vegetables, fresh fruits, dried fruits / nuts, live aquatic products, live livestock and poultry, eggs, coffee beans / cocoa, edible fungi, and other agricultural products. In addition to the above categories, standard food samples provided by national standards or other national standards can be used as a reference to calculate the freshness index.

[0035]

[0036] Measuring or acquiring visible spectrum images involves the following steps: Due to the properties of powdered products, pretreatment is necessary before measurement. Samples must be placed in transparent cuvettes or sample cups for measurement. Due to their properties, samples are affected by ambient light, measurement path length, and background. It is recommended that the sample thickness be uniform (50mm), and an opaque light shield be used to reduce these effects. Since the sample is non-uniform, multiple measurements must be taken and the average value calculated. Recommended instruments: Integrating sphere D / 8 type such as Hunterlab ETD0001, ETD0002, ETD0003; directional type such as ET04501, ET45001 (ring disk assembly recommended), etc.; Measurement modes: Reflectance color scale: CIELab (D65 / 10)WI, YI index, etc.

[0037] For opaque, irregularly shaped foods such as various dried fruits, beans, dried fruit, potato chips, biscuits, pet food, noodles, candy, bread, mooncakes, large biscuits, cheese, and root vegetables, due to the large particle size, pretreatment is necessary before measurement. Samples must be placed in a transparent sample cup for measurement. Because the samples are non-uniform with large particles, the gaps between samples in the sample cup are significant, and the gaps between samples vary with each placement, greatly affecting the measurement results. To eliminate these effects, large-area measurement should be used, and multiple measurements should be taken and averaged. Recommended instruments: directional ET4500UD, ET04501, ET45001; non-contact STHT; measurement mode: reflection; recommended color scale: CIELab (D65 / 10) index, etc.

[0038] refer to Figures 3 to 5 In step S200 of some embodiments of the present invention, the transformer neural network includes a first transformer neural network and a second transformer neural network. The first transformer neural network is used to identify the size and location of defects in the food sample to be tested; the second transformer neural network is used to identify the freshness index of the food sample to be tested.

[0039] Figure 3The basic structure of the transformer neural network in this disclosure is shown. The network mainly includes a backbone, a neck, and TPH (transformer prediction heads). The backbone includes focus, convolution, C3 (3-fold convolution and CS PBottleneck), SPP (spatial pyramid pooling), and the Trans operation.

[0040] Furthermore, the first transformer neural network is a YOLOv5-based transformer neural network. For details, please refer to [link / reference needed]. Figures 3 to 5 The aforementioned Transformer neural network uses a Transformer Encoder to replace some convolutional and CSP structures. Applying Transformers in vision is currently a mainstream trend, and Transformers possess a unique attention mechanism. CBAM (Convolutional Block Attention Module) is an example... Figure 4 As shown, before a feature map is input into the next processing unit, its channel attention and spatial attention are calculated in parallel, and then fused and reshaped. This makes subsequent processing units more focused on valuable target regions (Regions of Interest, ROIs).

[0041] refer to Figure 5 Each Transformer Encoder processes the initial EmbeddedPatches (sample features or multi-dimensional vectors) through LayerNorm (regularization), Multi-HeadAttention, Dropout, LayerNorm (regularization), MLP (Multi-Layer Perception), and Dropout operations, and finally outputs a feature vector.

[0042] It is understandable that surface-enhanced Raman spectroscopy images require markers or microorganisms on the food sample to determine the degree of spoilage (freshness index) of the food sample, while the freshness index can also be reflected from the visible light spectrum; therefore, the two form the basis for cross-validation, thereby improving universality.

[0043] In step S300 of some embodiments of the present invention, the step of sequentially determining whether to acquire a surface-enhanced Raman spectroscopy image and a confocal microscope image of the food sample to be tested based on the defect size and freshness index of the food sample to be tested includes: S301. If the maximum diameter of the defect in the food sample to be tested is less than or equal to a first threshold and the freshness index is lower than a second threshold, then acquire a surface-enhanced Raman spectroscopy image of the food sample to be tested; S302. If the maximum diameter of the defect in the food sample to be tested is less than or equal to a third threshold and the freshness index is lower than a fourth threshold, then acquire a confocal microscope image of the food sample to be tested. Further, the first threshold, second threshold, third threshold, and fourth threshold need to be specifically determined according to different foods to be tested.

[0044] Furthermore, the acquisition of the surface-enhanced Raman spectral image of the food sample to be tested includes: identifying the marker substance or microorganism of the food sample to be tested, and determining one or more substrate materials required for the surface-enhanced Raman spectroscopy based on it; constructing one or more specific probes based on the one or more substrate materials, and using the one or more specific probes to acquire the surface-enhanced Raman spectral image of the food sample to be tested.

[0045] It is understood that the substrate materials include, but are not limited to, gold and silver nanoparticles, transition metal materials such as platinum (Pt), ruthenium (Ru), rhodium (Rh), palladium (Pd), Fe, cobalt (Co), and nickel (Ni), or other metallic materials, non-metallic nanomaterials, and composite materials that have been proven to have high-quality SERS enhancement phenomena.

[0046] refer to Figure 6 The image shows a confocal microscope image of a food product at 1000x magnification, revealing a certain degree of mold growth (fungus) in the food. In step S400 of the above embodiment, the cross-validation of the transformer neural network's recognition results based on the visible light spectrum image, surface-enhanced Raman spectrum image, and confocal microscope image of the food sample to be tested, and the output of the detection results, includes: verifying the freshness index of the food sample by quantitatively calculating the characteristic peaks of the surface-enhanced Raman spectrum image of the microorganisms in the food sample; and verifying the defect size of the food sample by using the confocal microscope image of the microorganisms in the food sample.

[0047] Example 2

[0048] refer to Figure 7In a second aspect, the present invention provides a food detection system based on surface-enhanced Raman spectroscopy and visual recognition, comprising: an acquisition module 1 for acquiring multiple spectral images of a food sample to be detected, the spectral images including visible light spectral images; an identification module 2 for identifying the defect size, defect location, and freshness of the food sample to be detected using a trained transformer neural network; a judgment module 3 for determining, based on the defect size and freshness index of the food sample to be detected, whether to acquire surface-enhanced Raman spectral images and confocal microscope images of the food sample to be detected; and an output module 4 for cross-validating the recognition results of the transformer neural network based on the visible light spectral images, surface-enhanced Raman spectral images, and confocal microscope images of the food sample to be detected, and outputting the detection results.

[0049] Furthermore, the output module 4 includes: a first verification unit for verifying the freshness index of the food sample to be tested by quantitatively calculating the characteristic peaks of the surface-enhanced Raman spectroscopy image of the microorganisms in the food sample to be tested; and a second verification unit for verifying the defect size of the food sample to be tested by using the confocal microscopy image of the microorganisms in the food sample to be tested.

[0050] Example 3

[0051] refer to Figure 8 A third aspect of the present invention provides an electronic device comprising: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method of the first aspect of the present invention.

[0052] Electronic device 500 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 501, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 502 or a program loaded from storage device 508 into random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of electronic device 500. The processing unit 501, ROM 502, and RAM 503 are interconnected via bus 504. An input / output (I / O) interface 505 is also connected to bus 504.

[0053] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 508 including, for example, hard disks; and communication devices 509. Communication device 509 allows electronic device 500 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 8 An electronic device 500 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 8 Each box shown can represent a device or multiple devices as needed.

[0054] Specifically, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a ROM 502. When the computer program is executed by a processing device 501, it performs the functions defined in the methods of embodiments of this disclosure. It should be noted that the computer-readable medium described in embodiments of this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0055] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more computer programs, which, when executed by the electronic device, cause the electronic device to:

[0056] Computer program code for performing the operations of embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, C++, and Python—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0057] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0058] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A food detection method using surface-enhanced Raman spectroscopy and visual recognition, characterized in that, include: Acquire multiple spectral images of the food sample to be tested, including visible light spectral images; Using a trained transformer neural network, the size, location, and freshness index of defects in the food samples to be tested are identified. Based on the defect size and freshness index of the food sample to be tested, it is determined in sequence whether to acquire surface-enhanced Raman spectroscopy and confocal microscopy images of the food sample to be tested: if the maximum diameter of the defect in the food sample to be tested is less than or equal to the first threshold and the freshness index is lower than the second threshold, then acquire the surface-enhanced Raman spectroscopy image of the food sample to be tested; if the maximum diameter of the defect in the food sample to be tested is less than or equal to the third threshold and the freshness index is lower than the fourth threshold, then acquire the confocal microscopy image of the food sample to be tested. Based on the visible light spectrum image, surface-enhanced Raman spectrum image, and confocal microscope image of the food sample to be tested, the recognition results of the transformer neural network are cross-validated, and the detection results are output.

2. The food detection method based on surface-enhanced Raman spectroscopy and visual recognition according to claim 1, characterized in that, The transformer neural network includes a first transformer neural network and a second transformer neural network. The first transformer neural network is used to identify the size and location of defects in the food sample to be tested; The second transformer neural network is used to identify the freshness index of the food sample to be tested.

3. The food detection method based on surface-enhanced Raman spectroscopy and visual recognition according to claim 2, characterized in that, The first transformer neural network is a YOLOv5-based transformer neural network.

4. The food detection method based on surface-enhanced Raman spectroscopy and visual recognition according to claim 1, characterized in that, The acquisition of the surface-enhanced Raman spectral image of the food sample to be tested includes: Identify the marker substances or microorganisms in the food sample to be tested, and determine one or more substrate materials required for surface-enhanced Raman spectroscopy based on them; One or more specific probes are constructed based on the one or more substrate materials, and surface-enhanced Raman spectral images of the food sample to be tested are obtained using the one or more specific probes.

5. The food detection method based on surface-enhanced Raman spectroscopy and visual recognition according to any one of claims 1 to 4, characterized in that, The recognition results of the transformer neural network are cross-validated based on the visible light spectrum image, surface-enhanced Raman spectrum image, and confocal microscope image of the food sample to be tested, and the output detection results include: The freshness index of the food sample was verified by quantitatively calculating the characteristic peaks of the surface-enhanced Raman spectroscopy images of microorganisms in the food sample to be tested. The size of defects in the food sample was verified by using confocal microscopy images of the microorganisms in the food sample to be tested.

6. A food detection system combining surface-enhanced Raman spectroscopy and visual recognition, characterized in that, include: The acquisition module is used to acquire multiple spectral images of the food sample to be tested, including visible light spectral images; The identification module is used to identify the size, location, and freshness index of defects in the food sample to be tested using a trained transformer neural network. The judgment module is used to determine whether to acquire a surface-enhanced Raman spectroscopy image and a confocal microscope image of the food sample to be tested based on the size of the defects and the freshness index: if the maximum diameter of the defects in the food sample to be tested is less than or equal to the first threshold and the freshness index is lower than the second threshold, then the surface-enhanced Raman spectroscopy image of the food sample to be tested is acquired; if the maximum diameter of the defects in the food sample to be tested is less than or equal to the third threshold and the freshness index is lower than the fourth threshold, then the confocal microscope image of the food sample to be tested is acquired. The output module is used to cross-validate the recognition results of the transformer neural network based on the visible light spectrum image, surface-enhanced Raman spectrum image and confocal microscope image of the food sample to be tested, and output the detection results.

7. The food detection system based on surface-enhanced Raman spectroscopy and visual recognition according to claim 6, characterized in that, The output module includes: The first verification unit is used to verify the freshness index of the food sample by quantitatively calculating the characteristic peaks of the surface-enhanced Raman spectrum image of the microorganisms in the food sample to be tested. The second verification unit is used to verify the size of defects in the food sample by using confocal microscopy images of the microorganisms in the food sample.

8. An electronic device, comprising: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the food detection method of surface-enhanced Raman spectroscopy and visual recognition as described in any one of claims 1 to 5.

9. A computer-readable medium having a computer program stored thereon, wherein, When the computer program is executed by the processor, it implements the food detection method of surface-enhanced Raman spectroscopy and visual recognition as described in any one of claims 1 to 5.

Images