Food quality safety supervision system

Abstract

The invention relates to the field of food safety supervision, in particular to a food quality safety supervision system, which comprises a PDA (Personal Digital Assistant) terminal, comprising an integrated information collector for acquiring a food electronic certificate, an electronic nose for acquiring food smell data, a spectral imager for acquiring a food spectral image and a visual sensor for acquiring a food surface image, the supervision cloud platform is used for matching each big data evaluation model with the corresponding food type through the electronic certificate; and for a single type of food, the big data evaluation model judges the quality grade of the food based on the surface image of the food and re-determines the quality grade of the food based on the smell data and the spectral image, and if the quality grade is unqualified, a food processing suggestion is generated based on the quality grade. Rapid detection is carried out through the PDA terminal, and the quality grade of the food is generated through the supervision cloud platform, so that portable and rapid detection of the food quality and efficient supervision of multi-link detection are realized.

Description

Technical Field

[0001] The present invention relates to the field of food safety supervision, and in particular to a food quality and safety supervision system. Background Art

[0002] As people's living standards improve, they pay more and more attention to food safety. However, food is at risk of contamination in many links, including production, storage, transportation and sales, which requires monitoring by multiple departments and links. Although traditional laboratory testing is highly accurate, it has high construction costs and long testing cycles, making it difficult to meet the needs of rapid and large-scale testing.

[0003] Therefore, portable and rapid detection equipment integrating multiple sensors plays an increasingly important role in food safety testing.

[0004] For example, the invention document with application number 202010997539.4 discloses an intelligent food receiving system for a group canteen, which is only interoperable with the supply chain of the group canteen.

[0005] Therefore, how to make the rapid detection terminal both interconnected with the supply chain and facilitate on-site rapid testing of food is a technical problem that needs to be solved. Summary of the invention

[0006] To this end, the present invention provides a food quality and safety supervision system, which performs rapid detection through a PDA terminal and generates food quality grades through a supervision cloud platform, thereby realizing portable and rapid detection of food quality, as well as remote and efficient supervision of multi-link detection.

[0007] To achieve the above object, the present invention proposes a food quality and safety supervision system, comprising:

[0008] A PDA terminal for on-site rapid inspection of food, including an integrated information collector for obtaining electronic food vouchers, an electronic nose for obtaining food odor data, a spectral imager for obtaining food spectral images, and a visual sensor for obtaining food surface images;

[0009] A supervision cloud platform is connected to the PDA terminal and includes several big data evaluation models. Each big data evaluation model is matched with a corresponding food category through the electronic certificate. For a single category of food, the big data evaluation model determines the quality grade of the food based on the surface image of the food, and redetermines the quality grade of the food based on the odor data and the spectral image. If the quality grade is unqualified, a food processing suggestion is generated based on the quality grade.

[0010] The big data evaluation model of the regulatory cloud platform determines whether food is adulterated based on the characteristic values ​​of the odor data.

[0011] Furthermore, the big data evaluation model includes a neural network sub-model that can characterize the nonlinear relationship between the characteristic wavelength of pesticides and pesticide residues, which is used to extract the characteristic wavelength of pesticides from the full band of the spectral image, and determine the pesticide residue based on the characteristic wavelength of pesticides. If the pesticide residue exceeds the pesticide residue safety value, it is determined that pesticide residues exist in the food, and pesticide type information is generated.

[0012] Furthermore, the big data evaluation model also includes a correction sub-model, which is used to determine whether the pesticide residue is an abnormal detection pesticide based on the ratio of the area corresponding to the characteristic wavelength of the pesticide to the total area of ​​the spectral image. If so, the characteristic wavelength of the pesticide is not input into the neural network sub-model; otherwise, the characteristic wavelength of the pesticide is input into the neural network sub-model.

[0013] Furthermore, the food includes fruit, and the big data evaluation model also includes a soluble solids extraction sub-model, which is used to extract content characteristic wavelengths associated with soluble solids content from the full band of the spectral image, and determine the soluble solids content based on the content characteristic wavelengths, and judge whether the soluble solids content exceeds the transportation safety value, and if it does not exceed the transportation safety value, adjust the quality grade to unqualified.

[0014] In the above scheme, the spectral imaging device of the PDA terminal realizes dual monitoring of pesticide residues and soluble solids content. Compared with multiple detections using the spectral imaging device, it greatly simplifies the detection process and further realizes rapid detection.

[0015] Furthermore, the big data evaluation model also includes:

[0016] An image feature extraction mechanism for extracting defect features of the surface image through a neural network;

[0017] An attention mechanism is used to generate correlation features between multiple defect features;

[0018] A three-dimensional convolution mechanism is used to combine and distinguish the associated features to generate a food surface image with defect type information and defect degree values ​​marked;

[0019] The quality grade determination mechanism is used to determine the quality grade according to the defect type information and the defect degree value.

[0020] Furthermore, the quality grade determination mechanism is equipped with a grade generation algorithm, which determines setting parameters of the grade generation algorithm according to the defect type information, and inputs the defect degree value into the grade generation algorithm to obtain the quality grade.

[0021] Furthermore, the supervision cloud platform also includes a boundary extraction module, which is used to perform optical correction and background deletion on the spectral image to determine the food boundary of the surface image.

[0022] In the above scheme, accurate judgment of the quality grade of food is achieved through visual sensors.

[0023] Furthermore, the supervision cloud platform also includes an early warning module, which determines the link where the quality level is reduced based on the comparison results of multiple quality levels of food in multiple links of the supervision cloud platform, and generates early warning information.

[0024] Furthermore, the controller of the PDA terminal communicates with the supervision cloud platform based on the LPWAN Internet of Things. The supervision cloud platform also includes a supervision report generation module, which is used to generate a topological supervision report based on the electronic credentials, odor data, spectral image, surface image and quality grade of food in multiple links, and send and display the inspection report to the user terminal.

[0025] Furthermore, the multiple links include food production, storage, transportation and sales.

[0026] In the above solution, the supervision cloud platform is used to realize the interconnection and interoperability of supervision information in the production, storage, transportation and sales links of food, realize intelligent judgment of potential risk points, and provide decision-making support for regulatory authorities.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] 1. Rapid testing is carried out through PDA terminals, and the quality grade of food is generated through the supervision cloud platform, which realizes the portable and rapid testing of food quality, as well as the remote and efficient supervision of multi-link testing.

[0029] 2. The spectral imaging device of the PDA terminal realizes dual monitoring of pesticide residues and soluble solids content. Compared with multiple detections using the spectral imaging device, it greatly simplifies the detection process and further realizes rapid detection.

[0030] 3. Accurate judgment of food quality level is achieved through visual sensors.

[0031] 4. Through the supervision cloud platform, the interconnection and interoperability of supervision information in the production, storage, transportation and sales links of food is realized, and intelligent judgment of potential risk points is realized, providing decision-making support for regulatory authorities. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1A schematic diagram of the general structure of a food quality and safety supervision system according to an embodiment of the present invention;

[0033] Figure 2 A detailed flowchart of a food quality and safety supervision system according to an embodiment of the present invention;

[0034] Figure 3 A schematic diagram of a spectral image of pesticide residues in a food quality and safety monitoring system according to an embodiment of the present invention;

[0035] Figure 4 It is a detailed structural diagram of the food quality and safety supervision system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0036] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0037] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0038] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0039] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0040] like Figures 1 to 4 As shown, the present invention provides a food quality and safety supervision system, which performs rapid detection through a PDA terminal and generates food quality grades through a supervision cloud platform, thereby realizing portable and rapid detection of food quality and remote and efficient supervision of multi-link detection.

[0041] in, Figure 1 A schematic diagram of the general structure of a food quality and safety supervision system according to an embodiment of the present invention; Figure 2 A detailed flowchart of a food quality and safety supervision system according to an embodiment of the present invention; Figure 3 A schematic diagram of a spectral image of pesticide residues in a food quality and safety monitoring system according to an embodiment of the present invention; Figure 4 It is a detailed structural diagram of the food quality and safety supervision system according to an embodiment of the present invention.

[0042] Embodiment 1:

[0043] like Figures 1 to 4 As shown, this embodiment proposes a food quality and safety supervision system, including: a PDA terminal, which is used to perform on-site rapid inspection of food, including an integrated information collector for obtaining food electronic certificates, an electronic nose for obtaining food odor data, a spectral imager for obtaining food spectral images, and a visual sensor for obtaining food surface images;

[0044] A supervision cloud platform is connected to the PDA terminal and includes several big data evaluation models. Each big data evaluation model is matched with a corresponding food category through the electronic certificate. For a single category of food, the big data evaluation model determines the quality grade of the food based on the surface image of the food, and redetermines the quality grade of the food based on the odor data and the spectral image. If the quality grade is unqualified, a food processing suggestion is generated based on the quality grade.

[0045] It should be noted that see Figure 4 In some implementations, the information collector, electronic nose, spectral imager and visual sensor reuse functions such as battery power supply and external communication. In another implementation, the information collector, electronic nose, spectral imager and visual sensor only communicate with a single-chip microcomputer to achieve unified communication and distributed control.

[0046] It can be understood that by classifying the odor data detected by the electronic nose based on multi-dimensional odor characteristics, it is possible to detect whether the ingredients of food are single and whether the food is adulterated.

[0047] It is understandable that the food quality and safety supervision system of this embodiment can be applied to new technologies such as the Internet of Things and blockchain, thereby enabling food safety traceability to move from post-search to real-time monitoring. By embedding smart sensors in PDA terminals, the automatic collection and real-time upload of key data in the food supply chain can be achieved; blockchain technology is used to ensure the authenticity, immutability and traceability of traceability information. This ensures that every flow of food in the supply chain is under the supervision of the regulator. Once an abnormality is found, it can be quickly identified, accurately traced, and handled promptly.

[0048] Specifically, the detection principle of the electronic nose is to record the response signal of the gas sensor array to the sample. The response signal is processed on the computer to give a characteristic value, and the odor characteristic value is input into the pattern recognition library to give the result of type judgment, thereby achieving the purpose of sample type judgment. By judging the type of sample and the type of food carried by the electronic voucher, it is determined whether it is adulterated.

[0049] Furthermore, the big data evaluation model includes a neural network sub-model that can characterize the nonlinear relationship between the characteristic wavelength of pesticides and pesticide residues, which is used to extract the characteristic wavelength of pesticides from the full band of the spectral image, and determine the pesticide residue based on the characteristic wavelength of pesticides. If the pesticide residue exceeds the pesticide residue safety value, it is determined that pesticide residues exist in the food, and pesticide type information is generated.

[0050] Furthermore, the big data evaluation model also includes a correction sub-model, which is used to determine whether the pesticide residue is an abnormal detection pesticide based on the ratio of the area corresponding to the characteristic wavelength of the pesticide to the total area of ​​the spectral image. If so, the characteristic wavelength of the pesticide is not input into the neural network sub-model; otherwise, the characteristic wavelength of the pesticide is input into the neural network sub-model.

[0051] Furthermore, the food includes fruit, and the big data evaluation model also includes a soluble solids extraction sub-model, which is used to extract content characteristic wavelengths associated with soluble solids content from the full band of the spectral image, and determine the soluble solids content based on the content characteristic wavelengths, and judge whether the soluble solids content exceeds the transportation safety value, and if it does not exceed the transportation safety value, adjust the quality grade to unqualified.

[0052] It can be understood that the soluble solid content refers to the percentage of water-soluble sugars, acids, vitamins, minerals, etc. contained in food to the total mass or weight of the food. It is an important indicator reflecting the maturity of the fruit. The higher the maturity, the higher the soluble solid content. Studies have shown that vibration stress and temperature changes during fruit transportation can promote the increase of soluble solid content in the fruit, accelerate the ripening and aging of the fruit, and reduce its storage resistance. Therefore, it is necessary to ensure that the soluble solid content is not so high that the fruit cannot meet the transportation conditions, so as to avoid the fruit being found to be deteriorating after transportation and difficult to sell.

[0053] See also Figure 3 , which represents the pseudo-color images of apple samples with different contamination levels, Figure 3The part selected by the dotted line is the part determined to have pesticide residues. The pseudo-color image is obtained by extracting the corresponding characteristic wavelength / content characteristic wavelength of the pesticide from the full band of the spectral imager, that is, the characteristic wavelength of the pesticide is set to the characteristic wavelength range of the pesticide, that is, 340nm to 690nm, and the types of pesticides such as cypermethrin, imidacloprid, benzyl·dichlor, soybean seed dressing agent and organophosphorus pesticides can be detected. According to whether the characteristic wavelength of the pesticide overlaps with the characteristic wavelength of cypermethrin, imidacloprid, benzyl·dichlor, soybean seed dressing agent and organophosphorus pesticides, the pesticide type information is generated, and the PDA terminal sends the pesticide type information to the supervision cloud platform for display, thereby realizing the supervision of the types of pesticide residues in food. The spectral features of the full wavelength of the set area within the characteristic wavelength range of the pesticide are extracted. A neural network sub-model is established to characterize the nonlinear relationship between the spectral characteristics of pesticides and pesticide residues. The spectral characteristics of pesticides are input into the neural network sub-model to determine the pesticide residues. The more pesticide residues there are, the color in the pseudo-color image gradually changes from blue to yellow and then to red. In this way, the grayscale image predicted by pseudo-color processing can be used to intuitively judge the degree of internal contamination of the fruit.

[0054] By comparing the pseudo-color images, it is found that the colors of some areas appear vertical lines. It is understandable that the unique uneven texture characteristics of the fruit surface collected by the visual sensor will have a certain impact on the presentation of the pseudo-color image, so the use of a spectral imager can obtain the accurate degree of contamination inside the fruit.

[0055] Specifically, the spectral imager is preferably an ultra-high performance liquid chromatograph with a PDA detector, which is arranged in an Empower2 chromatography workstation.

[0056] It is understandable that since the spectral wavelengths of some pigments and biological molecules (such as proteins, nucleic acids, etc.) overlap with the characteristic wavelengths of pesticides (340nm to 690nm), pigments and biological molecules can be eliminated by judging the area ratio to avoid misjudging them as pesticide residues.

[0057] In the above scheme, the spectral imaging device of the PDA terminal realizes dual monitoring of pesticide residues and soluble solids content. Compared with multiple detections using the spectral imaging device, it greatly simplifies the detection process and further realizes rapid detection.

[0058] Furthermore, the big data evaluation model also includes: an image feature extraction mechanism, which is used to extract defect features of the surface image through a neural network; an attention mechanism, which is used to generate correlation features between multiple defect features; a three-dimensional convolution mechanism, which is used to splice and distinguish the correlation features to generate a food surface image marked with defect type information and defect degree values; a quality grade determination mechanism, which is used to determine the quality grade based on the defect type information and the defect degree value.

[0059] Furthermore, the attention mechanism (CAM) includes multiple sets of parallel compression networks (Squeeze) and excitation networks (Excitation); sample data of food surfaces with depressions, corrosion, discoloration, and deformation are used to train the image feature extraction mechanism, the excitation network, and the three-dimensional convolution mechanism. Each set of compression networks and excitation networks processes the depression features, corrosion features, discoloration features, and deformation features determined by the image feature extraction mechanism, respectively, thereby enhancing the correlation features between different features.

[0060] Specifically, the excitation network obtains the weight value of each channel in the corresponding feature map according to the compression process, normalizes the weight value through the Sigmoid function to obtain a weight vector, and multiplies all elements in the weight vector by the corresponding channel of the input feature.

[0061] Furthermore, the 3D Convolution Module (3D-CM) can perform deep splicing and feature discrimination on multiple feature vectors according to the associated features. In the 3D convolution mechanism, after feature splicing of the concave features, corrosion features, discoloration features, and deformation features, the defects on the food surface are comprehensively discriminated according to the associated features, and a comprehensive judgment result integrating the common defect features of concave, corrosion, discoloration, and deformation is generated, and the defect type information and defect degree value are comprehensively judged.

[0062] Furthermore, the feature annotation mechanism of the three-dimensional convolution mechanism includes at least a pooling layer, a fully connected layer and a Sigmoid activation function; the associated features are written into the pooling layer, the fully connected layer and the activation function in sequence to generate a food surface image with defect type information and defect degree values, and the food surface image is sent to the cloud platform server, and then displayed on the user side, so that the user can easily determine the defect type and defect degree of the food, thereby eliminating the process of the user checking the food one by one with the naked eye, so that the user only needs to collect food information through PDA terminal communication, which greatly reduces the user's workload.

[0063] It can be understood that the attention mechanism and the three-dimensional convolution mechanism adopt an early fusion strategy to deeply fuse related visual features at an earlier stage and maintain semantic consistency between various features, which is more conducive to downstream tasks.

[0064] Furthermore, if Figure 4As shown, this embodiment is based on the existing food defect image library and uses the functions in the relevant resource library based on Python to expand the sample size. The functions include adding noise data, changing the original sample pixel size, changing the original sample direction position, changing the original sample integrity, adding Gaussian noise, adding salt and pepper noise, adding Poisson noise, etc. This solves the problem that some of the food image samples obtained are single and it is difficult to provide a large number of data samples for training the convolutional neural network image recognition model through simple collection.

[0065] Furthermore, the quality grade determination mechanism is equipped with a grade generation algorithm, which determines setting parameters of the grade generation algorithm according to the defect type information, and inputs the defect degree value into the grade generation algorithm to obtain the quality grade.

[0066] Specifically, the quality level generation algorithm is: In the formula, Q is the quality grade, S is the defect area, W is the detection window area, the defect degree is expressed by the ratio of the defect area to the detection window area, L is the correction value for different spectral imagers, k is the setting parameter, and k is set according to the defect type information. More specifically, the setting parameters k for the defect type information of dent, corrosion, discoloration, and deformation are 1, 8, 5, and 2, respectively. The value of the unqualified quality grade is greater than 15, the value of the risky quality grade is less than or equal to 15 and greater than 5, and the value of the good quality grade is less than or equal to 5.

[0067] When the quality level is unqualified, a food handling recommendation is generated to recommend discarding the food; when there is a risk in the quality level, a food handling recommendation is generated to recommend proper storage; when the quality level is good, a food handling recommendation is generated to recommend maintaining the current storage method.

[0068] Specifically, the quality grade is the weighted average of the pre-quality grade and the above-mentioned quality grade Q. In the production stage, pesticide residues may be too much and will be removed by themselves in the subsequent transportation stage, so image detection is more preferred to prevent food with obvious defects from circulation. At this time, the weights are 0.3 and 0.7 respectively; in the storage stage, image detection is more preferred, and the weights are 0.4 and 0.6 respectively; in the transportation stage, spectral detection is more preferred, and the weights are 0.6 and 0.4 respectively, to avoid further circulation of food with pesticide residues; in the sales stage, spectral detection and image detection are equally important, and their weights are 0.5 and 0.5 respectively.

[0069] Furthermore, the supervision cloud platform also includes a boundary extraction module, which is used to perform optical correction and background deletion on the spectral image to determine the food boundary of the surface image.

[0070] In the above scheme, accurate judgment of the quality grade of food is achieved through visual sensors.

[0071] Furthermore, the supervision cloud platform also includes an early warning module, which determines the link where the quality level is reduced based on the comparison results of multiple quality levels of food in multiple links of the supervision cloud platform, and generates early warning information.

[0072] Furthermore, the controller of the PDA terminal communicates with the supervision cloud platform based on the LPWAN Internet of Things. The supervision cloud platform also includes a supervision report generation module, which is used to generate a topological supervision report based on the electronic credentials, odor data, spectral image, surface image and quality grade of food in multiple links, and send and display the inspection report to the user terminal.

[0073] Furthermore, the multiple links include food production, storage, transportation and sales.

[0074] It should be noted that the controller is preferably a single chip microcomputer.

[0075] In the above solution, the supervision cloud platform is used to realize the interconnection and interoperability of supervision information in the production, storage, transportation and sales links of food, realize intelligent judgment of potential risk points, and provide decision-making support for regulatory authorities.

[0076] See also Figure 4 In this embodiment, the perception layer is mainly composed of sensors, microcontrollers, relays, etc. The data collected by the sensor is uploaded to the application layer through the transmission layer through wireless sensing technology or wired communication technology. The application layer obtains the quality inspection report after data analysis and processing, and displays the quality inspection report on the user end and PC end, and stores it in the memory. The perception layer is the basis for realizing multi-link monitoring and control of the food quality and safety supervision system.

[0077] The transport layer is composed of LoRa communication nodes, LoRa gateways and NB-IoT communication nodes. LoRa gateways and NB-IoT nodes use star networking to collect data from sensors in multiple links scattered in different places. The NB-IoT communication module is responsible for uploading data to the cloud platform. The transport layer is the communication link between the perception layer and the application layer, and plays an important role in monitoring data upload and control command issuance.

[0078] The application layer is composed of cloud platform, application server, database and user terminal. The server analyzes and processes the data collected by the system and stores it in the database, dynamically displays it to the user in the form of a topology map, and sends the control decision results to the perception layer to remotely monitor and automatically control the detection process, thus realizing remote management of detection.

[0079] It should be noted that LoRa technology is a low-power long-distance communication technology based on LoRa spread spectrum modulation technology first launched by SemTech in the United States. Compared with sub-GHz wireless communication technology, LoRa based on linear Chip spread spectrum modulation (CSS) demodulation technology has a longer effective communication distance, higher sensitivity, stronger anti-interference ability, and a significant improvement in data transmission distance.

[0080] It is understandable that a rapid response and traceability mechanism should be established to control the development of the incident to the greatest extent possible and reduce the losses and impacts of food safety incidents. Food companies should formulate detailed food safety emergency plans, clarify the emergency organizational structure, division of responsibilities, handling procedures and information reporting, etc. Once a food safety incident occurs, an emergency response should be initiated as soon as possible, and measures such as product recall, flow investigation, and pollution source control should be taken quickly to control the spread of the incident.

[0081] It is understandable that the quality level can be matched with food safety management systems such as HACCP and ISO22000, meet the analysis results of key links based on HACCP hygiene requirements, formulate scientific operating procedures and monitoring plans, and achieve real-time monitoring of key parameters affecting food safety. Once deviations are found, corrective measures are taken in a timely manner to ensure the stability of product quality. With Yili Group as the target, a full-chain quality control system from "farm to table" has been established.

[0082] Specifically, see Figure 4 The microcontroller is a low-power single-chip computer powered by solar energy, which has a sleep mode and a communication mode to achieve interconnection between the perception layer, the transmission layer and the application layer without power supply or charging.

[0083] Furthermore, the spectral imager of the PDA terminal can also detect foodborne microorganisms in specific production and sales links, and determine whether chemical pollution, foodborne pathogenic bacteria contamination and veterinary drug residues occur according to the types of foodborne microorganisms.

[0084] Furthermore, the cloud platform at the application layer has a built-in one-click feedback module. When it receives a one-click feedback command from the user end or PC end, it automatically fills in the feedback form required by the regulatory department. After the user confirms, it is sent to the regulatory department with one click, thereby realizing rapid feedback of food problems and preventing the circulation of problematic foods.

[0085] Specifically, regulatory authorities can use the system to view historical test data, analyze risk trends, and formulate targeted regulatory measures. In addition, the cloud platform also uses electronic certificates to conduct full traceability to ensure that food quality is controllable.

[0086] In this embodiment, by integrating rapid detection instruments, detection cloud platforms, data management modules and intelligent analysis modules, rapid detection, real-time supervision and full traceability of food safety are achieved. The system has the advantages of wide detection range, high accuracy and easy operation, which can significantly improve the efficiency and level of food safety supervision and ensure the food safety of the people.

[0087] In this embodiment, rapid detection is performed through a PDA terminal, and the quality grade of food is generated through a regulatory cloud platform, thereby achieving portable and rapid detection of food quality, as well as remote and efficient supervision of multi-link detection. The spectral imager of the PDA terminal realizes dual monitoring of pesticide residues and soluble solids content. Compared with multiple detections using a spectral imager, it greatly simplifies the detection process and further realizes rapid detection. Accurate judgment of the quality grade of food is achieved through visual sensors. Through the regulatory cloud platform, the interconnection of regulatory information in the production, storage, transportation and sales links of food is achieved, and intelligent judgment of potential risk points is realized, providing decision-making support for regulatory authorities.

[0088] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0089] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A food quality and safety supervision system, characterized in that: include: A PDA terminal for on-site rapid inspection of food, including an integrated information collector for obtaining electronic food vouchers, an electronic nose for obtaining food odor data, a spectral imager for obtaining food spectral images, and a visual sensor for obtaining food surface images; A supervision cloud platform is connected to the PDA terminal and includes several big data evaluation models. Each big data evaluation model is matched with a corresponding food category through the electronic certificate. For a single category of food, the big data evaluation model determines the quality grade of the food based on the surface image of the food, and redetermines the quality grade of the food based on the odor data and the spectral image. If the quality grade is unqualified, a food processing suggestion is generated based on the quality grade.

2. The food quality and safety supervision system according to claim 1, characterized in that: The big data evaluation model includes a neural network sub-model that can characterize the nonlinear relationship between the characteristic wavelength of pesticides and the pesticide residue, which is used to extract the characteristic wavelength of pesticides from the full band of the spectral image, and determine the pesticide residue based on the characteristic wavelength of pesticides. If the pesticide residue exceeds the pesticide residue safety value, it is determined that pesticide residues exist in the food, and pesticide type information is generated.

3. The food quality and safety supervision system according to claim 2 is characterized in that: The big data evaluation model also includes a correction sub-model, which is used to determine whether the pesticide residue is an abnormal detection pesticide based on the ratio of the area corresponding to the characteristic wavelength of the pesticide to the total area of ​​the spectral image. If so, the characteristic wavelength of the pesticide is not input into the neural network sub-model; otherwise, the characteristic wavelength of the pesticide is input into the neural network sub-model.

4. The food quality and safety supervision system according to claim 1, characterized in that: The food includes fruit, and the big data evaluation model also includes a soluble solids extraction sub-model, which is used to extract content characteristic wavelengths associated with soluble solids content from the full band of the spectral image, and determine the soluble solids content based on the content characteristic wavelengths, and judge whether the soluble solids content exceeds the transportation safety value, and if it does not exceed the transportation safety value, adjust the quality grade to unqualified.

5. The food quality and safety supervision system according to claim 1, characterized in that: The big data evaluation model also includes: An image feature extraction mechanism for extracting defect features of the surface image through a neural network; An attention mechanism is used to generate correlation features between multiple defect features; A three-dimensional convolution mechanism is used to combine and distinguish the associated features to generate a food surface image with defect type information and defect degree values ​​marked; The quality grade determination mechanism is used to determine the quality grade according to the defect type information and the defect degree value.

6. The food quality and safety supervision system according to claim 5, characterized in that: The quality grade determination mechanism is equipped with a grade generation algorithm, which determines the setting parameters of the grade generation algorithm according to the defect type information, and inputs the defect degree value into the grade generation algorithm to obtain the quality grade.

7. The food quality and safety supervision system according to any one of claims 1 to 6, characterized in that: The supervision cloud platform also includes a boundary extraction module, which is used to perform optical correction and background deletion on the spectral image to determine the food boundary of the surface image.

8. The food quality and safety supervision system according to any one of claims 1 to 6, characterized in that: The supervision cloud platform also includes an early warning module, which determines the link where the quality level is reduced based on the comparison results of multiple quality levels of food in multiple links of the supervision cloud platform and generates early warning information.

9. The food quality and safety supervision system according to claim 8, characterized in that: The controller of the PDA terminal communicates with the supervision cloud platform based on the LPWAN Internet of Things. The supervision cloud platform also includes a supervision report generation module, which is used to generate a topological supervision report based on the electronic credentials, odor data, spectral image, surface image and quality grade of food in multiple links, and send and display the inspection report to the user terminal.

10. The food quality and safety supervision system according to claim 8, characterized in that: The multiple links include food production, storage, transportation and sales.

Images