Cold chain food material identification method and system based on image processing

Abstract

The invention relates to the technical field of cold chain food management, and discloses a cold chain food material identification method and system based on image processing, and the system comprises a data collection module and a data identification module. The identification method comprises the following steps: step 1, a data acquisition module reads a bar code of a cold chain food material through a bar code reader; 2, a data acquisition module acquires an image through a high-definition camera; 3, a data acquisition module is connected with a database through a network to acquire a reference data set; step 4, the data identification module analyzes and calculates the safety score Aqpf of the cold-chain food material; 5, the data identification module generates a residual block Cck and introduces a ResNet algorithm model, and deep learning keeps original features, so that the identification precision is higher; step 6, analyzing and calculating a quality score Zlpf of the cold-chain food materials according to the image data set; 7, the data identification module carries out classification marking, then food material management is carried out, multi-dimensional identification is not prone to wrong judgment or missing judgment, and the processing efficiency is high.

Description

Technical Field

[0001] The present invention relates to the technical field of cold chain food management, and specifically to a cold chain food identification method and system based on image processing. Background Art

[0002] Cold chain ingredients refer to ingredients that are always kept in a low temperature environment during production, storage, transportation, sales and consumption. This low temperature environment can effectively inhibit the growth and reproduction of microorganisms, extend the shelf life of ingredients, and maintain the freshness and nutritional value of ingredients. Cold chain ingredients mainly include frozen foods, refrigerated foods, fresh meat, aquatic products, dairy products, etc. The storage of cold chain ingredients should follow the principle of "four degrees and one light", namely temperature, humidity, air quality, light and packaging, to ensure the freshness and nutritional value of ingredients. The management and maintenance of storage facilities are also the key to ensuring the quality of cold chain ingredients, and regular inspections and maintenance are required to ensure their normal operation. Generally speaking, the storage temperature of frozen foods should be below -18°C, the storage temperature of refrigerated foods should be 0-4°C, the storage temperature of fresh meat and aquatic products should be 0-4°C, and the storage temperature of dairy products should be 2-6°C. The temperature control should be accurate to ±1°C to prevent the deterioration of ingredients and loss of nutrients. The storage humidity of cold chain ingredients should also be set according to their types and characteristics. The storage humidity of frozen food should be above 90%, that of refrigerated food should be between 85-95%, that of fresh meat and aquatic products should be between 85-95%, and that of dairy products should be between 85-90%. The control of humidity should be accurate to ±5% to prevent the drying and deterioration of food ingredients. The storage environment of cold chain food ingredients should be kept clean and free from pollution by harmful gases and odors. Storage facilities should be cleaned and disinfected regularly to maintain air quality. The storage environment of cold chain food ingredients should avoid direct strong light to prevent discoloration and deterioration of food ingredients. Storage facilities should use shading materials or equipment to reduce the impact of light. The packaging of cold chain food ingredients should have good sealing and heat preservation properties to prevent food contamination and temperature changes. Packaging materials should be non-toxic, odorless, and not easily damaged to ensure the safety and quality of food ingredients. The storage method of cold chain food ingredients should be selected according to their types and characteristics. Generally speaking, frozen food should be stored vertically, refrigerated food should be stored horizontally, fresh meat and aquatic products should be stored in a hanging manner, and dairy products should be stored in a stacked manner. The choice of storage method should take into account the convenience of accessing ingredients and the efficient use of space.

[0003] At present, traditional cold chain food identification methods based on image processing mostly rely on deep learning algorithm models combined with manual recognition methods to judge whether the quality of the food is qualified based on the food appearance image. However, in actual use, the deep learning algorithm model is prone to problems such as gradient vanishing and network degradation. The food appearance image is prone to subtle differences that are difficult for the human eye to distinguish, and it is easy to make misjudgments and missed judgments, resulting in low recognition and processing efficiency. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] In view of the shortcomings of the prior art, the present invention provides a cold chain food identification method and system based on image processing, which has the advantages of deep learning retaining the original features with higher recognition accuracy, multi-dimensional recognition not prone to false positives and high processing efficiency. It solves the problem that the algorithm model used in the traditional cold chain food identification method based on image processing is prone to gradient vanishing and network degradation, and manual intervention recognition is prone to false positives and low processing efficiency.

[0006] (II) Technical solution

[0007] To achieve the above object, the present invention provides the following technical solution: a cold chain food identification method based on image processing, comprising the following steps:

[0008] Step 1: The data collection module reads all the cold chain food barcodes through a barcode reader and forms a barcode data set T, which includes all the cold chain food data whose barcode information is successfully read;

[0009] Step 2: The data acquisition module uses a high-definition camera to capture the appearance images of all cold chain ingredients. The high-definition camera takes photos of each surface one by one according to the number of planes on the surface of each cold chain ingredient, and forms an image data set. The image data set includes the appearance images of all cold chain ingredients whose barcode information has been successfully read;

[0010] Step 3: The data collection module collects a reference data set by connecting to the database through a network. The reference data set includes the appearance characteristics of the cold chain ingredients during normal storage and the storage records of all cold chain ingredients.

[0011] Step 4: The data recognition module analyzes and calculates the safety score Aqpf of the cold chain food according to the barcode data set T and the reference data set. If the safety score Aqpf of the cold chain food is 1, the quality assessment of the cold chain food continues;

[0012] Step 5: The data recognition module is provided with a ResNet algorithm model. The data recognition module generates a residual block Cck according to a reference data set, and introduces the ResNet algorithm model to limit the gradient disappearance and network degradation of the model.

[0013] Step 6: The ResNet algorithm model with the residual block Cck introduced performs quality assessment on the image dataset and analyzes and calculates the quality score Zlpf of the cold chain food;

[0014] Step 7. The data identification module classifies cold chain ingredients according to their safety score Aqpf and quality score Zlpf, marks them as qualified or unqualified, and then manages the ingredients according to the different marks.

[0015] Preferably, in step 1, the expression of the barcode data set T is to Corresponding to each cold chain food whose barcode information is successfully read, m represents the information contained in the cold chain food barcode, the barcode information includes the food source, food name, production date and batch, and the digital numbers 1 to n represent that there are n cold chain foods whose barcode information is successfully read by the barcode reader.

[0016] Preferably, in step 2, the expression of the image data set is to The appearance images of cold chain food materials that successfully read the barcode information are sequentially corresponding to each image. w represents the number of appearance images of a single cold chain food material, and the number of images is the same as the number of planes on the surface of the cold chain food material. The digital numbers 1 to n represent that there are n cold chain food materials photographed by the high-definition camera. The digital numbers in the barcode dataset T correspond to the digital numbers in the image dataset one by one. and image datasets They all represent the same cold chain ingredients.

[0017] Preferably, in step 3, the expression of the reference data set is {CZ tz , CZ rk}, CZ tz Indicates the appearance characteristics of cold chain food during normal storage. The appearance characteristics of cold chain food during normal storage include normal color and normal size. CZ rk It represents the storage record of all cold chain ingredients, which includes the source, ingredient name, production date and batch of all cold chain ingredients in the cold chain warehouse.

[0018] Preferably, in step 4, the safety score Aqpf calculation formula is as follows:

[0019]

[0020] In the formula, Aqpf represents the safety score, T∩CZ rk The intersection of all the cold chain food data with successfully read barcode information in the barcode data set and the storage records of all cold chain food in the reference data set, that is, the safety score of all cold chain food with successfully read barcode information and existing in the storage records is 1, T-(T∩CZ rk ) represents the remaining part after removing the data in the barcode dataset that also appears in the reference dataset, that is, the safety score of all cold chain food ingredients whose barcode information is successfully read but does not exist in the warehouse entry record is 0.

[0021] Preferably, in step 5, the residual block Cck calculation formula is as follows:

[0022] Cck=h(CZ tz )+F(CZ tz )

[0023] In the formula, Cck represents the residual block, h(CZ tz ) represents the identity mapping, which refers to the appearance characteristics of cold chain food in normal storage in the reference dataset. tz ) represents the learning part of the residual block, which is the appearance features of cold chain food during normal storage after learning transformation.

[0024] Preferably, in step 6, the quality score Zlpf is calculated using the following formula:

[0025]

[0026] In the formula, Zlpf represents the quality score, represents the appearance image of the i-th cold chain food in the image dataset, Indicates that the appearance image of the i-th cold chain food is compared with the appearance features of the cold chain food in the reference dataset when it is normally stored. Indicates that the quality score of the i-th cold chain food is 1. Indicates that the quality score of the i-th cold chain food is 0, It indicates that the ResNet algorithm model with the introduced residual block Cck simultaneously calculates the comparison results of the appearance images of n cold chain ingredients and the appearance features of the cold chain ingredients when they are normally stored in the reference dataset.

[0027] Preferably, in step seven, if the safety score Aqpf and the quality score Zlpf of the cold chain food are both 1, the data recognition module will mark it as qualified and allow it to continue to be stored or shipped out for sale.

[0028] Preferably, in step seven, if the score of any one of the safety score Aqpf and the quality score Zlpf of the cold chain food is 0, the data identification module will mark it as unqualified, and will not allow it to be stored further, and will remind the management personnel to isolate and destroy it.

[0029] A cold chain food identification system based on image processing, comprising a data acquisition module and a data identification module;

[0030] The data acquisition module reads all cold chain food barcodes through a barcode reader. If the food barcode is successfully read by the barcode reader, the data acquisition module forms a barcode data set T. If there is a food barcode that is not read by the barcode reader, the data acquisition module marks it as abnormal and reminds the management personnel to conduct investigation and management. The data acquisition module connects to a high-definition camera through a network to collect an image data set, and the image data set includes all appearance images of cold chain food that have successfully read barcode information. The data acquisition module connects to a database through a network to collect a reference data set, and the reference data set includes the appearance characteristics of cold chain food during normal storage and the storage records of all cold chain food. The data acquisition module transmits the barcode data set T, the image data set and the reference data set to the data recognition module through the network;

[0031] The data recognition module analyzes and calculates the safety score Aqpf of the cold chain food according to the barcode data set T and the reference data set. If the safety score Aqpf of the cold chain food is 1, the quality assessment of the cold chain food continues. The data recognition module is provided with a ResNet algorithm model. The data recognition module generates a residual block Cck according to the reference data set and introduces the ResNet algorithm model. The ResNet algorithm model with the introduced residual block Cck performs quality assessment on the image data set, analyzes and calculates the quality score Zlpf of the cold chain food. The data recognition module classifies the cold chain food according to the safety score Aqpf and the quality score Zlpf, marks them as qualified or unqualified, and then manages the food according to different marks.

[0032] Compared with the prior art, the present invention provides a cold chain food identification method and system based on image processing, which has the following beneficial effects:

[0033] 1. The present invention uses a barcode reader to read all cold chain food barcodes. If the food barcode is successfully read by the barcode reader, the data acquisition module will compose the barcode data set T. The data acquisition module uses a high-definition camera to capture the appearance images of all cold chain food. The high-definition camera takes photos of each surface one by one according to the number of planes on the surface of each cold chain food, and composes them into an image data set. The data acquisition module connects to the database through a network to collect a reference data set. The reference data set includes the appearance characteristics of the cold chain food during normal storage and the storage records of all cold chain food. The data recognition module analyzes the barcode data set T and the reference data set. Calculate the safety score Aqpf of cold chain ingredients. If the safety score Aqpf of cold chain ingredients is 1, and ingredients with clear traceable sources can ensure that the ingredients meet food safety and quality standards, then continue to conduct quality assessment on cold chain ingredients. The data recognition module is equipped with a ResNet algorithm model. The data recognition module generates a residual block Cck based on a reference data set, and introduces the ResNet algorithm model to limit the gradient vanishing and network degradation of the model, maintain the integrity of the input information, and deal with industrial automation problems such as food picking. It can better meet the high efficiency and high precision requirements of modern industrial production, and deep learning retains the original features for higher recognition accuracy.

[0034] 2. The present invention performs quality assessment on the image data set through the ResNet algorithm model that has introduced the residual block Cck, analyzes and calculates the quality score Zlpf of the cold chain food, and the data recognition module classifies the cold chain food according to the safety score Aqpf and the quality score Zlpf, and marks it as qualified or unqualified. If the safety score Aqpf and the quality score Zlpf of the cold chain food are both 1, the data recognition module marks it as qualified and allows it to continue to be stored or shipped out for sale. If any one of the safety score Aqpf and the quality score Zlpf of the cold chain food is 0, the data recognition module marks it as unqualified and does not allow it to continue to be stored, and reminds the management personnel to isolate and destroy it. Multi-dimensional recognition is not easy to make mistakes and omissions, and the processing efficiency is high. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a step diagram of the method of the present invention;

[0036] Figure 2 It is a schematic diagram of the system flow of the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] Since traditional cold chain food identification methods based on image processing mostly rely on deep learning algorithm models combined with manual recognition methods to judge whether the food quality is qualified based on the food appearance image, but in actual use, the deep learning algorithm model is prone to gradient disappearance and network degradation problems, and the food appearance image is prone to subtle differences that are difficult for the human eye to distinguish, and it is easy to make misjudgments and missed judgments, and the recognition processing efficiency is low. Therefore, a cold chain food identification method and system based on image processing is provided. Please refer to Figure 1-Figure 2 , the identification method comprises the following steps:

[0039] Step 1: The data collection module reads all the cold chain food barcodes through the barcode reader and forms a barcode data set T. The barcode data set T includes all the cold chain food data whose barcode information is successfully read. The expression of the barcode data set T is: to Corresponding to each cold chain food whose barcode information is successfully read, m represents the information contained in the cold chain food barcode. The barcode information includes the source of the food, the name of the food, the production date and the batch, which is convenient for quickly tracing the production information of the cold chain food and facilitating the subsequent accurate judgment of the safety of the cold chain food. The digital numbers 1 to n represent that there are n cold chain food whose barcode information is successfully read by the barcode reader;

[0040] Step 2: The data acquisition module uses a high-definition camera to capture the appearance images of all cold chain ingredients. The high-definition camera takes photos of each side one by one according to the number of planes on the surface of each cold chain ingredient. For example, when the cold chain ingredient is half a pork loin, the high-definition camera will take photos of the front and back of the half pork loin one by one and form an image dataset. The image dataset includes all the appearance images of the cold chain ingredients that have successfully read the barcode information. The expression of the image dataset is: to The appearance images of cold chain food materials that successfully read the barcode information are sequentially corresponding to each image. w represents the number of appearance images of a single cold chain food material, and the number of images is the same as the number of planes on the surface of the cold chain food material. The digital numbers 1 to n represent that there are n cold chain food materials photographed by the high-definition camera. The digital numbers in the barcode dataset T correspond to the digital numbers in the image dataset one by one. and image datasets They all represent the same cold chain food, with one item and one code, and then corresponding to a detailed appearance image, providing a reliable data source for subsequent identification and evaluation;

[0041] Step 3: The data collection module collects the reference data set through the network connection database. The reference data set includes the appearance characteristics of cold chain ingredients during normal storage and the storage records of all cold chain ingredients. The expression of the reference data set is {CZ tz, CZ rk}, CZ tz Indicates the appearance characteristics of cold chain food during normal storage. The appearance characteristics of cold chain food during normal storage include normal color and normal size. For example, the normal color of chicken in cold chain food should be pinkish white with a slight pink color. If the appearance color of cold chain chicken is found to be yellow, black, gray, green or brown during the identification process, it means that mold spots and deterioration have occurred. CZ rk Indicates the storage records of all cold chain ingredients. The storage records include the source, ingredient name, production date and batch of all cold chain ingredients in the cold chain warehouse, which is convenient for matching and identifying ingredient barcode information according to the storage records to prevent the ingredient barcode information from being maliciously tampered with;

[0042] Step 4: The data recognition module analyzes and calculates the safety score Aqpf of cold chain food according to the barcode data set T and the reference data set. The calculation formula is as follows:

[0043]

[0044] In the formula, Aqpf represents the safety score, T∩CZ rk It represents the intersection of all the cold chain food data with successfully read barcode information in the barcode data set and the storage records of all the cold chain food in the reference data set, that is, all the cold chain food safety scores of successfully read barcode information and existing in the storage records are 1. Food with clear traceable sources can ensure that the food meets food safety and quality standards and respond to recalls or safety issues in a timely manner. rk ) indicates the remaining part after removing the data in the barcode data set that also appears in the reference data set, that is, the safety score of all cold chain ingredients whose barcode information is successfully read but does not exist in the storage record is 0. The production source and production date of ingredients of unknown origin cannot be traced, and the food safety risk is high;

[0045] If the cold chain food safety score Aqpf is 1, it means that the food has a clear source, and the quality assessment of the cold chain food will continue;

[0046] Step 5: The data recognition module is equipped with a ResNet algorithm model. The data recognition module generates a residual block Cck according to the reference data set, and introduces the ResNet algorithm model to limit the gradient disappearance and network degradation of the model. The calculation formula is as follows:

[0047] Cck=h(CZ tz )+F(CZ tz )

[0048] In the formula, Cck represents the residual block. The core idea of ​​the residual block is to allow the input data to be directly transmitted to the subsequent layers. No matter how many layers of processing are passed in the middle, an identity mapping can be retained. h(CZtz ) represents the identity mapping, which refers to the appearance characteristics of cold chain food in normal storage in the reference dataset. tz represents the learning part of the residual block, that is, the appearance features of cold chain food in normal storage after learning transformation, even if F(CZ tz )The learned weights are close to zero. Without any complex transformation, the entire ResNet algorithm model can still be obtained through h(CZ tz ) Maintaining the integrity of input information, when dealing with industrial automation problems such as food picking, it can better meet the high efficiency and high precision requirements of modern industrial production. Deep learning retains the original features with higher recognition accuracy;

[0049] Step 6. The ResNet algorithm model with the residual block Cck introduced performs quality assessment on the image dataset and analyzes and calculates the quality score Zlpf of the cold chain food. The calculation formula is as follows:

[0050]

[0051] In the formula, Zlpf represents the quality score, represents the appearance image of the i-th cold chain food in the image dataset, Indicates that the appearance image of the i-th cold chain food is compared with the appearance features of the cold chain food in the reference dataset when it is normally stored. The appearance image of the cold chain food is identical to the appearance features of the cold chain food during normal storage, indicating that the quality score of the i-th cold chain food is 1. The appearance image of the cold chain food is different from the appearance features of the cold chain food during normal storage, indicating that the quality score of the i-th cold chain food is 0. It indicates that the ResNet algorithm model with the introduced residual block Cck simultaneously calculates the comparison results of the appearance images of n cold chain ingredients and the appearance features of the cold chain ingredients when they are normally stored in the reference data set. The network depth of the ResNet algorithm model with the introduced residual block Cck is greatly increased, and it can better capture subtle image differences during the image content comparison process;

[0052] Step 7. The data recognition module classifies the cold chain ingredients according to their safety score Aqpf and quality score Zlpf. If both the safety score Aqpf and the quality score Zlpf of the cold chain ingredients are 1, the data recognition module marks them as qualified and allows them to continue to be stored or shipped out for sale. If any of the safety score Aqpf and the quality score Zlpf of the cold chain ingredients is 0, the data recognition module marks them as unqualified and does not allow them to continue to be stored. The management personnel are reminded to isolate and destroy them. Multi-dimensional recognition is not prone to errors and omissions and has high processing efficiency.

[0053] The above-mentioned cold chain food identification method based on image processing is implemented by a cold chain food identification system based on image processing, which includes a data acquisition module and a data identification module;

[0054] The data acquisition module reads all the cold chain food barcodes through the barcode reader. If the food barcode is successfully read by the barcode reader, the data acquisition module forms a barcode data set T. If there is a food barcode that is not read by the barcode reader, the data acquisition module marks it as abnormal, reminding the management personnel to conduct investigation and management. It is necessary to manually investigate whether the food barcode is lost or tampered with. The data acquisition module connects to the high-definition camera through the network to collect the image data set. The image data set includes all the appearance images of the cold chain food that successfully read the barcode information. The data acquisition module connects to the database through the network to collect the reference data set. The reference data set includes the appearance characteristics of the cold chain food during normal storage and the storage records of all the cold chain food. The data acquisition module transmits the barcode data set T, the image data set and the reference data set to the data recognition module through the network;

[0055] The data recognition module analyzes and calculates the safety score Aqpf of the cold chain ingredients based on the barcode data set T and the reference data set. If the safety score Aqpf of the cold chain ingredients is 1, the quality assessment of the cold chain ingredients continues. The data recognition module is provided with a ResNet algorithm model. The data recognition module generates a residual block Cck based on the reference data set and introduces the ResNet algorithm model. Deep learning retains the original features for higher recognition accuracy. The ResNet algorithm model with the introduced residual block Cck performs quality assessment on the image data set and analyzes and calculates the quality score Zlpf of the cold chain ingredients. The data recognition module classifies the cold chain ingredients according to the safety score Aqpf and the quality score Zlpf, marks them as qualified or unqualified, and then manages the ingredients according to different marks. Multi-dimensional recognition is not prone to errors and omissions and has high processing efficiency.

[0056] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A cold chain food identification method based on image processing, characterized in that: The following steps are involved: Step 1: The data collection module reads all the cold chain food barcodes through a barcode reader and forms a barcode data set z. The barcode data set T includes all the cold chain food data whose barcode information is successfully read; Step 2: The data acquisition module uses a high-definition camera to capture the appearance images of all cold chain ingredients. The high-definition camera takes photos of each surface one by one according to the number of planes on the surface of each cold chain ingredient, and forms an image data set. The image data set includes the appearance images of all cold chain ingredients whose barcode information has been successfully read; Step 3: The data collection module collects a reference data set by connecting to the database through a network. The reference data set includes the appearance characteristics of the cold chain ingredients during normal storage and the storage records of all cold chain ingredients. Step 4: The data recognition module analyzes and calculates the safety score Aqpf of the cold chain food according to the barcode data set T and the reference data set. If the safety score Aqpf of the cold chain food is 1, the quality assessment of the cold chain food continues; Step 5: The data recognition module is provided with a ResNet algorithm model. The data recognition module generates a residual block Cck according to a reference data set, and introduces the ResNet algorithm model to limit the gradient disappearance and network degradation of the model. Step 6: The ResNet algorithm model with the residual block Cck introduced performs quality assessment on the image dataset and analyzes and calculates the quality score Zlpf of the cold chain food; Step 7. The data identification module classifies cold chain ingredients according to their safety score Aqpf and quality score Zlpf, marks them as qualified or unqualified, and then manages the ingredients according to the different marks.

2. According to claim 1, a cold chain food identification method based on image processing is characterized in that: In step 1, the expression of the barcode dataset T is: to Corresponding to each cold chain food whose barcode information is successfully read, m represents the information contained in the cold chain food barcode, the barcode information includes the food source, food name, production date and batch, and the digital numbers 1 to n represent that there are n cold chain foods whose barcode information is successfully read by the barcode reader.

3. The cold chain food identification method based on image processing according to claim 2 is characterized in that: In step 2, the expression of the image data set is: to The appearance images of cold chain food materials that successfully read the barcode information are sequentially corresponding to each image. w represents the number of appearance images of a single cold chain food material, and the number of images is the same as the number of planes on the surface of the cold chain food material. The digital numbers 1 to n represent that there are n cold chain food materials photographed by the high-definition camera. The digital numbers in the barcode dataset T correspond to the digital numbers in the image dataset one by one. and image datasets They all represent the same cold chain ingredients.

4. The cold chain food identification method based on image processing according to claim 3 is characterized in that: In step 3, the expression of the reference data set is {CZ tz , CZ rk }, CZ tz Indicates the appearance characteristics of cold chain food during normal storage. The appearance characteristics of cold chain food during normal storage include normal color and normal size. CZ rk It represents the storage record of all cold chain ingredients, which includes the source, ingredient name, production date and batch of all cold chain ingredients in the cold chain warehouse.

5. The cold chain food identification method based on image processing according to claim 4 is characterized in that: In step 4, the safety score Aqpf is calculated as follows: In the formula, Aqpf represents the safety score, T∩CZ rk The intersection of all the cold chain food data with successfully read barcode information in the barcode data set and the storage records of all cold chain food in the reference data set, that is, the safety score of all cold chain food with successfully read barcode information and existing in the storage records is 1, T-(T∩CZ rk ) represents the remaining part after removing the data in the barcode dataset that also appears in the reference dataset, that is, the safety score of all cold chain food ingredients whose barcode information is successfully read but does not exist in the warehouse entry record is 0.

6. The cold chain food identification method based on image processing according to claim 5 is characterized in that: In step 5, the residual block Cck is calculated as follows: Cck=h(CZ tz )+F(CZ tz ) In the formula, Cck represents the residual block, h(CZ tz ) represents the identity mapping, which refers to the appearance characteristics of cold chain food in normal storage in the reference dataset. tz ) represents the learning part of the residual block, which is the appearance features of cold chain food during normal storage after learning transformation.

7. The cold chain food identification method based on image processing according to claim 6 is characterized in that: In step 6, the quality score Zlpf is calculated as follows: In the formula, Zlpf represents the quality score, represents the appearance image of the i-th cold chain food in the image dataset, Indicates that the appearance image of the i-th cold chain food is compared with the appearance features of the cold chain food in the reference dataset when it is normally stored. Indicates that the quality score of the i-th cold chain food is 1. Indicates that the quality score of the i-th cold chain food is 0, It indicates that the ResNet algorithm model with the introduced residual block Cck simultaneously calculates the comparison results of the appearance images of n cold chain ingredients and the appearance features of the cold chain ingredients when they are normally stored in the reference dataset.

8. The cold chain food identification method based on image processing according to claim 7 is characterized in that: In step seven, if the safety score Aqpf and the quality score Zlpf of the cold chain food are both 1, the data recognition module will mark it as qualified and allow it to continue to be stored or shipped out for sale.

9. The cold chain food identification method based on image processing according to claim 8, characterized in that: In step seven, if any of the safety score Aqpf and quality score Zlpf of the cold chain food is 0, the data identification module will mark it as unqualified, not allowing it to be stored further, and reminding the management personnel to isolate and destroy it.

10. A cold chain food identification system based on image processing, applied to the cold chain food identification method based on image processing according to claims 1-9, characterized in that: It includes a data acquisition module and a data identification module; The data acquisition module reads all cold chain food barcodes through a barcode reader. If the food barcode is successfully read by the barcode reader, the data acquisition module forms a barcode data set T. If there is a food barcode that is not read by the barcode reader, the data acquisition module marks it as abnormal and reminds the management personnel to conduct investigation and management. The data acquisition module connects to a high-definition camera through a network to collect an image data set, and the image data set includes all appearance images of cold chain food that have successfully read barcode information. The data acquisition module connects to a database through a network to collect a reference data set, and the reference data set includes the appearance characteristics of cold chain food during normal storage and the storage records of all cold chain food. The data acquisition module transmits the barcode data set T, the image data set and the reference data set to the data recognition module through the network; The data recognition module analyzes and calculates the safety score Aqpf of the cold chain food according to the barcode data set T and the reference data set. If the safety score Aqpf of the cold chain food is 1, the quality assessment of the cold chain food continues. The data recognition module is provided with a ResNet algorithm model. The data recognition module generates a residual block Cck according to the reference data set and introduces the ResNet algorithm model. The ResNet algorithm model with the introduced residual block Cck performs quality assessment on the image data set, analyzes and calculates the quality score Zlpf of the cold chain food. The data recognition module classifies the cold chain food according to the safety score Aqpf and the quality score Zlpf, marks them as qualified or unqualified, and then manages the food according to different marks.

Images