Food additive detection method based on artificial intelligence

Abstract

The invention belongs to the technical field of food additive detection, and particularly relates to a food additive detection method based on artificial intelligence, which comprises the following steps: S1, shooting a package ingredient table to obtain a package ingredient table photo; s2, transmitting the shot photo to a cloud for AI identification, and extracting additive names and dosage data on the photo; s3, establishing a food additive early warning database according to WHO, EFSA and GB2760 international and domestic authoritative databases; s4, the extracted additive data is compared with data in a food additive early warning database, and related information output and early warning are carried out. People can rapidly and conveniently know the risk levels, effects and potential influences on health of various additives in the food purchasing process; therefore, people are helped to purchase and select food, and body health and safety are guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of food additive detection, and in particular to a food additive detection method based on artificial intelligence. Background Art

[0002] Food additives refer to chemically synthesized or naturally occurring substances added to food to improve its quality, color, aroma, and flavor, as well as for preservation and processing needs. Many food additives are toxic, some even highly toxic. However, if they are used within the prescribed limits and dosages, they are harmless to the human body. Commonly Used Food Additives: China's commodity classification includes 35 categories of food additives, including flavor enhancers, defoaming agents, leavening agents, colorants, and preservatives. Over 10,000 foods contain additives. Of these, the "Standards for the Use of Food Additives" and Ministry of Health announcements allow for use in 23 categories, totaling over 2,400 types. National or industry quality standards have been established for 364 types. There are mainly 23 categories including acidity regulators, anti-caking agents, defoaming agents, antioxidants, bleaching agents, leavening agents, base substances in gum-based candies, colorants, color preservatives, emulsifiers, enzyme preparations, flavor enhancers, flour treatment agents, coating agents, moisture retention agents, nutritional enhancers, preservatives, stabilizers and coagulants, sweeteners, thickeners, food flavorings, processing aids for the food industry, and others.

[0003] Preservatives - Commonly used ones include sodium benzoate, potassium sorbate, sulfur dioxide, and lactic acid. They are used in the processing of foods such as jams and preserves. Antioxidants - Similar to preservatives, they can extend the shelf life of foods. Commonly used ones include vitamin C and isotocin. Colorants - Commonly used synthetic pigments include carmine, amaranth, tartrazine, and indigo. They can change the appearance of food and make it more appetizing. Thickeners and stabilizers - They can improve or stabilize the physical properties of cold drinks, making them smooth and delicate. They help frozen foods such as ice cream maintain a soft, loose texture for a long time. Raising agents - Raising agents are added to some candies and chocolates to encourage sugars to produce carbon dioxide, thereby causing the food to expand. Commonly used raising agents include sodium bicarbonate, ammonium bicarbonate, and compound raising agents. Sweeteners - Commonly used synthetic sweeteners include sodium saccharin and sodium cyclamate. Their purpose is to increase the sweetness. Acidulants - Some beverages and candies often use acidulants to adjust and improve flavor. Commonly used are citric acid, tartaric acid, malic acid, and lactic acid. Whitening agents - Benzoyl peroxide is the main ingredient in flour whitening agents. The maximum dosage allowed in Chinese food flour is 0.06g / kg. Excessive amounts of whitening agents can destroy the nutritional value of flour, and the benzoic acid produced after hydrolysis can damage the liver. Benzoyl peroxide has been banned as a food additive in developed countries such as the European Union. my country also banned benzoyl peroxide as a whitening agent in May 2011. Flavorings - Flavorings are available in a variety of flavors, both synthetic and natural. The various flavors of chocolate commonly consumed by consumers utilize a wide range of spices during production, giving them their unique flavors.

[0004] The existing technology cannot help people understand the obscure names of food additives in the ingredient lists of various food packages, cannot enable people to quickly and conveniently understand the risk levels, functions and potential health effects of various additives during the food purchasing process, and cannot help people make food purchase choices to ensure their own health and safety. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of not being able to help people understand the obscure names of food additives in the ingredient lists of various food packages, not being able to let people quickly and conveniently understand the risk levels, functions and potential health effects of various additives during the food procurement process, and not being able to help people make food purchase choices and protect their own health and safety. The food additive detection method based on artificial intelligence is proposed.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] The food additive detection method based on artificial intelligence includes the following steps:

[0008] S1. Take a picture of the packaging ingredient list to obtain a photo of the packaging ingredient list;

[0009] S2. The captured photos are transmitted to the cloud for AI recognition to extract the names and dosage data of the additives in the photos;

[0010] S3. Establish a food additive early warning database based on the WHO, EFSA, and GB2760 international and domestic authoritative databases;

[0011] S4. Compare the extracted additive data with the data in the food additive early warning database, and output relevant information and issue early warnings.

[0012] Preferably, in S1, a K210 development board is used with an integrated audio-visual architecture and a neural network accelerator KPU as the core, and the YOLO algorithm is run at millisecond level to realize dynamic target detection and complete the photo shooting function.

[0013] Preferably, in S2, the WiFi function of the ESP32 chip is used, and the WPA2 and WPA3 encryption algorithms are used to upload the photo information to Baidu AI. The algorithms used by Baidu AI graphic and text recognition mainly include multimodal fusion architecture, dynamic handwriting modeling, and small sample learning capabilities.

[0014] Preferably, in said S3, a food additive early warning database is established based on the international and domestic authoritative databases of WHO, EFSA and GB2760, and the latest additive limit data of GB2760, EFSA and WHO are integrated in real time to update the data in the food additive early warning database.

[0015] Preferably, in S2, the taken photo is transmitted to the cloud for AI recognition, the additive name data on the photo is extracted, and the food is directly scanned by near-infrared spectroscopy or Raman spectroscopy technology to verify the consistency of the actual ingredients of the additives with the packaging label.

[0016] Preferably, in said S4, the extracted additive data is compared with the data in the food additive early warning database, and relevant information is output and early warning is performed, and the information output and early warning methods include vision, vibration, and voice.

[0017] Preferably, in said S4, the extracted data on the names and amounts of additives are compared with the data in the food additive early warning database, the differences between domestic and foreign standards are compared, a threshold is set, and when the extracted data on the names and amounts of additives exceed the threshold, an early warning is triggered.

[0018] Preferably, the extracted additive data is compared with the data in the food additive early warning database, and relevant information is output and early warning is carried out. The early warning adopts a graded system, including general warning, low-risk warning and high-risk warning.

[0019] Compared with the prior art, the application has the advantages that:

[0020] Quick identification: can quickly and conveniently know what food additives are in the food to be eaten, whether it is safe, especially for those who are prone to gout, asthma, allergies, pregnant people, so that they can reasonably choose food purchase, protect their own health and safety;

[0021] Scientific classification: spread scientific knowledge of food additives to everyone, because food additives have very strict use specifications in China, and most food additives will not harm our body if used within the safe range, so we should avoid the behavior of excessive, excessive, non-standard use of food additives;

[0022] Focus on food safety: through the popularization of food additive knowledge, let people pay attention to their food health and safety, can eat less snacks, will eat snacks, away from junk food.

[0023] The application enables people to quickly and conveniently understand the risk level, effect and potential impact on health of various additives during food purchase, so as to help people make food purchase selection and protect their own health and safety. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 The flowchart of the food additive detection method based on artificial intelligence provided by the application. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments will be described clearly and completely in combination with the drawings in the embodiments. Obviously, the described embodiments are only part of the embodiments, not all the embodiments.

[0026] Embodiment one

[0027] Referring to Figure 1 The food additive detection method based on artificial intelligence comprises the following steps:

[0028] S1, photographing the packaging ingredient table to obtain a packaging ingredient table photo;

[0029] S2, transmitting the photographed photo to the cloud for AI recognition to extract the additive name and dosage data on the photo;

[0030] S3, establishing a food additive early warning database according to WHO, EFSA and GB2760 international and domestic authoritative databases;

[0031] S4, compare the extracted additive data with the data in the food additive early warning database, and output and warn related information.

[0032] In this embodiment, in S1, the K210 development board is used to take photos with the photo shooting function realized by using the YOLO algorithm with the millisecond-level running of the audio-visual integrated architecture and the neural network accelerator KPU as the core.

[0033] In this embodiment, in S2, the WiFi function of the ESP32 chip is used, and the WPA2 and WPA3 encryption algorithms are used to upload the photo information to Baidu AI. The algorithms used by Baidu AI graphic character recognition mainly include a multi-modal fusion architecture, dynamic handwriting modeling, and small sample learning capability.

[0034] In this embodiment, in S3, the food additive early warning database is established according to the WHO, EFSA, and GB2760 international and domestic authoritative databases, and the latest additive limit data of GB2760, EFSA, and WHO are integrated in real time to update the data in the food additive early warning database.

[0035] In this embodiment, in S2, the photographed photos are transmitted to the cloud for AI recognition, and the additive name data on the photos are extracted. Through near-infrared spectrum or Raman spectrum technology, the consistency of the actual ingredients of the additive and the packaging label is directly scanned and verified.

[0036] In this embodiment, in S4, the extracted additive data is compared with the data in the food additive early warning database, and related information is output and warned. The information output and warning methods include vision, vibration, and voice.

[0037] In this embodiment, in S4, the extracted additive name and dosage data are compared with the data in the food additive early warning database, the differences between domestic and foreign standards are compared, a threshold value is set, and when the extracted additive name and dosage data exceed the threshold value, the warning is triggered.

[0038] In this embodiment, in S4, the extracted additive data is compared with the data in the food additive early warning database, and related information is output and warned. The warning adopts a hierarchical system, including general warning, low-risk warning, and high-risk warning.

[0039] Embodiment Two

[0040] The food additive detection method based on artificial intelligence comprises the following steps:

[0041] S1, take a photo of the packaging ingredient table, use image denoising, enhancement, and segmentation technology to improve the recognition rate of reflective packaging or blurred fonts, and obtain a packaging ingredient table photo;

[0042] S2. The captured photos are transmitted to the cloud for AI recognition to extract the names and dosage data of the additives in the photos;

[0043] S3. Establish a food additive early warning database based on the WHO, EFSA, and GB2760 international and domestic authoritative databases;

[0044] S4. Compare the extracted additive data with the data in the food additive early warning database, and output relevant information and issue early warnings.

[0045] In this embodiment, in S2, the WiFi function of the ESP32 chip is used, and the WPA2 and WPA3 encryption algorithms are used to upload the photo information to Baidu AI. The algorithms used by Baidu AI graphic and text recognition mainly include multimodal fusion architecture, dynamic handwriting modeling, and small sample learning capabilities.

[0046] In this embodiment, in S3, a food additive early warning database is established based on the international and domestic authoritative databases of WHO, EFSA, and GB2760, and the latest additive limit data of GB2760, EFSA, and WHO are integrated in real time to update the data in the food additive early warning database.

[0047] In this embodiment, in S2, the captured photo is transmitted to the cloud for AI recognition, the additive name data on the photo is extracted, and the food is directly scanned through near-infrared spectroscopy or Raman spectroscopy technology to verify the consistency of the actual ingredients of the additives with the packaging label.

[0048] In this embodiment, in S4, the extracted additive data is compared with the data in the food additive warning database, and relevant information is output and warning is issued. The information output and warning methods include vision, vibration, and voice.

[0049] In this embodiment, in S4, the extracted additive name and usage data are compared with the data in the food additive early warning database, and the differences between domestic and foreign standards are compared. A threshold is set, and when the extracted additive name and usage data exceed the threshold, an early warning is triggered.

[0050] In this embodiment, in S4, the extracted additive data is compared with the data in the food additive early warning database, and relevant information is output and early warning is issued. The early warning adopts a graded system, including general warning, low-risk warning and high-risk warning.

[0051] Example 3

[0052] The food additive detection method based on artificial intelligence includes the following steps:

[0053] S1. Take a picture of the packaging ingredient list to obtain a photo of the packaging ingredient list;

[0054] S2, the photographed photo is transmitted to the cloud for AI recognition, a transfer learning optimized OCR model is introduced to adapt to different languages and font formats of the ingredient table, to reduce the recognition error rate, and to extract the additive name and dosage data on the photo;

[0055] S3, a food additive early warning database is established according to WHO, EFSA and GB2760 international and domestic authoritative databases;

[0056] S4, the extracted additive data is compared with the data in the food additive early warning database, and relevant information output and early warning are performed.

[0057] In this embodiment, in S1, a K210 development board is used to take photos with a visual and audio integrated architecture and a neural network accelerator KPU as the core, a millisecond-level YOLO algorithm is used to realize dynamic target detection, and a photo shooting function is completed.

[0058] In this embodiment, in S3, a food additive early warning database is established according to WHO, EFSA and GB2760 international and domestic authoritative databases, GB2760, EFSA and WHO latest additive limit data are integrated in real time, and the data in the food additive early warning database is updated.

[0059] In this embodiment, in S2, the photographed photo is transmitted to the cloud for AI recognition, additive name data on the photo is extracted, and near-infrared spectrum or Raman spectrum technology is used to directly scan the food to verify the consistency of the actual ingredient of the additive and the packaging label.

[0060] In this embodiment, in S4, the extracted additive data is compared with the data in the food additive early warning database, and relevant information output and early warning are performed, and the information output and early warning methods include vision, vibration and voice.

[0061] In this embodiment, in S4, the extracted additive name and dosage data are compared with the data in the food additive early warning database, the differences between domestic and foreign standards are compared, a threshold value is set, and when the extracted additive name and dosage data exceed the threshold value, early warning is triggered.

[0062] In this embodiment, in S4, the extracted additive data is compared with the data in the food additive early warning database, and relevant information output and early warning are performed, and the early warning adopts a grading system, including general early warning, low-risk early warning and high-risk early warning.

[0063] The above is only a preferred specific embodiment of the present embodiment, but the protection scope of the present embodiment is not limited thereto, any person skilled in the art can make equivalent replacement or change according to the technical scheme and inventive concept of the present embodiment within the technical range disclosed by the present embodiment, which should be covered within the protection scope of the present embodiment.

Claims

1. A food additive detection method based on artificial intelligence, characterized in that: The following steps are involved: S1. Take a picture of the packaging ingredient list to obtain a photo of the packaging ingredient list; S2. The captured photos are transmitted to the cloud for AI recognition to extract the names and dosage data of the additives in the photos; S3. Establish a food additive early warning database based on the WHO, EFSA, and GB2760 international and domestic authoritative databases; S4. Compare the extracted additive data with the data in the food additive early warning database, and output relevant information and issue early warnings.

2. The food additive detection method based on artificial intelligence according to claim 1, characterized in that: In the S1, the K210 development board is used with an integrated audio-visual architecture and a neural network accelerator KPU as the core. The YOLO algorithm runs in milliseconds to achieve dynamic target detection and complete the photo shooting function.

3. The food additive detection method based on artificial intelligence according to claim 1, characterized in that: In S2, the WiFi function of the ESP32 chip is used, and the WPA2 and WPA3 encryption algorithms are used to upload photo information to Baidu AI. The algorithms used by Baidu AI for graphic and text recognition mainly include multimodal fusion architecture, dynamic handwriting modeling, and small sample learning capabilities.

4. The food additive detection method based on artificial intelligence according to claim 1, characterized in that: In said S3, a food additive early warning database is established based on the international and domestic authoritative databases of WHO, EFSA and GB2760, and the latest additive limit data of GB2760, EFSA and WHO are integrated in real time to update the data in the food additive early warning database.

5. The food additive detection method based on artificial intelligence according to claim 1, characterized in that: In S2, the captured photos are transmitted to the cloud for AI recognition, and the name data of the additives in the photos are extracted. By using near-infrared spectroscopy or Raman spectroscopy technology, the food is directly scanned to verify the consistency between the actual ingredients of the additives and the packaging labels.

6. The food additive detection method based on artificial intelligence according to claim 1, characterized in that: In said S4, the extracted additive data is compared with the data in the food additive early warning database, and relevant information is output and early warning is carried out, and the information output and early warning methods include visual, vibration, and voice.

7. The food additive detection method based on artificial intelligence according to claim 1, characterized in that: In said S4, the extracted data on the names and amounts of additives are compared with the data in the food additive early warning database, and the differences between domestic and foreign standards are compared. A threshold is set, and when the extracted data on the names and amounts of additives exceed the threshold, an early warning is triggered.

8. The food additive detection method based on artificial intelligence according to claim 1, characterized in that: In said S4, the extracted additive data is compared with the data in the food additive early warning database, and relevant information is output and early warning is carried out. The early warning adopts a graded system, including general warning, low-risk warning and high-risk warning.

Images