Cooking robot food material freshness identification method and system based on AI

Abstract

The invention relates to the technical field of machine learning, in particular to a cooking robot food material freshness identification method and system based on AI. The method comprises the following steps: acquiring a food material shooting image; performing food material feature point extraction on the food material shooting image, and performing food material type judgment on the food material shooting image according to the food material feature points to obtain a single food material judgment result and a mixed food material judgment result; performing similar food material freshness identification on the standard food material shot image based on the single food material judgment result to obtain single food material freshness identification data; performing volatile gas data acquisition on the food materials based on the mixed food material judgment result to obtain mixed food material volatile gas data; abnormal gas screening is conducted on the mixed food material volatile gas data, and mixed food material abnormal gas screening data are obtained. By combining food material feature point extraction, volatile gas analysis, temperature change monitoring and freshness change prediction, the accuracy of identifying the freshness of the mixed food material is improved.

Description

Technical Field

[0001] The present invention relates to the field of machine learning technology, and in particular to an AI-based cooking robot food freshness recognition method and system. Background Art

[0002] Early food freshness detection mainly relied on manual sensory evaluation and traditional chemical analysis methods, such as smell, appearance and texture. However, these methods are highly subjective and time-consuming, making it difficult to achieve efficient and real-time detection. With the development of image processing technology and machine learning, food freshness recognition began to use computer vision technology. By taking surface images of food with a camera and using image recognition algorithms, the color, shape and surface characteristics of food can be initially judged. At this stage, research focused on convolutional neural networks (CNNs) based on deep learning, which can train the ability to recognize the freshness of food by analyzing large amounts of data. In recent years, deep learning technology in the field of AI has developed rapidly. Combined with the Internet of Things (IoT) and sensor technology, the freshness recognition of food has become more accurate. However, when dealing with mixed ingredients, the existing technologies usually lack accurate identification and freshness evaluation methods for mixed ingredients. At the same time, traditional methods focus on static food freshness judgment and lack dynamic prediction of food freshness changes, resulting in low accuracy of freshness recognition. Summary of the invention

[0003] Based on this, it is necessary to provide an AI-based cooking robot food freshness identification method and system to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a cooking robot food freshness identification method based on AI is provided, the method comprising the following steps: Step S1: Acquire food images; extract food feature points from the food images, and identify food types based on the food feature points to obtain single food identification results and mixed food identification results; identify the freshness of similar food in standard food images based on the single food identification results to obtain single food freshness identification data; Step S2: based on the mixed food identification result, the volatile gas data of the food is collected to obtain the volatile gas data of the mixed food; the volatile gas data of the mixed food is screened for abnormal gas to obtain the abnormal gas screening data of the mixed food; based on the abnormal gas screening data of the mixed food, the temperature change analysis of the abnormal gas volatilization time is performed to generate the temperature change data of the mixed food; the mixed food temperature change data is classified into the storage mode of the mixed food to generate the fresh mode and the corruption mode; Step S3: marking corrupt ingredients on the standard ingredient images based on the corrupt mode to obtain corrupt ingredient marking data; predicting the change of mixed ingredient freshness on the standard ingredient images based on the fresh mode to generate mixed ingredient freshness change prediction data; Step S4: Integrate the mixed ingredient freshness change prediction data and the corrupt ingredient marking data into mixed ingredient freshness identification data, and upload and store the mixed ingredient freshness identification data and the single ingredient freshness identification data to obtain an ingredient freshness identification report.

[0005] The present invention can accurately identify single ingredients and mixed ingredients through the extraction of ingredient feature points and identification of ingredient types in step S1, solving the challenge that traditional technologies are difficult to distinguish mixed ingredients. Based on this judgment, the freshness of single ingredients can be effectively identified through image analysis, while the freshness of mixed ingredients is further analyzed through subsequent steps. Through the collection and abnormal screening of volatile gas data (step S2), combined with the analysis of gas volatilization time and temperature changes, the patent not only relies on visual information, but also introduces odor and physical and chemical characteristics to achieve a more comprehensive judgment of the freshness of ingredients. This multi-dimensional information fusion makes freshness identification more accurate, especially for the dynamic change prediction of mixed ingredients. Through the classification of corruption mode and fresh mode (step S3), combined with temperature change data and freshness change prediction, the patent realizes real-time monitoring of food freshness changes. In particular, during the storage process of mixed ingredients, the freshness change trend of ingredients can be detected in real time, so as to identify potential corruption risks in advance and deal with them in a timely manner. After the corruption ingredients are marked and the freshness change is predicted, the patent integrates all data through step S4 to generate a detailed freshness identification report. The report uniformly stores the freshness identification data of single ingredients and mixed ingredients, providing an important basis for subsequent intelligent decision-making. This process of data integration and report generation improves the overall efficiency of the system, making the ingredient management and cooking process more intelligent and automated. Through the detailed classification of corruption mode and fresh mode, the system can adapt to a variety of ingredient types and different storage conditions, effectively handle the freshness assessment of ingredients under different states, and enhance the flexibility and wide applicability of the system. Therefore, the present invention improves the accuracy of freshness identification of mixed ingredients by combining ingredient feature point extraction, volatile gas analysis, temperature change monitoring and freshness change prediction.

[0006] Preferably, step S1 comprises the following steps: Step S11: using the camera module on the cooking robot to obtain food images; Step S12: performing image preprocessing on the food image to generate a standard food image, wherein the image preprocessing includes image filtering, image brightness enhancement and image geometric transformation; Step S13: extracting food feature points from the standard food image, and matching the food feature points with a preset food resource library to obtain feature point matching data; performing food recognition on the standard food image based on the feature point matching data to obtain food recognition data; Step S14: Using the food identification data, the food type is identified on the standard food images to obtain single food identification results and mixed food identification results; based on the single food identification results, the freshness of the same type of food is identified on the standard food images to obtain single food freshness identification data.

[0007] The present invention can effectively eliminate noise and interference in the image through image preprocessing and feature point matching, and ensure the accuracy of food identification, which is very important for food classification, distinguishing single food and mixed food, especially when applied in kitchen robots, it can significantly improve the reliability of its operation. The freshness recognition of a single food can provide users with accurate food quality feedback through in-depth analysis of feature points and recognition data, help optimize the selection and storage of food, and improve food safety and consumer satisfaction. By distinguishing between single food and mixed food, different cooking needs can be met. For example, single food and mixed food are different in processing methods, which provides more flexibility for the robot to choose cooking methods. The whole process has a high degree of automation, which can greatly reduce the burden of manual operation and improve the intelligence level of kitchen robots. This efficient food recognition and freshness discrimination can support more intelligent cooking tasks, such as automatic classification of food, dynamic adjustment of cooking strategies, etc. This method can be self-adjusted according to the expansion of the food library. For example, as the food library increases, the feature point matching and freshness discrimination modules can continuously optimize the recognition effect and accuracy by updating the database.

[0008] Preferably, the freshness identification of similar ingredients based on the single ingredient identification result by shooting an image of a standard ingredient includes: Extracting images of the same food type as the standard food image from a preset food resource library, marking them as reference food images, and performing image area calibration between the reference food image and the standard food image to obtain an image calibration area; Perform pixel color depth detection on the standard food image to generate pixel color depth detection data; perform adjacent pixel color depth difference calculation on the pixel color depth detection data to obtain adjacent pixel color depth difference calculation data; According to a preset color depth difference threshold, the color depth difference calculation data of adjacent pixel points are confirmed to have a difference range, and color depth abnormal range data is obtained; the image calibration area is marked as an abnormal area using the color depth abnormal range data, and an image calibration abnormal area is obtained; Key visual information is extracted from the abnormal area of ​​image calibration to obtain key visual information extraction data; freshness of standard food images is judged based on the key visual information extraction data to obtain single food freshness identification data.

[0009] The present invention can ensure that the two are compared under the same standard by calibrating the image area of ​​the standard food image and the reference food image. The precise demarcation of the image calibration area provides a reliable basis for the subsequent color depth detection and abnormal area marking, avoiding errors caused by image distortion or deformation. The detection of pixel color depth and the calculation of the color depth difference of adjacent pixels provide more detailed information for the judgment of the freshness of food. This method can capture the slight changes in the color of the food surface, which usually reflect the quality changes of the food such as oxidation and corruption, and enhance the ability to judge the freshness. Through the color depth abnormal range confirmed based on the color depth difference threshold, the abnormal area in the image can be accurately marked, which can not only help distinguish between fresh and stale food, but also accurately determine the changes in certain specific areas on the surface of the food, providing a more comprehensive food quality analysis. By extracting the key visual information in the image calibration abnormal area, the freshness of the food can be further confirmed. Key visual information (such as color spots, shape changes, etc.) is crucial for judging the aging, corruption and other conditions of food, and can provide multi-angle support for freshness judgment. The method can complete efficient food freshness judgment without manual intervention through the automated process of image processing. It can quickly process a large number of food images, adapt to different types of food, and continuously optimize the judgment results through algorithms, thereby improving the intelligence level of the entire cooking process. This method can also adapt well to different food ingredients and different storage environments. Through the diversity of benchmark food images and the universality of color depth detection, freshness can be judged under a variety of cooking conditions to avoid misjudgments caused by changes in food types or environments.

[0010] Preferably, extracting data from key visual information to determine the freshness of standard food photographs includes: By extracting data from key visual information, continuous detection of color difference changes in standard food images is performed to obtain a first food freshness discrimination value; Using key visual information extraction data to perform texture degradation judgment on the standard food image, and obtain a second food freshness judgment value; By extracting key visual information data, edge deformation of the standard food images is judged to obtain the freshness judgment value of the third food; Based on the first food freshness discrimination value, the second food freshness discrimination value and the third food freshness discrimination value, a weighted average calculation is performed on the standard food photographed image to generate single food freshness identification data.

[0011] The present invention can effectively reflect the color change of the food surface over time through the change of color difference. The continuous detection of color difference change can timely detect the aging, oxidation and other changes of food, which is particularly effective for color-sensitive food such as fruits and vegetables. Texture degradation on the surface of food is often an important sign that the food loses its freshness. Blurred, falling or changed texture can be used as an indication of freshness decline. Through texture degradation detection, the freshness problem of food can be identified at an early stage to prevent the use of expired food. The deformation of the edge of food is usually a direct manifestation of decay or deterioration. Especially for leafy vegetables, fish and other easily damaged food, the deformation of the edge can provide crucial clues for freshness determination. The weighted average calculation of multiple freshness indicators such as color difference change, texture degradation and edge deformation can avoid misjudgment caused by a single feature by comprehensively analyzing the changes of different features. For example, some food causes color difference fluctuations due to changes in light or environmental conditions, but the combination of texture and edge deformation can more accurately reflect its actual freshness. This weighted average calculation can effectively balance the importance of different indicators and improve the accuracy and robustness of food freshness identification. By combining multiple visual information features (color difference, texture, edge), the system can cope with the diversity of different types of ingredients and flexibly judge the freshness of various ingredients. For example, for ingredients that are easy to change color, such as tomatoes and peppers, color difference changes are particularly important; and for perishable ingredients, such as fish and meat, the judgment of texture degradation and edge deformation is more critical. This comprehensive judgment method can effectively handle the characteristic differences of different types of ingredients and improve the comprehensiveness and accuracy of freshness judgment. This automatic freshness judgment method based on visual features can not only reduce the error of manual judgment, but also realize high-speed and efficient food freshness analysis, which is especially suitable for large-scale food management and fast cooking processes.

[0012] Preferably, step S2 comprises the following steps: Step S21: Based on the mixed food identification result, the gas module of the cooking robot collects volatile gas data of the food to obtain volatile gas data of the mixed food; Step S22: performing gas component identification on the mixed food volatile gas data to obtain mixed food gas component identification data; performing abnormal gas screening on the mixed food gas data according to the mixed food gas component identification data to obtain mixed food abnormal gas screening data; Step S23: confirming the volatilization time of the abnormal gas based on the abnormal gas screening data of the mixed food, and performing temperature change analysis on the volatilization time of the abnormal gas through the temperature module on the cooking robot to generate temperature change data of the mixed food; Step S24: converting the temperature change data of the mixed food into an infrared thermal map to generate a temperature change thermal map of the mixed food, and classifying the temperature change data of the mixed food into mixed food storage modes through the temperature change thermal map of the mixed food to generate a fresh mode and a spoilage mode.

[0013] The present invention collects gas data volatilized by mixed food through a gas module and performs component analysis on it, which can accurately determine the changes in volatile substances (such as ethanol, ammonia, etc.) in the food. These gas components are usually related to the freshness and corruption degree of the food. Especially for perishable mixed food, the change in gas composition can directly reflect the quality change of the food. By screening the collected gas data, environmental interference and irrelevant gas components can be effectively eliminated, so as to accurately identify abnormal gas components related to the freshness of the food, and further improve the accuracy of food freshness identification. Through the temperature module on the cooking robot, combined with the abnormal gas volatilization time to perform temperature change analysis, the storage environment of the food and its response to temperature changes can be monitored in real time. This analysis helps to identify whether the food is within the appropriate storage temperature range, especially when judging whether the food is corrupted early due to inappropriate temperature changes, it provides an important basis. Over time, the temperature change of the food will affect its freshness and storage state. Through the careful analysis of the temperature change data, abnormal temperature fluctuations can be discovered earlier, and then the corruption process of the food can be predicted and judged. Converting temperature change data into infrared heat maps provides an intuitive way to view the temperature distribution of ingredients during storage. This heat map can help identify temperature abnormality areas more clearly and provide a scientific basis for subsequent food storage and processing. Heat map analysis makes the spatial distribution of temperature changes visual, which is convenient for quick positioning and decision-making. By analyzing the temperature change heat map, the system can automatically classify mixed ingredients into fresh mode or corruption mode. This classification can help the cooking robot automatically identify the current storage status of the ingredients and provide corresponding processing or cooking suggestions. This classification method is simple and efficient, which helps to quickly screen the freshness of ingredients and avoid using corrupt or stale ingredients. Through automated processes such as gas component identification, abnormal gas screening, and temperature change analysis, the system can achieve real-time monitoring and management of the freshness of mixed ingredients without manual intervention. This automated monitoring system can effectively improve work efficiency and reduce the time and error of manual inspection, especially for large-scale smart kitchens or catering management. Based on the temperature change heat map and gas screening results, the system can provide real-time feedback on the freshness status of ingredients and provide decision support for cooking robots. For example, when ingredients enter the corruption mode, the system can automatically adjust the cooking process or prompt the user to handle it in time.

[0014] Preferably, step S23 includes the following steps: Step S231: Analyze the release curve of abnormal gas based on the abnormal gas screening data of mixed food, and generate the abnormal gas release curve; confirm the time point of gas concentration increase on the abnormal gas release curve, and obtain the start time of abnormal gas volatilization; Step S232: extracting the time window of the start time of abnormal gas volatilization, and monitoring the temperature change of the food in the time window through the temperature module on the cooking robot to generate temperature change monitoring data; Step S233: temperature stratify the temperature change monitoring data to obtain food surface temperature data and food internal temperature data; perform gas reaction temperature gradient change analysis on the food surface temperature data and food internal temperature data according to the mixed food abnormal gas screening data to generate food temperature gradient change data; Step S234: performing temperature change rate analysis on the food temperature gradient change data to generate mixed food temperature change data.

[0015] By analyzing the release curve of abnormal gas, the system can accurately identify the early signs of food spoilage. The change in gas concentration and the time point of release are direct manifestations of food spoilage, which provides a clear time starting point for subsequent temperature and freshness analysis. By determining the starting time of abnormal gas volatilization, the system can accurately capture the beginning stage of food spoilage. This early warning helps to take appropriate measures before the food is obviously corrupted, reducing waste and losses. Using the time window of abnormal gas volatilization to monitor the temperature of food can ensure that the temperature data is closely related to the changes in the food spoilage process. Through precise temperature monitoring, the system can capture the impact of temperature fluctuations on the freshness of food and further understand the temperature change characteristics during the corruption process. Through the hierarchical processing of temperature data, the system can obtain the temperature changes on the surface and inside of the food respectively, revealing the different reactions of the internal and surface temperatures of the food, which is crucial to evaluating the freshness of the food, especially for different types of food, the changes in surface and internal temperatures will be different. By analyzing the gas reaction temperature gradient of the surface and internal temperature data of the food, the system can gain a deep understanding of how the internal and external temperature changes of the food interact with the release of corruption gas. This analysis helps reveal the dynamic relationship between temperature and gas volatilization, which helps optimize the storage and handling strategies of food. This temperature gradient change analysis can help the system accurately determine the degree of corruption of the food. For example, the increase in the surface temperature of the food indicates that it is in a state of overheating or improper storage, while the change in the internal temperature of the food indicates that its internal structure is changing, leading to corruption or deterioration. By analyzing the temperature change rate, the system can further reveal the temperature change trend during the food corruption process. Different types of food have different corruption rates. Rate analysis can effectively predict future changes in food and help the system respond in advance. Changes in the temperature change rate provide important information for the storage and use of food. If the temperature changes too quickly, it means that the food is not stored under appropriate conditions; on the contrary, if the temperature changes too slowly, it means that the food is still in a relatively fresh state. Through accurate rate analysis, the system can take appropriate measures before the food is corrupted to avoid food waste. Comprehensive analysis of gas volatilization, temperature changes and gas reaction temperature gradients can fully grasp the quality of food and ensure that food is always in the best condition during storage. Timely detection of abnormal temperature fluctuations and release of spoilage gases can help the system take necessary measures to avoid food safety issues.

[0016] Preferably, step S24 includes the following steps: Step S241: converting the temperature change data of the mixed food into an infrared thermal map, thereby generating a temperature change thermal map of the mixed food; annotating the temperature change thermal map of the mixed food with a hue, thereby obtaining a hue of the temperature change thermal map, wherein the hue of the temperature change thermal map includes blue, green, yellow and red; Step S242: setting a mode classification standard for the mixed food temperature change data by using the temperature change heat map color phase: if the blue-green hue ratio in the mixed food temperature change heat map is greater than 50% and the food temperature gradient change data is less than 1.5°C, and the mixed food temperature change data is less than 0.1°C / min, then a fresh mode discrimination rule is obtained; Step S243: If the yellow-red hue ratio in the mixed food temperature change heat map is greater than 50% and the food temperature gradient change data is greater than 2.0°C, and the mixed food temperature change data is greater than 0.15°C / min, then a corruption mode discrimination rule is obtained; Step S244: Integrate the mixed food storage mode based on the fresh mode discrimination rule and the corruption mode discrimination rule to generate a fresh mode and a corruption mode.

[0017] The present invention can intuitively display the spatial distribution and change trend of the temperature of the food by converting the temperature change data of the mixed food into an infrared heat map. This visualization method makes the temperature change easier to understand, helps users quickly identify the hot and cold areas of the food, and provides a visualization basis for subsequent storage mode determination. Hue labeling (blue, green, yellow, red) of the temperature change heat map helps to further distinguish the different temperature ranges of the food, thereby improving the recognition of the heat map. Each hue represents a different temperature level, helping the system and users to intuitively perceive the temperature state of the food. When the blue-green hue in the mixed food temperature change heat map accounts for more than 50%, and the temperature gradient change is less than 1.5°C and the temperature change rate is less than 0.1°C / min, the system will be judged as "fresh mode", which indicates that the food is in a stable and low temperature change state, indicating that the food is still in a good storage state. When the yellow-red hue in the temperature change heat map accounts for more than 50%, and the temperature gradient changes by more than 2.0°C and the temperature change rate is more than 0.15°C / min, the system will determine it as "corruption mode". This mode indicates that the food has experienced significant temperature fluctuations, indicating that the food is rapidly deteriorating or corrupting and needs to be processed or discarded immediately. Based on the temperature change heat map and pattern classification, the system can automatically monitor the temperature changes of food ingredients and classify them in real time. For the catering industry or home users, this automated monitoring and classification can effectively prevent food spoilage, reduce waste and improve food utilization efficiency. By classifying the temperature changes of mixed ingredients, the system can optimize the storage conditions of the ingredients in real time, such as adjusting the refrigeration temperature, allocating food storage space, etc., thereby extending the freshness of the ingredients and improving storage efficiency.

[0018] Preferably, step S3 comprises the following steps: Step S31: marking corrupted food in the standard food image based on the corrupted pattern to obtain corrupted food marking data; Step S32: extracting freshness change features of standard food images based on the freshness mode to obtain freshness change feature data; dividing the freshness change feature data into data sets to generate a model training set and a model test set; Step S33: training the model training set using a long short-term memory neural network algorithm to generate a food mixture freshness change prediction pre-model; optimizing and iterating the food mixture freshness change prediction pre-model using a model test set to generate a food mixture freshness change prediction model; Step S34: importing the standard food image corresponding to the mixed food identification result into the food mixture freshness change prediction model to perform food mixture freshness change prediction and generate mixed food freshness change prediction data.

[0019] The present invention can quickly identify corrupted ingredients by marking the corruption pattern of food image, and help users automatically remove spoiled ingredients in mixed food processing, which not only saves time for manual screening, but also effectively reduces food safety problems caused by corrupted ingredients. By extracting freshness change feature data from standard food images, the freshness change trend of the food can be accurately captured, which is convenient for subsequent prediction and analysis. Such feature extraction provides important input data for the machine learning model and enhances the accuracy of the prediction. The freshness change feature data is divided into a training set and a test set to ensure the scientific nature of the model training and the representativeness of the data. By using the long short-term memory neural network (LSTM) algorithm, the model can learn the change law of food freshness from historical data, and optimize and iterate on the test set, so as to generate a more accurate food mixed freshness change prediction model. Using the LSTM model to repeatedly train the training set and iteratively optimize the test set can effectively avoid overfitting and underfitting problems, ensure that the model has strong generalization ability and stability, which provides more reliable technical support for the prediction of freshness change of mixed ingredients. By inputting standard food images into the food mix freshness change prediction model, the system can predict the change trend of mixed food in real time and generate new freshness prediction data. This prediction provides users with dynamic monitoring of the status of mixed food, allowing users to adjust food processing and storage methods according to the prediction results. Using the prediction ability of the LSTM model, the system can identify the freshness changes of food in advance and provide suggestions or warnings. For example, when food is about to go bad, the system can notify users in advance to handle it, avoid food waste or deterioration, and improve food utilization and storage time. Traditional food freshness judgment often relies on manual judgment, which is not only time-consuming but also easily affected by human factors. Through automated model prediction, the system can efficiently and accurately analyze the freshness changes of food, reducing manual intervention and the complexity of operation.

[0020] Preferably, step S4 comprises the following steps: Step S41: Integrate the mixed ingredient freshness change prediction data and the corrupt ingredient marking data into mixed ingredient freshness identification data, and upload the mixed ingredient freshness identification data and the single ingredient freshness identification data to the data processing module of the cooking robot for data storage, thereby generating ingredient freshness identification data; Step S42: Visualize the food freshness identification data to obtain a food freshness identification report.

[0021] The present invention integrates the prediction data of the freshness change of mixed ingredients with the marking data of corrupt ingredients, so that the system can comprehensively and accurately identify the overall freshness of mixed ingredients. This comprehensive method of identifying the freshness of ingredients not only considers the trend of freshness change of ingredients, but also integrates the marking of corrupt ingredients, providing users with more accurate judgment basis. By integrating the freshness identification data of mixed ingredients and single ingredients, the system can make a more accurate assessment of the freshness of ingredients. This comprehensive assessment helps users to grasp the quality of ingredients in real time and provides a scientific basis for subsequent processing, use or storage. Uploading the freshness identification data of mixed ingredients and the freshness identification data of single ingredients to the data processing module of the cooking robot for storage can ensure the centralized management of data. This centralized storage not only improves the security of data, but also facilitates users to query, analyze and monitor at any time. Through centralized storage, users can retrieve the freshness data of ingredients at any time as needed. This convenient data management method provides users with a high degree of operational freedom, and is particularly suitable for large-scale kitchens or catering environments that need to monitor the status of ingredients in real time. The integrated freshness identification data of ingredients provides an accurate basis for the quality control of ingredients. Users can rely on this data to quickly identify changes in the freshness of ingredients and take timely measures to handle the ingredients to prevent the quality of ingredients from being affected by improper storage. The system helps users make more scientific decisions during the cooking process through intelligent processing of ingredient freshness identification data. For example, users can decide whether to use a certain ingredient based on the freshness report, or adjust the ingredient storage and use strategy, thereby reducing ingredient waste and ensuring the quality of the meal.

[0022] In this specification, a cooking robot food freshness identification system based on AI is provided, which is used to execute the above-mentioned cooking robot food freshness identification method based on AI. The cooking robot food freshness identification system based on AI includes: The food identification module is used to obtain food images; extract food feature points from the food images, and identify the food types based on the food feature points to obtain single food identification results and mixed food identification results; identify the freshness of the same type of food for standard food images based on the single food identification results to obtain single food freshness identification data; The storage mode analysis module is used to collect volatile gas data of the food based on the mixed food identification result to obtain volatile gas data of the mixed food; screen the volatile gas data of the mixed food for abnormal gas to obtain abnormal gas screening data of the mixed food; perform temperature change analysis on the volatile time of the abnormal gas based on the abnormal gas screening data of the mixed food to generate temperature change data of the mixed food; classify the storage mode of the mixed food based on the temperature change data of the mixed food to generate a fresh mode and a corruption mode; A freshness prediction module is used to mark corrupted ingredients based on the corrupted pattern of standard ingredients and obtain corrupted ingredient marking data; to predict the change of mixed freshness of ingredients based on the fresh pattern of standard ingredients and generate mixed ingredient freshness change prediction data; The data storage module is used to integrate the mixed ingredient freshness change prediction data and the corrupt ingredient marking data into mixed ingredient freshness identification data, and upload and store the mixed ingredient freshness identification data and the single ingredient freshness identification data to obtain an ingredient freshness identification report.

[0023] The beneficial effect of the present invention is that the food identification module can accurately shoot and analyze the food, accurately extract the characteristic points of the food and distinguish the type of food, which not only improves the accuracy of food identification, but also helps to distinguish single food from mixed food, and provides clearer data support for subsequent management. The freshness identification of food is based on the standard food shooting image, combined with the single food identification result, it can effectively identify the freshness of different food and avoid food waste. The storage mode analysis module can deeply understand the corruption process of food by collecting food volatile gas data, screening abnormal gas and analyzing temperature changes. This multi-dimensional analysis not only focuses on the visual appearance changes, but also comprehensively considers the smell and ambient temperature changes to comprehensively evaluate the freshness of food. The temperature change analysis and storage mode classification of mixed food further enhance the prediction ability of the system, can accurately judge the storage state of food, and provide a more scientific basis for subsequent processing. The freshness prediction module can predict the freshness change of food in advance by analyzing the changes of food in corruption mode and fresh mode. This dynamic prediction not only helps users understand the health status of current food, but also provides suggestions based on time and environmental changes to avoid food corruption in advance. The prediction data provides more information for the management of ingredients, thereby improving the efficiency of inventory management and the rationality of ingredient use. The data storage module integrates and stores all key data (such as ingredient freshness identification data, corruption mark data, etc.) to ensure the long-term traceability and accessibility of the data. The ingredient freshness identification report provides users with an intuitive and detailed ingredient health report to help users take appropriate storage and handling measures based on the specific ingredients. Whether it is a home kitchen or commercial ingredient management, the freshness of ingredients can be monitored and managed in real time to avoid ingredient waste and improve ingredient utilization. Therefore, the present invention improves the accuracy of freshness identification of mixed ingredients by combining ingredient feature point extraction, volatile gas analysis, temperature change monitoring and freshness change prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of the steps of a method for identifying the freshness of ingredients of a cooking robot based on AI; Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart; Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG. The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0025] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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 technicians in this field without creative work are within the scope of protection of the present invention.

[0026] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0027] It should be understood that although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit, and the term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0028] To achieve this, please refer to Figures 1 to 3 , a cooking robot food freshness recognition method based on AI, the method comprising the following steps: Step S1: Acquire food images; extract food feature points from the food images, and identify food types based on the food feature points to obtain single food identification results and mixed food identification results; identify the freshness of similar food in standard food images based on the single food identification results to obtain single food freshness identification data; Step S2: based on the mixed food identification result, the volatile gas data of the food is collected to obtain the volatile gas data of the mixed food; the volatile gas data of the mixed food is screened for abnormal gas to obtain the abnormal gas screening data of the mixed food; based on the abnormal gas screening data of the mixed food, the temperature change analysis of the abnormal gas volatilization time is performed to generate the temperature change data of the mixed food; the mixed food temperature change data is classified into the storage mode of the mixed food to generate the fresh mode and the corruption mode; Step S3: marking corrupt ingredients on the standard ingredient images based on the corrupt mode to obtain corrupt ingredient marking data; predicting the change of mixed ingredient freshness on the standard ingredient images based on the fresh mode to generate mixed ingredient freshness change prediction data; Step S4: Integrate the mixed ingredient freshness change prediction data and the corrupt ingredient marking data into mixed ingredient freshness identification data, and upload and store the mixed ingredient freshness identification data and the single ingredient freshness identification data to obtain an ingredient freshness identification report.

[0029] The present invention can accurately identify single ingredients and mixed ingredients through the extraction of ingredient feature points and identification of ingredient types in step S1, solving the challenge that traditional technologies are difficult to distinguish mixed ingredients. Based on this judgment, the freshness of single ingredients can be effectively identified through image analysis, while the freshness of mixed ingredients is further analyzed through subsequent steps. Through the collection and abnormal screening of volatile gas data (step S2), combined with the analysis of gas volatilization time and temperature changes, the patent not only relies on visual information, but also introduces odor and physical and chemical characteristics to achieve a more comprehensive judgment of the freshness of ingredients. This multi-dimensional information fusion makes freshness identification more accurate, especially for the dynamic change prediction of mixed ingredients. Through the classification of corruption mode and fresh mode (step S3), combined with temperature change data and freshness change prediction, the patent realizes real-time monitoring of food freshness changes. In particular, during the storage process of mixed ingredients, the freshness change trend of ingredients can be detected in real time, so as to identify potential corruption risks in advance and deal with them in a timely manner. After the corruption ingredients are marked and the freshness change is predicted, the patent integrates all data through step S4 to generate a detailed freshness identification report. The report uniformly stores the freshness identification data of single ingredients and mixed ingredients, providing an important basis for subsequent intelligent decision-making. This process of data integration and report generation improves the overall efficiency of the system, making the ingredient management and cooking process more intelligent and automated. Through the detailed classification of corruption mode and fresh mode, the system can adapt to a variety of ingredient types and different storage conditions, effectively handle the freshness assessment of ingredients under different states, and enhance the flexibility and wide applicability of the system. Therefore, the present invention improves the accuracy of freshness identification of mixed ingredients by combining ingredient feature point extraction, volatile gas analysis, temperature change monitoring and freshness change prediction.

[0030] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a schematic diagram of a step flow of a method for identifying freshness of ingredients of a cooking robot based on AI according to the present invention. In this example, the method for identifying freshness of ingredients of a cooking robot based on AI includes the following steps: Step S1: Acquire food images; extract food feature points from the food images, and identify food types based on the food feature points to obtain single food identification results and mixed food identification results; identify the freshness of similar food in standard food images based on the single food identification results to obtain single food freshness identification data; In an embodiment of the present invention, a clear image of the food is captured by using a high-resolution camera (such as a digital camera, a mobile phone camera, or an industrial camera). Ensure that the shooting angles are diverse so that detailed features of the food can be obtained from different perspectives. In order to improve the image quality, image enhancement techniques (such as denoising, contrast adjustment, etc.) can be considered. Use computer vision algorithms (such as SIFT, SURF, or ORB) to extract key feature points in the food image. Feature points can be corner points, edges, or textures, etc. The extracted feature points are further described, and feature descriptors are constructed for use in discrimination and classification in subsequent steps. The extracted feature points are trained using machine learning or deep learning models (such as convolutional neural networks CNN) to identify the types of food. This step can divide the food into two categories: single food and mixed food according to its visual features. For a single food, its type (such as apples, carrots, etc.) can be directly identified. For mixed food, it is necessary to combine multiple image areas for discrimination, and even combine other information (such as weight, packaging information, etc.) to assist in judgment. Based on the discrimination results of a single food, select relevant standard food shooting images as a reference. By comparing the visual features of newly photographed ingredients with those of standard ingredients (such as color, texture, surface defects, etc.), image processing algorithms are used to identify their freshness. Common methods for identifying freshness include color analysis (fresh ingredients are usually bright in color, while old ingredients are darker in color), surface gloss, crack detection, etc. These visual features are input into the trained model to evaluate the freshness of the ingredients and obtain the freshness identification data of a single ingredient. Finally, the single ingredient identification results and mixed ingredient identification results, as well as the freshness identification data, are collated and output for subsequent decision-making or operation (such as ingredient classification, inventory management, or quality control).

[0031] Step S2: based on the mixed food identification result, the volatile gas data of the food is collected to obtain the volatile gas data of the mixed food; the volatile gas data of the mixed food is screened for abnormal gas to obtain the abnormal gas screening data of the mixed food; based on the abnormal gas screening data of the mixed food, the temperature change analysis of the abnormal gas volatilization time is performed to generate the temperature change data of the mixed food; the mixed food temperature change data is classified into the storage mode of the mixed food to generate the fresh mode and the corruption mode; In an embodiment of the present invention, the food material that needs to be gas-collected is determined according to the mixed food material discrimination result obtained in step S1. A gas sensor array (such as a gas sensor module, an electronic nose, or other volatile gas detection equipment) is used to collect gas. Common gas sensors include sensors with high sensitivity to volatile organic compounds (VOCs), carbon dioxide (CO2), ammonia (NH3), etc. The volatile gas data of food materials in different time periods are collected to ensure a comprehensive understanding of the changing trend of gas release. The collected gas data is sorted using data preprocessing methods (such as data cleaning, denoising, smoothing, etc.). Anomaly detection algorithms (such as isolation forest, support vector machine (SVM) anomaly detection, k-means clustering, etc.) are applied to filter out abnormal values ​​in the gas data. Abnormal values ​​indicate corruption or other abnormal changes (such as fermentation, oxidation, etc.) of food materials. Relevant characteristic parameters (such as gas concentration change rate, peak concentration, etc.) are extracted to further analyze which gases belong to abnormal gases. The screened abnormal gas data is associated with the temperature data to analyze the relationship between the volatilization time of the abnormal gas and the temperature change. Use temperature sensors (such as thermocouples, infrared thermometers, etc.) to monitor the temperature around or directly on the food. Based on the time series and temperature data of gas volatilization, analyze whether the temperature fluctuation is related to abnormal gas emissions, and use regression analysis or time series analysis to generate temperature change data of mixed food. These data can reveal the impact of temperature changes on gas volatilization and further help evaluate the storage and freshness of food. Based on the temperature change data, use machine learning classification algorithms (such as decision trees, random forests, support vector machines, K-nearest neighbor algorithms, etc.) for pattern recognition to classify the storage mode of food into "fresh mode" and "corruption mode". The fresh mode is usually characterized by small temperature fluctuations, more regular gas volatilization, and no abnormal peak gas concentration. The corruption mode is usually characterized by large temperature fluctuations, unstable gas volatilization, accompanied by abnormal peak gas concentrations, especially related to the release of corruption-related gases such as organic acids, hydrogen sulfide, and ammonia.

[0032] Step S3: marking corrupt ingredients on the standard ingredient images based on the corrupt mode to obtain corrupt ingredient marking data; predicting the change of mixed ingredient freshness on the standard ingredient images based on the fresh mode to generate mixed ingredient freshness change prediction data; In an embodiment of the present invention, the characteristics of the corruption pattern are identified by analyzing the mixed food in step S2. The corruption pattern is usually manifested as changes in the appearance of the food (such as discoloration, deformation, moisture, mold, etc.), odor changes (changes in volatile gases), and abnormal temperature fluctuations. Using the characteristics of the corruption pattern recognition, the corrupt area or features are marked in the standard food image. The corruption features (such as color changes, rotten spots, foreign matter, etc.) can be marked by image processing methods (such as edge detection, region segmentation, target detection, etc.). The marking results are output in the form of data, such as the marked area coordinates, the type of corruption features (such as rot, mold, color change, etc.) and the degree of corruption. The generated corrupt food labeling data can be used as the basic data set for training models or data input for subsequent quality control. The characteristics of the fresh mode are identified by comparing the normal appearance characteristics of the standard food and its color, morphology, surface gloss and other characteristics in the fresh state. Standard food usually has bright color, regular shape, no obvious damage on the surface, etc. Collect image data of standard ingredients at different time points and under different storage conditions. These data should include the freshness status of the ingredients (such as fresh, slightly spoiled, slightly corrupted, etc.) and their corresponding image features. Use deep learning models (such as convolutional neural networks (CNN), long short-term memory networks (LSTM), etc.) to train standard ingredient images and identify the law of freshness changes. By learning the process of ingredients from fresh to corrupt, the model can predict the freshness changes of ingredients at a certain moment in the future or under specific conditions. Based on the trained model, predict the standard ingredient images and generate mixed freshness change data of ingredients. The prediction results should include: freshness prediction of ingredients at future time points (such as freshness level in a few hours or days). Factors affecting freshness changes (such as temperature, humidity, storage environment, etc.).

[0033] Step S4: Integrate the mixed ingredient freshness change prediction data and the corrupt ingredient marking data into mixed ingredient freshness identification data, and upload and store the mixed ingredient freshness identification data and the single ingredient freshness identification data to obtain an ingredient freshness identification report.

[0034] In an embodiment of the present invention, the mixed food freshness change prediction data generated in step S3 is integrated with the corrupt food marking data. The integration method can combine the two data forms together through a data fusion method (such as weighted average, data merging, etc.) to ensure that the overall freshness status of each food or food group is accurately identified. The above two data are fused through an algorithm (such as a weighted average method, a rule engine, etc.) to generate freshness identification data of the mixed food. The data specifically includes: the current state of the mixed food (fresh, slightly corrupt, corrupt, etc.). The freshness score of the mixed food (based on the weighted score of the prediction model and the corruption mark). Identify the time range and change speed of the freshness change of the food. The integrated mixed food freshness identification data and the freshness identification data of the single food are uploaded to the storage system. Before uploading, the data can be compressed and encrypted to ensure data security and save storage space. A cloud storage platform, a database, or a distributed file system can be selected for storage. Common storage methods include NoSQL databases (such as MongoDB, Cassandra), relational databases (such as MySQL, PostgreSQL), or cloud storage (such as AWS S3, Google Cloud Storage). Automatically generate food freshness identification report based on uploaded and stored data.

[0035] Preferably, step S1 comprises the following steps: Step S11: using the camera module on the cooking robot to obtain food images; Step S12: performing image preprocessing on the food image to generate a standard food image, wherein the image preprocessing includes image filtering, image brightness enhancement and image geometric transformation; Step S13: extracting food feature points from the standard food image, and matching the food feature points with a preset food resource library to obtain feature point matching data; performing food recognition on the standard food image based on the feature point matching data to obtain food recognition data; Step S14: Using the food identification data, the food type is identified on the standard food images to obtain single food identification results and mixed food identification results; based on the single food identification results, the freshness of the same type of food is identified on the standard food images to obtain single food freshness identification data.

[0036] In the embodiment of the present invention, by using the high-definition camera module integrated in the cooking robot, the camera should support autofocus and high-resolution imaging. The robot should have the ability to automatically adjust the shooting angle, distance and field of view during the shooting process to ensure that the multi-angle images of the food are clear and correct. The camera module needs to have a stable light perception ability to adapt to the imaging requirements under different ambient light conditions, such as low light, strong light, etc. De-noising techniques such as Gaussian filters and mean filters are applied to remove noise and interference in the image and improve the clarity of the image. De-noising techniques such as Gaussian filters and mean filters are applied to remove noise and interference in the image and improve the clarity of the image. Including rotation, scaling, cropping, etc., to unify the image size and proportion, and ensure that the main features of the food are not distorted. Feature point detection algorithms (such as SIFT, SURF, ORB, etc.) are used to extract key feature points in the food image. The extracted feature points are compared with the preset food resource library, and matching items are found using feature matching algorithms (such as brute force matching, FLANN matching). The matching process can use convolutional neural networks (CNN) in deep learning to further improve the accuracy of feature extraction and matching. Based on the food identification data, multi-classification algorithms (such as support vector machine (SVM), decision tree, random forest, etc.) are used to identify the type of food. If the food image contains multiple ingredients, image segmentation technology can be used to separate different types of food and identify their types separately. For images of mixed ingredients, target detection algorithms in deep learning (such as YOLO and Faster R-CNN) are combined to identify the boundaries of different ingredients and determine their categories. Image analysis technology (such as color analysis and texture analysis) is combined with machine learning models (such as support vector regression SVR or deep neural network) to evaluate the freshness of food. The basis for freshness judgment includes color change (such as yellowing of green leafy vegetables), texture analysis (such as texture change of meat), etc. Statistical modeling is performed on the freshness data of the same type of food, and a food freshness index model is constructed. The actual freshness score of each food is obtained based on the model.

[0037] Preferably, the freshness identification of similar ingredients based on the single ingredient identification result by shooting an image of a standard ingredient includes: Extracting images of the same food type as the standard food image from a preset food resource library, marking them as reference food images, and performing image area calibration between the reference food image and the standard food image to obtain an image calibration area; Perform pixel color depth detection on the standard food image to generate pixel color depth detection data; perform adjacent pixel color depth difference calculation on the pixel color depth detection data to obtain adjacent pixel color depth difference calculation data; According to a preset color depth difference threshold, the color depth difference calculation data of adjacent pixel points are confirmed to have a difference range, and color depth abnormal range data is obtained; the image calibration area is marked as an abnormal area using the color depth abnormal range data, and an image calibration abnormal area is obtained; Key visual information is extracted from the abnormal area of ​​image calibration to obtain key visual information extraction data; freshness of standard food images is judged based on the key visual information extraction data to obtain single food freshness identification data.

[0038] In an embodiment of the present invention, by searching for food images that match the type of images shot by standard food in the food resource library, the image type is ensured to be consistent (such as fruits, vegetables, meat, etc.). The found image is marked as a reference food image as a comparison standard for subsequent image area calibration and freshness assessment. The standard food image is aligned with the reference food image using image registration technology. Common registration methods include feature point-based registration (such as SIFT, ORB) or image content-based registration (such as mutual information method). After calibration, the corresponding area in the image (such as similar parts of the appearance, color, texture, etc. of the food) is extracted to obtain an image calibration area, which will be used as a key analysis area in subsequent color depth detection and freshness assessment. The color depth of each pixel in the image calibration area is detected, and the color value of the pixel (such as RGB value, HSV value, etc.) is calculated. According to the definition of the color space (such as RGB), the brightness or color difference value of each pixel is calculated to generate pixel color depth detection data. Usually, these data will include information such as color difference, brightness value, saturation, etc. of each pixel. The color depth difference of adjacent pixels is calculated by means of adjacent pixel pairs. Common calculation methods include using Euclidean distance to measure the color difference between two pixels (for example, calculating the Euclidean distance of each pixel in RGB space). For each pair of adjacent pixels, the color depth difference is calculated to obtain a set of color depth difference calculation data for subsequent detection of areas with abnormal color depth changes. A threshold range for color depth difference is set based on experience or training data sets. The range can be set based on the freshness changes of food ingredients, such as dullness or changes in color indicating food spoilage. If the color depth difference of adjacent pixels exceeds the threshold, the area has color depth abnormality and is marked as an abnormal area. Through this step, it is possible to accurately determine which areas have abnormal changes and confirm the cause of the change (such as food discoloration or decay) through color depth differences. Based on the obtained color depth abnormality range data, the abnormal areas within the image calibration area are marked. The marking process includes distinguishing areas with abnormal color depth from normal areas. Through this step, areas with decay, discoloration or other quality problems are identified. In the image calibration abnormal area, key visual information is extracted, such as color changes, texture changes or shape changes. The visual features of abnormal areas can be extracted through image analysis techniques (such as edge detection, texture analysis, color space conversion, etc.) for subsequent accurate judgment. Use machine learning models (such as convolutional neural networks (CNN)) to classify and analyze the extracted visual information to extract signs of decay and deterioration. Based on the extracted key visual information, use machine learning or deep learning models (such as support vector machines (SVM), convolutional neural networks (CNN), random forests, etc.) to determine freshness. When training the model, a large amount of labeled food image data (including samples of fresh and spoiled food) can be used for supervised learning to establish a high-precision food freshness recognition model.Ultimately, the model outputs a freshness score or classification result for a single ingredient (such as "fresh", "expired", "about to go bad", etc.).

[0039] Preferably, extracting data from key visual information to determine the freshness of standard food photographs includes: By extracting data from key visual information, continuous detection of color difference changes in standard food images is performed to obtain a first food freshness discrimination value; Using key visual information extraction data to perform texture degradation judgment on the standard food image, and obtain a second food freshness judgment value; By extracting key visual information data, edge deformation of the standard food images is judged to obtain the freshness judgment value of the third food; Based on the first food freshness discrimination value, the second food freshness discrimination value and the third food freshness discrimination value, a weighted average calculation is performed on the standard food photographed image to generate single food freshness identification data.

[0040] In an embodiment of the present invention, the color difference change of each pixel in the image is calculated by using a color space (such as RGB, HSV, etc.). A local color difference measurement method (such as Euclidean distance) can be used to detect the change of the surface color of the food over time or region, and identify the degree of color change. For continuous food image capture, the time evolution of the color difference can be tracked to detect color degradation caused by long-term placement or storage. If the color difference change exceeds the preset threshold, it means that the food has deteriorated, reducing its freshness score, and this detection result generates a first food freshness discrimination value. Texture analysis technology (such as gray level co-occurrence matrix, local binary pattern LBP, Gabor filter, etc.) is used to extract the texture features of the image. Fresh food usually has a delicate and regular texture, while the texture of rotten food will degenerate and become blurred or irregular. By analyzing the detailed changes in the surface texture of the food, the freshness of the food can be effectively identified. For example, the fineness of the surface texture of meat or vegetables will gradually lose over time, and the texture will become rough or disordered. When obvious texture degradation is detected, a second food freshness discrimination value is generated. Generally, degraded texture has a negative impact on freshness judgment. Edge information in the food image is extracted using edge detection algorithms (such as Canny edge detection, Sobel operator, etc.). Fresh food usually has clearer edges, while the edges of rotten or spoiled food are blurred or deformed. Analyze edge deformation, especially for food with signs of spoilage on the surface such as fruits, vegetables or meat. Edge contraction, expansion or blurring indicates that its quality has declined. If the edge deformation exceeds the predetermined threshold, it means that the food has expired or spoiled, and a third food freshness discrimination value is generated. According to the relative importance of each freshness discrimination value, a weight is set for weighted averaging. The specific weight can be determined by experiments or previous data analysis. For example, color difference changes can affect the freshness of some food more than edge deformation, so the weight of color difference changes is higher. For example, assuming that the weight of color difference change is 0.4, the weight of texture degradation is 0.3, and the weight of edge deformation is 0.3, the freshness judgment formula of a single food is: single food freshness = 0.4 × first food freshness + 0.3 × second food freshness + 0.3 × third food freshness. The final freshness score is calculated by weighted averaging. The higher the score, the fresher the food. The final result will generate freshness identification data for a single food, such as "fresh", "slightly spoiled", "severely spoiled", etc., or output a specific numerical score.

[0041] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes: Step S21: Based on the mixed food identification result, the gas module of the cooking robot collects volatile gas data of the food to obtain volatile gas data of the mixed food; Step S22: performing gas component identification on the mixed food volatile gas data to obtain mixed food gas component identification data; performing abnormal gas screening on the mixed food gas data according to the mixed food gas component identification data to obtain mixed food abnormal gas screening data; Step S23: confirming the volatilization time of the abnormal gas based on the abnormal gas screening data of the mixed food, and performing temperature change analysis on the volatilization time of the abnormal gas through the temperature module on the cooking robot to generate temperature change data of the mixed food; Step S24: converting the temperature change data of the mixed food into an infrared thermal map to generate a temperature change thermal map of the mixed food, and classifying the temperature change data of the mixed food into mixed food storage modes through the temperature change thermal map of the mixed food to generate a fresh mode and a spoilage mode.

[0042] In an embodiment of the present invention, the volatile gas of the mixed food during the cooking process is collected in real time by using a gas sensor (such as an electronic nose or a gas sensor array) built into the cooking robot. Commonly used gas sensors can detect a variety of volatile compounds (such as methane, carbon dioxide, ethanol, etc.). The volatile gas data is collected by the gas sensor at different cooking stages (such as heating, stirring, etc.) and stored as a sequence of gas concentration changes over time. These data can reflect the law of gas volatilization of the food during the cooking process. According to the type, state, and cooking process of the food, different time periods are selected for gas data collection to obtain a complete gas volatilization map. The collected gas data is analyzed using a gas chromatograph or a machine learning model to identify the specific components of the volatile gas. Common gas components include water vapor, carbon dioxide, volatile organic compounds (VOCs), ammonia, etc. The gas data is processed using a feature extraction algorithm (such as principal component analysis PCA or independent component analysis ICA) to extract key gas component feature data, which can reflect the differences in chemical components volatilized by different food. The gas component identification data finally output will include the concentration change curve of each component and its fluctuation pattern during the cooking process. According to experimental data or known food gas characteristics, the normal range and abnormal range of gas composition are set. For example, the increased concentration of certain gases (such as hydrogen sulfide, ammonia, etc.) indicates that the food is spoiled or corrupted. According to the gas composition identification data, check which components have concentrations exceeding the predetermined normal range and mark them as abnormal gases. These abnormal gases are closely related to the freshness and quality of the food. The screened abnormal gas data will include the concentration, time point and fluctuation range of the excessive gas, providing a basis for the subsequent temperature change analysis. According to the screened abnormal gas data, determine the time period when the gas concentration exceeds the standard. These time periods correspond to the abnormal gas volatilization of the food, which is usually related to the corruption or deterioration stage of the food. During the time period of abnormal gas volatilization, the temperature change data on the surface of the food is collected using the temperature sensor module (such as infrared temperature probe, thermocouple, etc.) built into the cooking robot. By analyzing the changes in the surface temperature of the food during the abnormal gas volatilization period, the relationship between temperature fluctuations and gas concentration changes is identified, which helps to determine whether the food is accelerated due to excessively high temperatures. The final generated mixed food temperature change data will include a temperature change curve and a temperature fluctuation range corresponding to the abnormal gas volatilization. Convert temperature change data into infrared heat maps, which can be achieved through infrared sensors or by mapping temperature data to the color scale of the heat map. Infrared heat maps can intuitively show the spatial distribution of food surface temperature and its changes over time. Based on the generated temperature change heat map, combined with gas volatilization data and food type, classification algorithms (such as support vector machine SVM, decision tree, neural network, etc.) are used to classify storage modes. Through the learning model, it is determined whether the food is in "fresh mode" or "corruption mode".Based on the heat map analysis results and the classification model, the storage mode of the food is output. Fresh food usually shows a stable temperature distribution, while spoiled food shows uneven or abnormally high temperature. The classification results can be output with labels such as "fresh", "about to spoil", or "spoiled".

[0043] Preferably, step S23 includes the following steps: Step S231: Analyze the release curve of abnormal gas based on the abnormal gas screening data of mixed food, and generate the abnormal gas release curve; confirm the time point of gas concentration increase on the abnormal gas release curve, and obtain the start time of abnormal gas volatilization; Step S232: extracting the time window of the start time of abnormal gas volatilization, and monitoring the temperature change of the food in the time window through the temperature module on the cooking robot to generate temperature change monitoring data; Step S233: temperature stratify the temperature change monitoring data to obtain food surface temperature data and food internal temperature data; perform gas reaction temperature gradient change analysis on the food surface temperature data and food internal temperature data according to the mixed food abnormal gas screening data to generate food temperature gradient change data; Step S234: performing temperature change rate analysis on the food temperature gradient change data to generate mixed food temperature change data.

[0044] In an embodiment of the present invention, a curve of gas concentration changing with time is drawn by using the abnormal gas data obtained by previous screening. This curve represents the release of abnormal gas during the cooking process. Generally, the concentration of abnormal gas will increase as the food starts to deteriorate. Smoothing filtering (such as Gaussian filtering) is used to remove noise in the abnormal gas data and draw the release curve more accurately. The inflection point of the release curve will reflect the critical moment when the food changes. The generated gas release curve contains the relationship between the time axis and the change in gas concentration, which can intuitively display the starting time of food deterioration. By analyzing the abnormal gas release curve, the time point when the gas concentration begins to rise significantly is identified. This time point usually indicates that the deterioration reaction begins to occur inside the food and the gas volatilization begins to accelerate. Using curve fitting or change point detection algorithm (such as CUSUM method or local linear regression), the starting point of the concentration rise is accurately determined and marked as the starting time of abnormal gas volatilization. The determined abnormal gas volatilization starting time will be used as the time window for subsequent analysis of temperature changes. Based on the abnormal gas volatilization starting time determined in step S231, a time window (for example, a few minutes before and after or according to the needs of the cooking process) is set to monitor the temperature changes during this period. The temperature sensor module (such as infrared sensor, temperature probe, etc.) on the cooking robot collects the surface and internal temperature data of the food in the time window in real time. The temperature module can provide high-precision temperature sampling to ensure the accuracy of the data. The output temperature change monitoring data will include the time-varying curves of the surface and internal temperature of the food, providing a basis for further temperature analysis. According to the temperature monitoring data, the temperature data of the surface and internal of the food are extracted in layers. The surface temperature of the food usually changes rapidly, while the internal temperature lags behind. The monitoring data is layered using a data processing algorithm to distinguish between the surface temperature and the internal temperature. This step can use time series analysis or a processing method based on spatial regions to ensure that the data in each region is correctly classified. The final output temperature data will be the temperature change data of the surface and internal of the food, respectively, for subsequent analysis. According to the surface and internal temperature data of the food obtained in step S233, the temperature gradient is calculated. The temperature gradient reflects the temperature difference between the surface and the inside of the food, and this difference will affect the volatilization of the gas. Combined with the abnormal gas data screened in step S232, the relationship between the gas concentration change and the temperature gradient is analyzed. Gas volatilization usually causes the surface temperature to rise. Especially when the food begins to deteriorate, the internal temperature rise lags behind the surface. The generated food temperature gradient change data will show the change of the temperature difference between the surface and the inside of the food over time, and combine the gas data to judge the change trend of the food. Based on the temperature gradient change data, calculate the rate of temperature change. The temperature rate reflects the speed of change of the surface and internal temperature of the food. The change in rate is usually closely related to the deterioration process of the food. Use the difference algorithm or regression analysis method to calculate the temperature change per unit time.A faster rate of temperature change means that the food is rapidly deteriorating or spoiling, and the generated mixed food temperature change data will include the temperature change rate of the food and how the rate affects the freshness judgment of the food.

[0045] Preferably, step S24 includes the following steps: Step S241: converting the temperature change data of the mixed food into an infrared thermal map, thereby generating a temperature change thermal map of the mixed food; annotating the temperature change thermal map of the mixed food with a hue, thereby obtaining a hue of the temperature change thermal map, wherein the hue of the temperature change thermal map includes blue, green, yellow and red; Step S242: setting a mode classification standard for the mixed food temperature change data by using the temperature change heat map color phase: if the blue-green hue ratio in the mixed food temperature change heat map is greater than 50% and the food temperature gradient change data is less than 1.5°C, and the mixed food temperature change data is less than 0.1°C / min, then a fresh mode discrimination rule is obtained; Step S243: If the yellow-red hue ratio in the mixed food temperature change heat map is greater than 50% and the food temperature gradient change data is greater than 2.0°C, and the mixed food temperature change data is greater than 0.15°C / min, then a corruption mode discrimination rule is obtained; Step S244: Integrate the mixed food storage mode based on the fresh mode discrimination rule and the corruption mode discrimination rule to generate a fresh mode and a corruption mode.

[0046] In an embodiment of the present invention, the temperature change data of the mixed ingredients (including the changes of the surface and internal temperature over time) collected by the cooking robot. The temperature data is converted into a heat map format using an infrared thermal imaging algorithm or a heat map generation method based on color mapping. Usually, the area with lower temperature is marked as blue, the area with higher temperature is marked as red, and the temperature zone in the middle is represented by green or yellow. For example, color mapping is implemented using Matplotlib or a similar heat map generation tool. The generated temperature change heat map is hue-labeled to determine the colors corresponding to different temperature ranges: blue: represents the area with lower temperature. Green: represents the area with moderate temperature. Yellow: represents the area with higher temperature but has not yet reached the limit. Red: represents the area with too high temperature, which is usually a sign of deterioration or corruption. This hue labeling can help quickly identify the distribution of ingredients in different temperature areas and their changing trends, and obtain a temperature change heat map with hue labeling for subsequent pattern classification and judgment. First, the proportion of blue and green hue areas in the temperature change heat map is counted. If the proportion of these areas is greater than 50% and the following two conditions are met: The temperature gradient change data of the food (i.e., the temperature difference between the surface and the inside) is less than 1.5°C, which indicates that the temperature change of the food is relatively gentle, and the temperature difference between the surface and the inside is relatively small, which usually means that the food is fresh. The temperature change rate of the mixed food is less than 0.1°C / min, which means that the temperature changes slowly, indicating that the food is in a relatively stable fresh state. If all the above conditions are met, the mixed food is judged to be in the fresh mode, that is, the food is fresh. Count the proportion of yellow and red areas in the heat map. If the proportion of these areas is greater than 50% and the following two conditions are met: The temperature gradient change data of the food is greater than 2.0°C, which indicates that the temperature difference between the surface and the inside is large, indicating that the food is deteriorating rapidly. The temperature change rate of the mixed food is greater than 0.15°C / min, which means that the temperature changes quickly, which usually means that the food is already in the deterioration or corruption stage. If the above conditions are met, the mixed food is judged to be in the corruption mode, that is, the food has begun to corrupt. By comparing the fresh mode discrimination rules and the corruption mode discrimination rules obtained in steps S242 and S243, it is determined whether the food is in a fresh or corrupt state. The temperature change data, temperature gradient data, gas composition data, etc. of the food are comprehensively analyzed, and the storage mode is classified according to the set mode discrimination rules.

[0047] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes: Step S31: marking corrupted food in the standard food image based on the corrupted pattern to obtain corrupted food marking data; Step S32: extracting freshness change features of standard food images based on the freshness mode to obtain freshness change feature data; dividing the freshness change feature data into data sets to generate a model training set and a model test set; Step S33: training the model training set using a long short-term memory neural network algorithm to generate a food mixture freshness change prediction pre-model; optimizing and iterating the food mixture freshness change prediction pre-model using a model test set to generate a food mixture freshness change prediction model; Step S34: importing the standard food image corresponding to the mixed food identification result into the food mixture freshness change prediction model to perform food mixture freshness change prediction and generate mixed food freshness change prediction data.

[0048] In the embodiment of the present invention, the standard food image is marked by the corruption mode discrimination result generated in step S24, and the corruption area or corruption feature is identified. For the corruption area, the standard food image is partially marked based on the characteristics of temperature change, color difference, texture degradation, etc., and the corruption area is marked. The marking data includes the coordinate information, color information, texture information, etc. of the corruption area, and a data set is generated for subsequent analysis. The generated corruption food marking data will be used to calibrate the corruption food area for further analysis and training of the prediction model. Based on the fresh mode discrimination result, the freshness change features such as color change, temperature change, surface texture change, etc. are extracted from the standard food image. The extraction process can be combined with computer vision technology, such as edge detection, texture analysis, color space conversion, etc., to extract key features reflecting the change in food freshness. The extracted feature data is divided into a training set and a test set. Generally, the data set can be used for training at 80% and for testing at 20% to ensure the effectiveness of the training and evaluation process. The data set can be divided by cross-validation or time series segmentation (if the data is time-dependent) to generate a food freshness change feature data set for training and testing for subsequent model training. The model training set is trained using a long short-term memory (LSTM) neural network. LSTM is a deep learning algorithm suitable for processing time series data and can capture the long-term dependency of freshness changes. The training set data is input into the LSTM model, and the model weights are adjusted using the back propagation algorithm so that the model can learn the pattern of freshness changes of ingredients. The trained model is evaluated using the test set, and the performance indicators of the model (such as accuracy, recall, F1 score, etc.) are calculated, and the model is optimized based on the evaluation results. The model is optimized through hyperparameter adjustment (such as learning rate, number of LSTM layers, number of nodes, etc.), regularization, early stopping and other technologies. Through continuous adjustment and optimization, the prediction accuracy of the model is gradually improved. The trained and optimized food mixture freshness change prediction model can be used to predict the future trend of food freshness changes. The standard food image corresponding to the mixed food discrimination result is imported into the trained and optimized food mixture freshness change prediction model. The input data includes the image data of the food and its corresponding freshness feature data (such as temperature change, texture, color difference, etc.). By inputting the feature data of mixed ingredients through the trained LSTM model, the model will predict the future trend of freshness changes of ingredients. The model will output the change data of ingredient freshness, usually including the freshness score, temperature change, corruption index, etc. at a future time point. The generated prediction data of mixed ingredient freshness change can provide users with decision-making basis for ingredient storage, cooking, etc., helping to extend the freshness period of ingredients or avoid ingredient corruption.

[0049] Preferably, step S4 comprises the following steps: Step S41: Integrate the mixed ingredient freshness change prediction data and the corrupt ingredient marking data into mixed ingredient freshness identification data, and upload the mixed ingredient freshness identification data and the single ingredient freshness identification data to the data processing module of the cooking robot for data storage, thereby generating ingredient freshness identification data; Step S42: Visualize the food freshness identification data to obtain a food freshness identification report.

[0050] In the embodiment of the present invention, the mixed food freshness change prediction data obtained by the food mixture freshness change prediction model (step S34) and the corrupt food marking data obtained by the corruption mode marking are integrated. Through data matching, feature fusion and other technologies, a complete set of mixed food freshness identification data is formed, which includes the overall freshness prediction, corruption situation, storage status, etc. of each food. In addition to the freshness identification data of the mixed food, the single food freshness identification data obtained by the single food discrimination result needs to be uploaded. This data comes from step S14 (single food freshness identification data) and contains the discrimination results of each individual food, including freshness status, type and other characteristics (such as color difference, texture change, etc.). The above integrated food freshness identification data (combination of mixed and single food) is uploaded to the data processing module of the cooking robot. This module is responsible for receiving, processing and storing these data for further calculation, analysis and operation. After the data storage is completed, the cooking robot can retrieve these data at any time according to user needs to make real-time decisions or predictions. Use data visualization tools (such as charts, heat maps, 3D graphics, etc.) to display the freshness identification data of the food. Different visualizations can be performed based on the food's corruption, freshness prediction, temperature changes, etc. Based on the above visualization results, a food freshness identification report is generated.

[0051] In this specification, a cooking robot food freshness identification system based on AI is provided, which is used to execute the above-mentioned cooking robot food freshness identification method based on AI. The cooking robot food freshness identification system based on AI includes: The food identification module is used to obtain food images; extract food feature points from the food images, and identify the food types based on the food feature points to obtain single food identification results and mixed food identification results; identify the freshness of the same type of food for standard food images based on the single food identification results to obtain single food freshness identification data; The storage mode analysis module is used to collect volatile gas data of the food based on the mixed food identification result to obtain volatile gas data of the mixed food; screen the volatile gas data of the mixed food for abnormal gas to obtain abnormal gas screening data of the mixed food; perform temperature change analysis on the volatile time of the abnormal gas based on the abnormal gas screening data of the mixed food to generate temperature change data of the mixed food; classify the storage mode of the mixed food based on the temperature change data of the mixed food to generate a fresh mode and a corruption mode; A freshness prediction module is used to mark corrupted ingredients based on the corrupted pattern of standard ingredients and obtain corrupted ingredient marking data; to predict the change of mixed freshness of ingredients based on the fresh pattern of standard ingredients and generate mixed ingredient freshness change prediction data; The data storage module is used to integrate the mixed ingredient freshness change prediction data and the corrupt ingredient marking data into mixed ingredient freshness identification data, and upload and store the mixed ingredient freshness identification data and the single ingredient freshness identification data to obtain an ingredient freshness identification report.

[0052] The beneficial effect of the present invention is that the food identification module can accurately shoot and analyze the food, accurately extract the characteristic points of the food and distinguish the type of food, which not only improves the accuracy of food identification, but also helps to distinguish single food from mixed food, and provides clearer data support for subsequent management. The freshness identification of food is based on the standard food shooting image, combined with the single food identification result, it can effectively identify the freshness of different food and avoid food waste. The storage mode analysis module can deeply understand the corruption process of food by collecting food volatile gas data, screening abnormal gas and analyzing temperature changes. This multi-dimensional analysis not only focuses on the visual appearance changes, but also comprehensively considers the smell and ambient temperature changes to comprehensively evaluate the freshness of food. The temperature change analysis and storage mode classification of mixed food further enhance the prediction ability of the system, can accurately judge the storage state of food, and provide a more scientific basis for subsequent processing. The freshness prediction module can predict the freshness change of food in advance by analyzing the changes of food in corruption mode and fresh mode. This dynamic prediction not only helps users understand the health status of current food, but also provides suggestions based on time and environmental changes to avoid food corruption in advance. The prediction data provides more information for the management of ingredients, thereby improving the efficiency of inventory management and the rationality of ingredient use. The data storage module integrates and stores all key data (such as ingredient freshness identification data, corruption mark data, etc.) to ensure the long-term traceability and accessibility of the data. The ingredient freshness identification report provides users with an intuitive and detailed ingredient health report to help users take appropriate storage and handling measures based on the specific ingredients. Whether it is a home kitchen or commercial ingredient management, the freshness of ingredients can be monitored and managed in real time to avoid ingredient waste and improve ingredient utilization. Therefore, the present invention improves the accuracy of freshness identification of mixed ingredients by combining ingredient feature point extraction, volatile gas analysis, temperature change monitoring and freshness change prediction.

[0053] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0054] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for identifying the freshness of ingredients for a cooking robot based on AI, characterized in that: The following steps are involved: Step S1: obtaining food image; extracting food feature points from the food image, and distinguishing the food type from the food image based on the food feature points, to obtain single food distinguishing results and mixed food distinguishing results; Based on the single food identification result, the freshness of the same type of food is identified by taking images of the standard food, and the single food freshness identification data is obtained; Step S2: based on the mixed food identification result, the volatile gas data of the food is collected to obtain the volatile gas data of the mixed food; the volatile gas data of the mixed food is screened for abnormal gas to obtain the abnormal gas screening data of the mixed food; based on the abnormal gas screening data of the mixed food, the temperature change analysis of the abnormal gas volatilization time is performed to generate the temperature change data of the mixed food; the mixed food temperature change data is classified into the storage mode of the mixed food to generate the fresh mode and the corruption mode; Step S3: marking corrupt ingredients on the standard ingredient images based on the corrupt mode to obtain corrupt ingredient marking data; predicting the change of mixed ingredient freshness on the standard ingredient images based on the fresh mode to generate mixed ingredient freshness change prediction data; Step S4: Integrate the mixed ingredient freshness change prediction data and the corrupt ingredient marking data into mixed ingredient freshness identification data, and upload and store the mixed ingredient freshness identification data and the single ingredient freshness identification data to obtain an ingredient freshness identification report.

2. The AI-based cooking robot food freshness identification method according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: using the camera module on the cooking robot to obtain food images; Step S12: performing image preprocessing on the food image to generate a standard food image, wherein the image preprocessing includes image filtering, image brightness enhancement and image geometric transformation; Step S13: extracting food feature points from the standard food image, and matching the food feature points with a preset food resource library to obtain feature point matching data; performing food recognition on the standard food image based on the feature point matching data to obtain food recognition data; Step S14: Using the food identification data, the food type is identified on the standard food images to obtain single food identification results and mixed food identification results; based on the single food identification results, the freshness of the same type of food is identified on the standard food images to obtain single food freshness identification data.

3. The AI-based cooking robot food freshness identification method according to claim 2 is characterized in that: Based on the single food identification result, the freshness identification of similar food by shooting images of standard food includes: Extracting images of the same food type as the standard food image from a preset food resource library, marking them as reference food images, and performing image area calibration between the reference food image and the standard food image to obtain an image calibration area; Perform pixel color depth detection on the standard food image to generate pixel color depth detection data; perform adjacent pixel color depth difference calculation on the pixel color depth detection data to obtain adjacent pixel color depth difference calculation data; According to a preset color depth difference threshold, the color depth difference calculation data of adjacent pixel points are confirmed to have a difference range, and color depth abnormal range data is obtained; the image calibration area is marked as an abnormal area using the color depth abnormal range data, and an image calibration abnormal area is obtained; Key visual information is extracted from the abnormal area of ​​image calibration to obtain key visual information extraction data; freshness of standard food images is judged based on the key visual information extraction data to obtain single food freshness identification data.

4. The AI-based cooking robot food freshness identification method according to claim 1 is characterized in that: Extracting data from key visual information to determine the freshness of standard food images includes: By extracting data from key visual information, continuous detection of color difference changes in standard food images is performed to obtain a first food freshness discrimination value; Using key visual information extraction data to perform texture degradation judgment on the standard food image, and obtain a second food freshness judgment value; By extracting key visual information data, edge deformation of the standard food images is judged to obtain the freshness judgment value of the third food; Based on the first food freshness discrimination value, the second food freshness discrimination value and the third food freshness discrimination value, a weighted average calculation is performed on the standard food photographed image to generate single food freshness identification data.

5. The AI-based cooking robot food freshness identification method according to claim 1 is characterized in that: Step S2 includes the following steps: Step S21: Based on the mixed food identification result, the gas module of the cooking robot collects volatile gas data of the food to obtain volatile gas data of the mixed food; Step S22: performing gas component identification on the mixed food volatile gas data to obtain mixed food gas component identification data; performing abnormal gas screening on the mixed food gas data according to the mixed food gas component identification data to obtain mixed food abnormal gas screening data; Step S23: confirming the volatilization time of the abnormal gas based on the abnormal gas screening data of the mixed food, and performing temperature change analysis on the volatilization time of the abnormal gas through the temperature module on the cooking robot to generate temperature change data of the mixed food; Step S24: converting the temperature change data of the mixed food into an infrared thermal map to generate a temperature change thermal map of the mixed food, and classifying the temperature change data of the mixed food into mixed food storage modes through the temperature change thermal map of the mixed food to generate a fresh mode and a spoilage mode.

6. The AI-based cooking robot food freshness identification method according to claim 4 is characterized in that: Step S23 includes the following steps: Step S231: analyzing the release curve of abnormal gas based on the abnormal gas screening data of mixed food, generating an abnormal gas release curve; confirming the time point of gas concentration increase on the abnormal gas release curve, and obtaining the start time of abnormal gas volatilization; Step S232: extracting the time window of the start time of abnormal gas volatilization, and monitoring the temperature change of the food in the time window through the temperature module on the cooking robot to generate temperature change monitoring data; Step S233: temperature stratify the temperature change monitoring data to obtain food surface temperature data and food internal temperature data; perform gas reaction temperature gradient change analysis on the food surface temperature data and food internal temperature data according to the mixed food abnormal gas screening data to generate food temperature gradient change data; Step S234: performing temperature change rate analysis on the food temperature gradient change data to generate mixed food temperature change data.

7. The AI-based cooking robot food freshness identification method according to claim 4 is characterized in that: Step S24 includes the following steps: Step S241: converting the temperature change data of the mixed food into an infrared thermal map, thereby generating a temperature change thermal map of the mixed food; annotating the temperature change thermal map of the mixed food with a hue, thereby obtaining a hue of the temperature change thermal map, wherein the hue of the temperature change thermal map includes blue, green, yellow and red; Step S242: setting a mode classification standard for the mixed food temperature change data by using the temperature change heat map color phase: if the blue-green hue ratio in the mixed food temperature change heat map is greater than 50% and the food temperature gradient change data is less than 1.5°C, and the mixed food temperature change data is less than 0.1°C / min, then a fresh mode discrimination rule is obtained; Step S243: If the yellow-red hue ratio in the mixed food temperature change heat map is greater than 50% and the food temperature gradient change data is greater than 2.0°C, and the mixed food temperature change data is greater than 0.15°C / min, then a corruption mode discrimination rule is obtained; Step S244: Integrate the mixed food storage mode based on the fresh mode discrimination rule and the corruption mode discrimination rule to generate a fresh mode and a corruption mode.

8. The AI-based cooking robot food freshness identification method according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: marking corrupted food in the standard food image based on the corrupted pattern to obtain corrupted food marking data; Step S32: extracting freshness change features of standard food images based on the freshness mode to obtain freshness change feature data; dividing the freshness change feature data into data sets to generate a model training set and a model test set; Step S33: training the model training set using a long short-term memory neural network algorithm to generate a food mixture freshness change prediction pre-model; optimizing and iterating the food mixture freshness change prediction pre-model using a model test set to generate a food mixture freshness change prediction model; Step S34: importing the standard food image corresponding to the mixed food identification result into the food mixture freshness change prediction model to perform food mixture freshness change prediction and generate mixed food freshness change prediction data.

9. The AI-based cooking robot food freshness identification method according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: Integrate the mixed ingredient freshness change prediction data and the corrupt ingredient marking data into mixed ingredient freshness identification data, and upload the mixed ingredient freshness identification data and the single ingredient freshness identification data to the data processing module of the cooking robot for data storage, thereby generating ingredient freshness identification data; Step S42: Visualize the food freshness identification data to obtain a food freshness identification report.

10. An AI-based cooking robot food freshness recognition system, characterized in that: Used to execute the AI-based cooking robot food freshness identification method as claimed in claim 1, the AI-based cooking robot food freshness identification system comprises: The food identification module is used to obtain food images; extract food feature points from the food images, and identify the food types based on the food feature points to obtain single food identification results and mixed food identification results; identify the freshness of the same type of food for standard food images based on the single food identification results to obtain single food freshness identification data; The storage mode analysis module is used to collect volatile gas data of the food based on the mixed food identification result to obtain volatile gas data of the mixed food; screen the volatile gas data of the mixed food for abnormal gas to obtain abnormal gas screening data of the mixed food; perform temperature change analysis on the volatile time of the abnormal gas based on the abnormal gas screening data of the mixed food to generate temperature change data of the mixed food; classify the storage mode of the mixed food based on the temperature change data of the mixed food to generate a fresh mode and a corruption mode; A freshness prediction module is used to mark corrupted ingredients based on the corrupted pattern of standard ingredients and obtain corrupted ingredient marking data; to predict the change of mixed freshness of ingredients based on the fresh pattern of standard ingredients and generate mixed ingredient freshness change prediction data; The data storage module is used to integrate the mixed ingredient freshness change prediction data and the corrupt ingredient marking data into mixed ingredient freshness identification data, and upload and store the mixed ingredient freshness identification data and the single ingredient freshness identification data to obtain an ingredient freshness identification report.

Images