Aviation catering quality monitoring method based on artificial intelligence

Abstract

The invention provides an aviation catering quality monitoring method based on artificial intelligence, and the method comprises the steps: obtaining flight delay information, passenger cabin configuration data and a real-time production line task sequence, evaluating the influence of delay on the task sequence and meal heat preservation time in combination with a preset meal standard, and generating a task recombination demand and a heat preservation time extension prediction value; carrying out real-time image acquisition on the meal semi-finished product required by task recombination, carrying out edge detection and texture analysis to obtain a raw material appearance quality index, and identifying a raw material attribute change trend based on the obtained raw material appearance quality index; and carrying out clustering analysis on the attribute change trend of the raw materials, identifying key characteristic parameters influencing the meal quality, and determining the quality risk level of the current production batch by combining the standardization requirement of the passenger cabin meal and the heat preservation time prolonging prediction value, so as to realize accurate control on the quality of the aviation meal.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an airline catering quality monitoring method based on artificial intelligence. Background Art

[0002] As a crucial component of airline service, airline catering directly impacts the passenger experience and airline brand image. Ensuring consistent production quality is crucial. With the rapid development of the aviation industry and rising passenger expectations for food quality, traditional airline catering quality control systems face significant challenges. Currently, airline catering production monitoring relies primarily on manual spot checks and automated equipment with fixed parameters. This approach is rigid and inefficient when adapting to dynamic flight conditions. Especially when flight delays occur, the existing system cannot flexibly adjust production processes and quality control standards, resulting in significant fluctuations in meal quality and an inability to meet the standardized requirements for catering across different aircraft cabin configurations. Production plan changes caused by flight delays necessitate real-time rescheduling of kitchen production lines, which directly impacts the holding time of semi-finished products, causing them to exhibit nonlinear characteristics. These variations in holding time, in turn, can lead to dynamic changes in the properties of meal ingredients, resulting in unpredictable fluctuations in key quality indicators such as taste, color, and nutritional content. This chain reaction makes it difficult for quality control systems based on fixed parameters to accurately assess the actual quality of meals and, therefore, to develop differentiated quality control strategies for meals of varying risk levels. Therefore, how to build an intelligent system that can perceive dynamic changes in flights and adaptively adjust production monitoring parameters to achieve accurate monitoring of meal quality under different aircraft cabin configurations, and dynamically adjust detection parameters and judgment thresholds according to quality risk levels in abnormal situations such as flight delays, has become a key issue in ensuring the consistency of airline catering production quality. Summary of the Invention

[0003] The present invention provides an artificial intelligence-based method for monitoring the quality of airline catering, which mainly includes: Obtain flight delay information, cabin configuration data, and real-time production line task sequencing. Combined with pre-set meal standards, the system assesses the impact of delays on task sequencing and meal holding times, generating predictions for task reorganization needs and extended holding times. Real-time image acquisition of semi-finished meals required for task reorganization is performed, along with edge detection and texture analysis to obtain raw material appearance quality indicators. Based on these indicators, the changing trends of raw material properties are identified. Conduct cluster analysis on the changing trends of raw material properties to identify key characteristic parameters that affect meal quality. Combined with cabin meal standardization requirements and predicted extended holding time, determine the quality risk level of the current production batch. For high-risk meals with a quality risk level greater than the preset level, the image acquisition parameters are adjusted based on the appearance deviation of the high-risk meals and the changing trends of the raw material properties to obtain a new set of detection parameters; the preset level can be determined based on actual conditions; A machine learning algorithm is used to optimize meal testing parameters for the new test parameter set, generating a dynamic threshold table adapted to current production conditions. Real-time test data is then compared with the threshold table to obtain meal quality consistency assessment results, which are used to evaluate whether meal quality meets airline catering standards. If the food quality does not meet airline catering standards, the production line task sequence will be adjusted based on the quality risk level, with high-risk meals being prioritized. The production process of these adjusted high-risk meals will be monitored, and standardized features of the production images will be extracted. Deviation data is extracted based on standardized features, and its correlation with holding time and raw material properties is analyzed. New production control instructions are generated and issued to the production line to adjust the holding time and raw material processing process to obtain food quality consistency data that meets airline catering standards.

[0004] Furthermore, flight delay information, cabin configuration data, and real-time production line task sequencing are obtained. Combined with preset meal standards, the impact of delays on task sequencing and meal holding times is evaluated to generate task reorganization requirements and predicted values for extended holding times. This includes: based on real-time flight delay time data and a set of cabin configuration data, obtaining meal production specification parameters corresponding to the flight from a preset meal specification library using data comparison association rules, and using a flight delay impact assessment matrix to perform correlation calculations between delay time and meal production specification parameters to obtain a dataset of delayed task sequencing changes. Based on the dataset of delayed task sequencing changes and the real-time production line capacity data, airport meal loading time limit parameters are obtained from a preset task window database, and a time window matching algorithm is used to calculate a set of candidate task sequences that meet the loading time limit constraints. Based on the candidate task sequence groups and meal production process data, a process time prediction model is used to calculate the actual execution time of each process, and the corresponding holding time limit parameters are obtained from the process rule library to obtain a dataset of process holding constraints. Based on the process heat preservation constraint dataset, the process execution threshold is obtained from the production line capacity constraint database. A multi-objective optimization algorithm is then used to screen candidate task sequence groups to obtain a set of task sequences that meet the heat preservation constraints. For this set of task sequences that meet the heat preservation constraints, a task time compensation model is used to calculate the heat preservation time extension caused by delays. This is then filtered using the heat preservation threshold judgment rule to obtain a task reorganization sequence and corresponding heat preservation time prediction values. Based on this task reorganization sequence and heat preservation time prediction values, task priority parameters are obtained from a preset task scheduling rule library, and a sequence compensation algorithm is used to generate a task execution schedule that takes heat preservation extension into account.

[0005] Furthermore, real-time image capture of semi-finished meals required for task reorganization is performed, followed by edge detection and texture analysis to obtain raw material appearance quality indicators. Based on the obtained raw material appearance quality indicators, the changing trends of raw material properties are identified. This process involves obtaining a set of delayed flight numbers from a real-time flight status database, reading delay time data and cabin seat distribution data through a flight status monitoring interface, generating a delay time and seat distribution mapping table using a flight data mapper, and extracting task ranking reference values from a preset task rule library. Based on the delay time and seat distribution mapping table, a cabin type distribution calculator is used to generate a seat distribution matrix for delayed flights. A task ranking evaluation function is used to calculate the impact of delays on the task ranking of each flight, generating a task impact score table. Based on the task impact score table, a task priority calculation unit generates priority scores based on flight delay time, number of seats, and cabin class. The corresponding production specifications are read from a meal specification database to generate a task priority sequence table. Based on the task priority sequence table, a meal production rule matcher is used to extract corresponding production line task parameters. A production resource scheduler is used to generate a task timing schedule, resulting in an initial task scheduling plan. According to the initial task scheduling plan, a support vector regression algorithm is used to establish a model for associating delay time with delivery time. A time prediction parameter set is obtained through training with historical delivery data. The task scheduling plan is time-corrected to obtain a meal delivery scheduling schedule.

[0006] Furthermore, cluster analysis is performed on the changing trends of raw material properties to identify key characteristic parameters affecting meal quality. Combined with cabin meal standardization requirements and predicted extended holding time, the quality risk level of the current production batch is determined. This involves collecting images of semi-finished meals from the meal production line based on task reorganization requirements. High-definition cameras positioned around the production line capture images from the top, bottom, left, and right directions. An image preprocessing unit performs Gaussian noise removal and brightness equalization to generate a preprocessed image set. A convolutional neural network is used to extract surface feature vectors from these preprocessed images. Edge detection is performed using the Sobel operator, and texture feature parameters are calculated using a gray-level co-occurrence matrix to generate an image feature description matrix. Based on this image feature description matrix, preset meal quality parameters are retrieved from a quality standard database. A feature similarity calculation unit performs a standard matching operation on this feature description matrix to generate a semi-finished product quality quantitative index. A time window splitter is used to construct a continuous monitoring data sequence for these semi-finished product quality quantitative indexes. A quality baseline calculation unit generates a standard reference interval to obtain a semi-finished product quality benchmark value. According to the semi-finished product quality benchmark value and the continuous monitoring data sequence, a quality change prediction model is established using the random forest algorithm. Real-time data updates are performed through a sliding time window to obtain a semi-finished product attribute change trend graph.

[0007] Furthermore, a cluster analysis is performed on the changing trends of raw material properties to identify key characteristic parameters affecting meal quality. Combined with cabin meal standardization requirements and the predicted extended holding time, the quality risk level of the current production batch is determined. This analysis involves: obtaining raw material color change curves, shape integrity data, and odor concentration values from a raw material attribute collector; collecting raw material surface feature data using a multispectral sensor; and grouping this raw material feature data using a k-means clustering algorithm to generate a raw material feature cluster table. Based on this raw material feature cluster table, cabin meal specification requirements are obtained from a meal standard database. A feature mapping unit establishes a mapping relationship between raw material features and meal specifications. A correlation calculator is used to generate a feature weight matrix to generate a meal quality key feature table. Based on this meal quality key feature table, feature standard thresholds are read from a quality assessment rule library. A feature quantification unit scores the three indicators of color, shape, and odor to generate a raw material quality feature score matrix. Based on this raw material quality feature score matrix and the predicted extended holding time data, a feature combination calculation unit generates a quality impact factor table. A factor weight allocator is used to determine the impact of each factor to generate a quality risk assessment matrix. According to the quality risk assessment matrix, a hierarchical clustering algorithm is used to divide the risk levels, and the quality risks are graded through the risk warning rule library to obtain the production batch quality risk level determination results.

[0008] Furthermore, for high-risk meals with a quality risk level greater than a preset level, image acquisition parameters are adjusted based on the appearance deviations and raw material attribute change trends of the high-risk meals to obtain a new set of detection parameters, including: based on the quality risk level determination result, if the quality risk level is higher than the preset warning value of the preset level, a parameter optimization unit is used to calculate new detection benchmark data. For the detection benchmark data, the appearance feature calculation unit compares the standard sample with the measured image, and a support vector machine algorithm is used to establish a raw material attribute deviation model to obtain a feature deviation quantitative table. Based on the feature deviation quantitative table, the time window calculation module generates an image acquisition frequency sequence, and a feature threshold calibrator is used to set new detection sensitivity parameters to obtain an optimized feature extraction rule set. For the feature extraction rule set, the detection parameter verification module calculates the feature recognition accuracy, and a deep convolutional network is used to train and optimize the feature extraction parameters to obtain a calibrated parameter combination table. Based on the calibrated parameter combination table, a new sampling instruction is issued to the image acquisition device via the task distribution module, and the feature recognition module executes the real-time detection task to obtain a new detection parameter set. For the new detection parameter set, the detection effect score is calculated by the parameter evaluation unit, and the historical detection records are obtained from the quality monitoring database.

[0009] Furthermore, computer vision-based meal appearance deviation data is acquired through a real-time production line monitoring system, raw material attribute change data is extracted from a raw material quality monitoring system, and the meal appearance deviation data and raw material attribute change data are standardized. The exposure parameters of an image acquisition device are dynamically adjusted based on the standardized deviation data. This includes: acquiring real-time images of the meal appearance from a production line monitoring database; identifying areas with abnormal color, shape, and texture on the meal surface using an image feature extractor; and performing area statistics and edge contour extraction on these abnormal areas using a multidimensional feature quantizer to generate an appearance deviation dataset. For this appearance deviation dataset, real-time monitoring records are read from a raw material quality monitoring database. An attribute analysis unit quantifies and calculates the degree of meat fiber breakage, starch texture softening ratio, water loss value, oil oxidation content, and residual spice intensity to generate a raw material quality variation table. Based on this raw material quality variation table, a numerical normalization processor performs interval mapping conversion on each indicator, and a range normalization method is used to generate a normalized data sequence to generate a standardized quality indicator set. A quality prediction model is established using a support vector regression algorithm for this standardized quality indicator set and the appearance deviation dataset. A deviation quantization unit calculates a comprehensive deviation score to generate a deviation grade assessment table. Based on the deviation level assessment table, an image acquisition parameter calculator generates exposure intensity adjustment values, contrast adjustment values, and color saturation adjustment values. These values are then converted into device control instructions using a parameter mapping unit to generate an image acquisition parameter update plan. For this image acquisition parameter update plan, historical adjustment records are retrieved from a device parameter database and evaluated for effectiveness using a parameter verification unit to generate an optimized image acquisition parameter set.

[0010] Furthermore, a machine learning algorithm is used to optimize meal inspection parameters using the new set of inspection parameters. A dynamic threshold table adapted to current production conditions is generated. Real-time inspection data is then compared with the threshold table to obtain meal quality consistency assessment results, which are used to evaluate whether meal quality meets airline catering standards. This process involves: Based on the new set of inspection parameters, a parameter feature extractor is used to extract shape, color, and texture feature data of the inspection parameters. A random forest algorithm is used to calculate feature weight distributions. Temperature, humidity, and light intensity parameters are retrieved from the production environment database to obtain an optimized inspection parameter matrix. Based on this optimized inspection parameter matrix, a threshold generation unit is used to construct a sequence of inspection thresholds. This sequence is dynamically updated using a time window sliding method. Qualified sample parameters are retrieved from a historical database to obtain a baseline threshold table. Based on this baseline threshold table, the thresholds are temperature-corrected and humidity-calibrated using an environmental compensator. A parameter adaptor is used to generate dynamic threshold intervals to obtain a real-time inspection threshold set. Based on this real-time inspection threshold set, meal inspection data is acquired through a data collector. A multi-dimensional comparator is used to calculate the degree of match between the inspection values and the thresholds to obtain a quality feature score table. Based on the quality characteristic scoring table, a score normalization processor generates a standardized score, and a support vector machine algorithm is used to establish a quality classification model to obtain a quality grading result. Based on the quality grading result, judgment rules are obtained from the quality standard library, and a rule matcher is used to generate a compliance assessment report to obtain the food quality judgment result.

[0011] Furthermore, if the food quality does not meet airline catering standards, the production line task sequence is adjusted based on the quality risk level, prioritizing high-risk meals. The production process of these adjusted high-risk meals is monitored, and standardized features are extracted from the production images. This process involves: calculating the quality risk value of each batch using a risk scoring unit based on the quality consistency assessment results, generating a task priority sequence using a priority calculator, and extracting a list of pending tasks from the production task database to obtain a high-risk task priority table. Based on this high-risk task priority table, a task timing planner generates a new production task sequence, and a capacity allocator allocates tasks based on the production line's processing capacity to obtain an optimized task execution plan. Based on this optimized task execution plan, an image acquisition device captures a sequence of production process images. A feature extractor acquires color distribution, shape contours, and texture structure data from the images to obtain a raw feature dataset. A data normalization processor maps the feature data to a standard interval, and a feature selector selects key feature indicators to obtain a standardized feature vector. Based on this standardized feature vector, a convolutional neural network extracts deep feature representations, and a feature aligner calculates the degree of match with the quality standard template to obtain a feature matching score table. For the feature matching score table, a multi-dimensional quality score is generated by a quality assessor, and a monitoring report is generated by a comprehensive determiner to obtain a production process quality feature set.

[0012] Furthermore, deviation data is extracted based on standardized features, and the correlation between the deviation data and holding time and raw material properties is analyzed. New production control instructions are generated and issued to the production line to adjust the holding time and raw material processing process to obtain food quality consistency data that meets airline catering standards. This includes: Based on the standardized feature data, a feature calculator extracts deviation values for color uniformity, shape integrity, and texture uniformity. A correlation analyzer is used to establish a correspondence between the deviation values and changes in holding time and raw material properties to generate a quality deviation mapping table. Based on this quality deviation mapping table, a neural network algorithm is used to establish temperature, time, and raw material impact models. A parameter optimizer is used to generate temperature control values, holding time values, and raw material processing values to obtain a production parameter adjustment plan. Based on this production parameter adjustment plan, a parameter converter generates equipment control instructions. A command checker verifies the compliance of the control instructions to obtain a new production control instruction set. Based on this standardized control instruction set, a task decomposer generates a sequence of equipment execution tasks, and a task scheduler is used to assign the execution tasks to obtain a production line execution plan. According to the production line implementation plan, a parameter collector collects temperature curves, holding time records, and raw material status data. A parameter validator performs threshold comparisons to generate a process monitoring dataset. A quality detector extracts the color, shape, and texture characteristics of the food from this process monitoring dataset. A random forest algorithm is used to classify and evaluate these features, generating a quality consistency evaluation table.

[0013] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: The present invention discloses an artificial intelligence-based method for monitoring the quality of airline catering. By acquiring flight delay information, cabin configuration data, and production line task sequencing, the method evaluates the impact of delays on tasks and holding time, and generates task reorganization requirements and holding time predictions. Computer vision technology is used to capture images of semi-finished meals in real time, and image features are analyzed in conjunction with a machine learning model to identify trends in raw material attribute changes. Cluster analysis is performed based on attribute change trends to determine quality risk levels and dynamically adjust detection parameters. Detection thresholds are optimized using a machine learning algorithm, and a dynamic threshold table is generated to assess meal quality consistency. For meals that do not meet standards, the production line task sequencing is adjusted, and secondary monitoring is performed. The correlation between deviation data and holding time and raw material attributes is analyzed, production parameters are optimized, and the holding time and raw material processing process are adjusted, ultimately achieving precise control and consistency of airline catering quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The present invention is a flowchart of an artificial intelligence-based airline catering quality monitoring method.

[0015] Figure 2The figure is a schematic diagram of an artificial intelligence-based airline catering quality monitoring method of the present invention. DETAILED DESCRIPTION

[0016] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0017] Airline catering quality monitoring involves real-time oversight of the meal production process to ensure that meals meet preset quality standards. The catering production system consists of a production line, monitoring equipment, and a data processing module. The production line is responsible for meal preparation, while the monitoring equipment collects production data. The data processing module analyzes the data and generates control instructions. Meal data is divided into raw material data and production data. Raw material data includes attributes such as color and taste, while production data includes parameters such as holding time and process duration. Flight delay information includes delay time and flight number, while cabin configuration data includes seat distribution and meal grade. Monitoring methods include manual sampling and automated testing. Traditional automated testing relies on fixed parameters and is difficult to adapt to dynamic changes.

[0018] like Figure 1 In this embodiment, an AI-based airline catering quality monitoring method may specifically include: S101 obtains flight delay information, cabin configuration data, and real-time production line task sequencing. Combined with preset meal standards, it evaluates the impact of delays on task sequencing and meal holding time, and generates predictions for task reorganization needs and extended holding time.

[0019] In an embodiment of the present invention, the catering production system has a built-in monitoring function, which can be implemented by system software or a dedicated application. The specific form is determined according to the actual scenario. The embodiment of the present invention does not impose too many restrictions on the startup method of the monitoring function. It can be triggered by manual operation by the user or by sending instructions through an external device. For example, when the monitoring function is integrated into the system software, the user can start the function through the production management interface or quickly activate it through a shortcut icon. When the function is embedded in an application, the developer can set the startup method according to needs, such as setting a monitoring button in the production scheduling interface, which the user can start by clicking.

[0020] S1011 obtains delay information and cabin configuration data based on the real-time flight status interface, combines it with the preset meal standard library, and generates a delayed task sorting change dataset through correlation analysis.

[0021] In this embodiment of the present invention, information such as delay time, flight number, and cabin seat distribution is obtained through the flight status interface. Corresponding meal preparation parameters, such as hold time and process requirements, are extracted from a pre-set meal standard library. A data association algorithm is used to match the delay time with meal parameters to generate a task order change dataset. For example, if a flight is delayed by 90 minutes and the hold time for business class meals is 120 minutes, the calculation indicates that the task needs to be adjusted in advance to meet the loading deadline.

[0022] S1012 uses the task sorting change dataset and the real-time production capacity of the production line to obtain the loading time limit parameters from the preset task window database and generates a candidate task sequence group through the time matching algorithm.

[0023] In this embodiment of the present invention, airport catering loading time constraints are extracted from the task window database based on the task sequence change dataset and production line capacity data. A time matching algorithm is then used to calculate candidate task sequences that meet these time constraints. For example, if an airport requires loading to be completed 90 minutes before takeoff, the task sequence can be adjusted based on the delayed takeoff time to ensure that production tasks meet the time constraint.

[0024] S1013 For the candidate task sequence group, a process prediction model is used to obtain the process execution threshold from the capacity constraint database to generate a process insulation constraint data set.

[0025] In this embodiment of the present invention, a preset process prediction model analyzes candidate task sequences and, combined with meal preparation process data, calculates the actual duration of each process. Process execution thresholds are obtained from a capacity constraint database to generate a heat preservation constraint dataset. For example, if business class meals take 45 minutes to prepare, the task completion time is determined based on the heat preservation time limit.

[0026] S1014: Based on the heat preservation constraint dataset, an optimization algorithm is used to screen candidate task sequence groups, generating task sequences that meet the heat preservation constraints and heat preservation time predictions. In this embodiment of the present invention, based on the process heat preservation constraint dataset, process execution thresholds are obtained from the production line capacity constraint database. A multi-objective optimization algorithm is then used to screen candidate task sequence groups, generating task sequences that meet the heat preservation constraints and corresponding heat preservation time predictions. For example, if business class meals are scheduled to begin production at 9:50 AM, the heat preservation time is extended to 135 minutes. The task time compensation model calculates the extended heat preservation time caused by delays and generates a task execution schedule.

[0027] S1015 obtains priority parameters from the task scheduling rule library according to the task sequence and the insulation time prediction value, and generates a task execution timing table through a sequence optimization algorithm.

[0028] In this embodiment, priority parameters are extracted from a task scheduling rule library based on the task sequence that meets the thermal insulation constraint, and a sequence optimization algorithm is used to generate a task execution schedule. For example, business class meals are prioritized for production, scheduled to be completed between 9:50 and 10:45, leaving time for emergency response.

[0029] In this embodiment of the present invention, flight delays are closely related to meal retention times. Through real-time data analysis, task sequencing is dynamically adjusted to ensure consistent meal quality. For example, if a flight is delayed by 90 minutes, task sequencing is optimized based on the retention times for business and economy class passengers to complete meal loading. Through intelligent task reorganization, this embodiment of the present invention effectively reduces the impact of delays on meal quality, improves production efficiency, ensures consistent in-flight meal quality, and provides passengers with a high-quality dining experience.

[0030] S102 collects real-time images of semi-finished meals required for task reorganization, performs edge detection and texture analysis, obtains raw material appearance quality indicators, and identifies raw material attribute change trends based on the obtained raw material appearance quality indicators.

[0031] In this embodiment of the present invention, high-definition cameras deployed on the production line capture images of semi-finished meals to address task reconfiguration requirements. Preprocessing techniques are then used to optimize image quality. Key features are extracted using a deep learning model to generate quantitative quality metrics. Furthermore, time series analysis is used to predict attribute trends. This approach enables real-time monitoring of semi-finished product quality, promptly identifying potential anomalies and ensuring meal production meets pre-set standards, providing precise assurance for airline catering quality.

[0032] like Figure 2 S1021 collects images of semi-finished meals from multiple angles through a high-definition camera device fixed on the production line, and uses an image preprocessing module to eliminate noise and equalize brightness to generate a standardized preprocessed image set.

[0033] In an embodiment of the present invention, high-resolution cameras installed around the production line are used to capture images of semi-finished meals from four directions: top, bottom, left, and right. The cameras use a 12-megapixel sensor to ensure clear image details. To eliminate the effects of uneven ambient lighting or equipment vibration, Gaussian filtering is used to remove random noise, and a brightness equalization algorithm is applied to adjust the image brightness to a standard range, such as a grayscale range of 180 to 220, to generate a pre-processed image set suitable for subsequent analysis. The embodiment of the present invention does not impose excessive restrictions on the deployment method of the camera device, and it can be adjusted according to the layout of the production line.

[0034] S1022 uses a convolutional neural network to extract surface feature vectors from preprocessed image sets, and generates an image feature description matrix through edge detection and texture analysis algorithms.

[0035] In an embodiment of the present invention, a convolutional neural network model is used to extract features from a preprocessed image set to generate a feature vector containing the surface shape, color, and arrangement regularity of the food. To further analyze the image details, the Sobel operator is used to perform edge detection and calculate the edge integrity score. For example, the edge score of rice grains in semi-finished rice products must be above 0.85. At the same time, the texture features are calculated using the grayscale co-occurrence matrix, including contrast, correlation, and entropy, to form an 8×8 feature description matrix. Contrast reflects the difference in brightness between pixels, correlation characterizes the regularity of arrangement, and entropy indicates the complexity of the texture. These parameters together constitute the feature description matrix, which provides a data basis for quality assessment.

[0036] S1023 obtains meal standard parameters from a preset quality standard database based on the feature description matrix, and calculates the quantitative quality indicators of the semi-finished product through a feature matching algorithm.

[0037] In this embodiment of the present invention, preset food quality parameters are extracted from a quality standard database. For example, semi-finished rice products must meet requirements such as a shape integrity rate of no less than 95% and a color uniformity of no less than 0.9. A cosine similarity algorithm is used to match the feature description matrix with the standard parameters to generate a quantitative indicator of the semi-finished product's quality. If the similarity score is less than 0.8, it indicates a quality deviation in the semi-finished product, and the degree of deviation in the quantitative indicator is recorded. This embodiment of the present invention does not impose excessive restrictions on the specific implementation of the matching algorithm, and technical personnel can optimize it based on actual scenarios.

[0038] S1024 combines the quantitative indicators of semi-finished product quality and uses time series analysis technology to construct a continuous monitoring data sequence to generate a quality benchmark interval.

[0039] In this embodiment of the present invention, a time window segmentation module is used to construct a continuous monitoring data series of quantitative indicators of semi-finished product quality at 5-minute intervals. By calculating the mean and standard deviation of the indicators over the past hour, a quality benchmark interval is generated. For example, the standard interval for rice color uniformity is 0.85 to 0.95. When the indicators for three consecutive time windows fall below the lower limit of the interval, a risk of quality fluctuation is determined. This embodiment of the present invention ensures real-time quality monitoring through time series analysis.

[0040] S1025 uses the random forest algorithm to build a quality change prediction model based on the continuous monitoring data sequence, updates the data through a sliding time window, and generates a semi-finished product attribute change trend chart.

[0041] In an embodiment of the present invention, a quality change prediction model is trained based on a random forest algorithm in combination with historical data, and input variables include raw material batches, processing time, and ambient temperature. The model updates data in real time through a 60-minute sliding time window to predict the changing trend of quality indicators within the next 30 minutes. For example, if the color uniformity of rice decreases by more than 15%, it is determined to be a quality abnormality. The generated trend graph uses time as the horizontal axis and the standardized quality score as the vertical axis to intuitively display the dynamic changes of attributes such as color and shape, which is convenient for production management personnel to monitor the status of semi-finished products in real time. The embodiment of the present invention effectively warns of quality risks through a predictive model and improves the stability of meal production.

[0042] In an embodiment of the present invention, the quality monitoring of semi-finished meals involves multi-dimensional feature analysis. Taking semi-finished rice as an example, four high-definition cameras capture images from different angles to ensure that the complete shape and color distribution of the rice grains are captured. The pre-processed images are subjected to feature extraction using a convolutional neural network, combined with edge detection and texture analysis to generate a high-precision feature description matrix. The quality standard database provides clear quantitative indicators, calculates the degree of deviation through feature matching, and generates a monitoring data sequence in real time. The random forest model predicts quality trends based on time series data and promptly detects anomalies, such as color changes caused by extended holding time. This multi-level analysis method can accurately identify the causes of quality fluctuations, ensure that meals meet airline catering standards, and improve production efficiency and passenger satisfaction.

[0043] S103 performs cluster analysis on the changing trends of raw material properties to identify key characteristic parameters that affect meal quality. Combined with the standardization requirements for cabin meals and the predicted value of extended holding time, the quality risk level of the current production batch is determined.

[0044] In this embodiment, a multispectral sensor is used to collect characteristic data based on the dynamic changes in raw material properties. A clustering algorithm is then used to analyze key indicators such as color, shape, and odor to generate a quality quantification matrix. This matrix, combined with cabin meal standards and predicted holding time, is then used to assess the quality risk level of production batches. This approach accurately identifies the causes of quality fluctuations and dynamically assigns risk levels, providing a basis for production optimization and ensuring the stability and consistency of airline catering quality.

[0045] S1031 collects raw material surface feature data through a multispectral sensor, groups the feature data using a k-means clustering algorithm, and generates a raw material feature classification table.

[0046] In an embodiment of the present invention, a multispectral sensor is used to collect surface data of raw materials, covering a wavelength range of 400 to 1000 nanometers, to capture features such as color and texture. For example, the color data of beef meals is analyzed using dominant wavelength analysis. Fresh beef has a dominant wavelength of approximately 650 nanometers, while spoiled beef has a wavelength shifted below 580 nanometers. Using a k-means clustering algorithm, the feature data is classified into three categories: Category A indicates high quality, Category B indicates average quality, and Category C indicates substandard quality. This generates a feature classification table. This embodiment of the present invention uses cluster analysis to perform preliminary screening of raw material quality, providing data support for subsequent evaluation.

[0047] S1032 extracts cabin meal specifications from the meal standards database based on the feature classification table, establishes a correspondence between raw material features and meal requirements through feature mapping technology, and generates a key feature weight table.

[0048] In this embodiment of the present invention, cabin meal specifications are obtained from a meal standards database. For example, first class requires Category A ingredients, business class accepts Category A or Category B, and economy class uses Category B. A feature mapping algorithm is used to establish associations between color, shape, and aroma and meal specifications. A correlation analysis is used to assign weights to each feature, such as a weight of 0.4 for color, 0.3 for shape, and 0.3 for aroma, generating a table of key feature weights. This embodiment of the present invention does not impose any specific restrictions on the specific implementation of the mapping algorithm; technical personnel may adjust it based on actual needs.

[0049] S1033 obtains standard thresholds from the quality assessment rule library for the key feature weight table, scores the color, shape, and smell indicators through quantitative analysis, and generates a raw material feature score matrix.

[0050] In this embodiment of the present invention, standard thresholds are extracted from a quality assessment rule library. For example, beef color uniformity must be above 0.9, shape integrity must be above 0.92, and the concentration of volatile odors must be below a preset value. A quantitative analysis module then assigns a percentage score to color, shape, and odor, generating an 8×8 feature score matrix. The color score includes sub-items for freshness, uniformity, and glossiness; the shape score includes integrity, regularity, and thickness uniformity; and the odor score includes odor detection values and volatile concentration. This multi-dimensional scoring ensures a comprehensive assessment.

[0051] S1034 combines the characteristic score matrix with the holding time prediction value, generates a quality impact factor table through risk factor analysis, determines the weight of each factor, and generates a quality risk assessment matrix.

[0052] In an embodiment of the present invention, a quality impact factor table is generated through a factor analysis module based on the feature score matrix and the predicted holding time. If the holding time is extended beyond 90 minutes, the odor factor weight is increased to 0.4, and the shape weight is reduced to 0.2. Based on historical data, the impact of each factor at different holding times is calculated to generate a quality risk assessment matrix. This embodiment of the present invention adapts to real-time changes in production conditions through dynamic weight adjustment, thereby improving the accuracy of risk assessment.

[0053] S1035 uses a hierarchical clustering algorithm to divide risk levels based on the quality risk assessment matrix, and combines it with the risk warning rule library to generate the production batch quality risk level determination results.

[0054] In an embodiment of the present invention, a hierarchical clustering algorithm is used to divide quality risks into four levels: green, yellow, orange, and red. Green indicates quality safety, yellow indicates minor risk, orange indicates medium risk, and red indicates major risk. Judgment criteria are extracted from the risk warning rule library. For example, when the color score is less than 75 points and the odor score is less than 80 points, it is judged as an orange risk. The generated risk level results directly guide subsequent quality control measures. For example, the production parameters of orange risk batches need to be adjusted first. The embodiment of the present invention ensures that high-risk batches are handled in a timely manner through refined risk classification.

[0055] In an embodiment of the present invention, multi-dimensional analysis of raw material properties is the core of airline catering quality control. Taking beef meals as an example, color changes are captured using a multispectral sensor, combined with an electronic nose sensor to detect volatile odors, ensuring the comprehensiveness of feature data. The combination of k-means clustering and hierarchical clustering algorithms enables progressive analysis from feature grouping to risk classification. The introduction of cabin meal standards ensures that assessment results match actual needs. A dynamic weight adjustment mechanism optimizes the impact of factors based on changes in holding time, generating an accurate risk assessment matrix. This method effectively identifies the causes of quality anomalies, such as odor changes caused by extended holding times, triggering timely production adjustments to ensure that meals meet airline service standards and enhance the passenger experience.

[0056] S104 For high-risk meals with a quality risk level greater than a preset level, the image acquisition parameters are adjusted according to the appearance deviation of the high-risk meals and the changing trend of the raw material properties to obtain a new set of detection parameters; the preset level can be determined based on actual conditions.

[0057] In this embodiment, for high-risk batches, multi-dimensional analysis and machine learning algorithms dynamically optimize detection parameters based on appearance deviations and changes in raw material properties, improving the sensitivity and accuracy of anomaly detection. This approach can adapt to changes in the production environment in real time, promptly identify quality issues, and guide process adjustments, thereby ensuring the stability of airline catering and passenger satisfaction.

[0058] S1041 If the quality risk level exceeds the preset level, the preset parameter optimization method is used to calculate new test benchmark data.

[0059] In an embodiment of the present invention, the quality risk level of the real-time image is obtained. If the quality risk level exceeds the preset level, the appearance integrity, color uniformity and shape regularity are analyzed through the risk scoring module. For example, taking salad making as an example, the standard lettuce leaf integrity must reach above 0.85, the color uniformity must reach 0.8, and the shape regularity must reach 0.75. If the detection value is lower than the standard, the risk score exceeds the warning threshold of 0.8, triggering the parameter optimization process. The parameter optimization method calculates new detection benchmark data based on the current detection data and historical standard samples. The preset level can be determined according to the actual situation.

[0060] S1042 uses the support vector machine algorithm to compare standard samples with measured images based on the detection benchmark data, generates a quantitative table of feature deviations, and optimizes image acquisition frequency and detection sensitivity through time series analysis.

[0061] In an embodiment of the present invention, a support vector machine algorithm is used to extract edge contours and color distribution features, compare the measured images with standard samples, and generate a quantitative table of feature deviations. For example, the edge defect rate of lettuce leaves is 15%, the browning area accounts for 12%, and the cutting size error is 8 mm. Through the time window calculation module, an image acquisition frequency sequence is generated at intervals of 60 seconds, and the frequency is increased from 180 seconds / time to 60 seconds / time. At the same time, the feature threshold calibration module adjusts the edge detection threshold to 0.8, the color segmentation threshold to 0.85, and the shape matching threshold to 0.9 to ensure that the detection sensitivity is adapted to high-risk scenarios.

[0062] S1043 extracts parameter features based on the feature deviation quantitative table, generates a calibrated parameter combination table, and sends it to the image acquisition device for real-time detection.

[0063] In an embodiment of the present invention, a deep convolutional neural network is used to train a feature extraction model based on 5,000 historical samples, including normal, minor defect and serious defect samples. After training, the model recognition accuracy reaches 97%, the missed detection rate is reduced to 2%, and the false detection rate is reduced to 1%. The generated parameter combination table includes exposure parameters, focus parameters, etc. For example, the exposure intensity is adjusted to 200-220, and the clarity is not less than 0.95. The task distribution module sends the parameters to 4 high-definition cameras, and the feature recognition module detects changes in lettuce leaves in real time. The embodiment of the present invention optimizes parameters through deep learning to improve detection accuracy.

[0064] S1044 uses multidimensional feature quantification and support vector regression algorithms to generate a standardized quality index set and deviation level evaluation table to guide further adjustment of image acquisition parameters based on changes in raw material properties.

[0065] In an embodiment of the present invention, the fiber breakage rate, starch texture softening ratio, water loss value, oil oxidation content and spice residual intensity of the beef meal are analyzed by a multidimensional feature quantification module. For example, the fiber breakage rate is 25%, the water loss is 8%, and the oil oxidation is 0.35. The numerical normalization module maps the indicators to the range of 0-1 and uses the range normalization method to generate a standardized quality indicator set. The support vector regression algorithm trains a quality prediction model based on historical data, and the deviation quantification module calculates the comprehensive deviation score. If the score reaches 0.82, it is judged to be high risk, and the exposure intensity is adjusted to 220, the contrast to 1.5, and the color saturation to 1.3.

[0066] In an embodiment of the present invention, the quality monitoring of airline catering needs to cope with complex production scenarios. Taking beef meals as an example, surface features are captured by high-resolution cameras to detect color, shape, and texture anomalies. When the area of the abnormal area exceeds 10% or the edge defect exceeds 15%, it is recorded as an appearance deviation. Analysis of raw material properties shows that the starch texture softening ratio has increased to 0.45 and the residual intensity of spices has decreased to 65%. The support vector regression model predicts quality trends and combines deep convolutional networks to optimize detection parameters, significantly improving the abnormality detection rate. After parameter adjustment, the detection records of the past 7 days show that the abnormality detection rate has increased by 25% and the accuracy has increased by 8%. The test results are fed back to the preprocessing link to guide the adjustment of the cutting tool sharpness and pre-cooling temperature to ensure the stability of meal quality and meet the standardized requirements of different cabins.

[0067] S105 uses a machine learning algorithm to optimize meal testing parameters for the new test parameter set, generates a dynamic threshold table adapted to current production conditions, compares real-time test data with the threshold table, and obtains meal quality consistency assessment results. These consistency assessment results are used to evaluate whether the meal quality meets airline catering standards.

[0068] In this embodiment of the present invention, multi-dimensional feature extraction and environmental calibration techniques, combined with machine learning algorithms, optimize detection thresholds for highly dynamic production scenarios, ensuring accurate and adaptable test results. This approach can respond in real time to environmental changes and raw material variations, dynamically adjusting test standards to effectively identify quality anomalies, improving the efficiency and reliability of airline catering quality control and providing passengers with a dining experience that meets their expectations.

[0069] S1051 obtains the shape, color, and texture feature data of the detection parameters through the feature extraction module. Combined with the production environment parameters, it uses the random forest algorithm to generate feature weight distribution and construct the detection parameter optimization matrix.

[0070] In an embodiment of the present invention, high-resolution image processing technology is used to extract shape, color, and texture features from food samples through a feature extraction module. Taking Western steak as an example, shape features include edge integrity, area regularity, and thickness uniformity; color features cover color uniformity, browning degree, and carbonized area ratio; and texture features involve texture directionality, surface roughness, and crack density. Temperature, humidity, and light intensity data are obtained from the production environment database, such as a temperature of 25 degrees Celsius, a humidity of 75%, and a light of 750 lux. A random forest algorithm analyzes feature importance and generates a weight distribution, such as a weight of 0.35 for color uniformity, 0.25 for edge integrity, and 0.15 for texture directionality. The weight distribution is combined with environmental parameters to generate a multi-dimensional detection parameter optimization matrix, providing a data basis for threshold adjustment.

[0071] S1052 constructs a detection threshold sequence based on the detection parameter optimization matrix using a threshold generation algorithm, and dynamically updates it using a time window sliding method to generate a benchmark threshold table.

[0072] In this embodiment of the present invention, a threshold generation algorithm is used to construct a threshold sequence containing multiple detection parameters, such as 36 feature thresholds, based on an optimization matrix. A sliding time window method uses a 30-minute cycle, updating threshold data every 5 minutes. This method, combined with historical qualified sample parameters, ensures that the threshold sequence reflects the latest production status. For example, the color uniformity threshold for standard steaks ranges from 0.85 to 0.95, and the sequence range is dynamically adjusted based on recent test data. The resulting baseline threshold table provides a reference for environmental calibration.

[0073] S1053 uses an environmental compensation algorithm to calibrate temperature and humidity for the benchmark threshold table, and combines it with an adaptive adjustment mechanism to generate dynamic threshold intervals to build a real-time detection threshold set.

[0074] In an embodiment of the present invention, an environmental compensation algorithm is used to adjust the threshold according to changes in the production environment. For example, for every 1 degree Celsius increase in temperature, the texture feature threshold is lowered by 4% to adapt to the softening of the meat; for every 1% increase in humidity, the color threshold is increased by 2% to offset the influence of surface moisture. If the light intensity is lower than 800 lux, the image contrast threshold is increased by 3%. The adaptive adjustment mechanism generates a dynamic threshold interval based on environmental trends. For example, the color uniformity interval is adjusted to 0.82 to 0.92. The real-time detection threshold set ensures that the detection parameters match the production conditions. The embodiment of the present invention improves the robustness of detection through environmental calibration.

[0075] S1054 obtains meal detection data through the data acquisition module based on the real-time detection threshold set, uses a multi-dimensional matching algorithm to calculate the matching degree between the detection value and the threshold, generates a quality feature score table, and divides the quality level through the classification model.

[0076] In an embodiment of the present invention, real-time meal detection data is obtained through a data acquisition module, covering shape, color and texture features. The cosine similarity algorithm is used to calculate the matching degree between the detection value and the threshold set, and a quality feature score table is generated. For example, the color uniformity score of the steak sample is 0.87, the edge integrity is 0.90, and the texture directionality is 0.84. The score normalization process maps the score to the range of 0-1 and inputs it into the support vector machine classification model. The model is trained based on historical data and divides the quality into three levels: excellent, qualified, and unqualified. First-class steak requires that all feature scores are higher than 0.9 for excellent and lower than 0.8 for unqualified. The embodiment of the present invention achieves accurate quality assessment through multi-dimensional matching and classification.

[0077] In an embodiment of the present invention, dynamic threshold optimization for airline catering is key to ensuring consistent quality. For example, for steak, multidimensional features are extracted from high-resolution images and weighted using a random forest algorithm to adapt to changes in the production environment. An environmental compensation mechanism dynamically adjusts thresholds based on temperature and humidity. For example, increased humidity results in an increase in the color threshold, ensuring detection sensitivity. A support vector machine model categorizes quality based on standardized scores. When color uniformity drops below 0.75, the meal is deemed unqualified and the reason is recorded. Test results are fed back to production to guide adjustments to processing temperatures or tool parameters, optimizing the process. This dynamic, intelligent detection solution effectively addresses quality fluctuations, ensures that meals meet the stringent standards of different cabins, and enhances the overall quality of airline service.

[0078] S106 If the quality of the food does not meet the airline catering standards, the production line task sequence will be adjusted according to the quality risk level, high-risk meals will be processed first, the production process of the adjusted high-risk meals will be monitored, and standardized features of the production images will be extracted.

[0079] In this embodiment of the present invention, for meal batches that fail to meet quality standards, production plans are dynamically adjusted using risk assessment and task optimization algorithms. High-precision image analysis technology is then combined with real-time monitoring of the production process to generate a multi-dimensional quality feature set. This approach enables rapid response to quality anomalies, optimized resource allocation, and precise identification of production deviations, thereby ensuring consistent quality for airline catering and improving production efficiency and passenger satisfaction.

[0080] S1061 Based on the results of food quality consistency assessment, the quality risk value of each batch is calculated through the risk analysis module, and the priority allocation algorithm is used to generate the task priority sequence to construct a high-risk task priority table.

[0081] In an embodiment of the present invention, key indicators of each batch are extracted from the food quality consistency assessment data, and the quality risk value is calculated by the risk analysis module. Taking the fruit platter as an example, if the freshness is lower than 0.85, the cutting regularity is lower than 0.8, or the color uniformity is lower than 0.75, the risk value exceeds the warning threshold of 0.8. The priority allocation algorithm is sorted according to the risk value, and the tasks to be processed are extracted from the production task database to generate a priority table for high-risk meal tasks. For example, 300 first-class, 500 business-class, and 1,200 economy-class platters are marked as high-risk and need to be handled as a priority. The embodiment of the present invention ensures that high-risk batches are handled in a timely manner through risk quantification.

[0082] S1062 uses the task scheduling algorithm to generate the production task sequence for the high-risk meal task priority table, and combines the production line capacity allocation tasks to generate an optimized task execution plan.

[0083] In an embodiment of the present invention, a task scheduling algorithm is used to generate a new production task sequence based on the production line's processing capacity and the urgency of the task. For example, the first-class platter task is scheduled to start at 9:00, the business class at 10:30, and the economy class at 13:00. The capacity allocation module allocates the high-precision first-class tasks to production line 1 and the remaining tasks to production lines 2 and 3 based on the characteristics of the production line to ensure maximum resource utilization. The optimized task execution plan can effectively shorten the processing cycle of high-risk batches. The embodiment of the present invention improves production flexibility through task optimization.

[0084] S1063 According to the task execution plan, high-frequency image acquisition equipment is used to obtain image sequences of the food production process, generate standardized feature vectors, and use deep learning models to perform feature analysis.

[0085] In an embodiment of the present invention, a high-definition image acquisition device is deployed on the production line to capture an image of the production process every 30 seconds to generate a continuous image sequence. The feature extraction module extracts color distribution, shape contour and texture structure data from the image, such as the color saturation of the fruit platter, the degree of browning of the edges, the consistency of the cut size and the clarity of the flesh texture. The data normalization process maps the eigenvalues to the range of 0-1, and the feature selection algorithm selects 6 key indicators, such as the color saturation of apple cuts must reach 0.85, and the consistency of the cuts must reach 0.9. The standardized feature vector is input into the convolutional neural network to extract a 32-dimensional deep feature representation. The features are analyzed by the deep learning model to improve the robustness of the feature extraction. The embodiment of the present invention ensures the traceability of the production process through high-frequency monitoring.

[0086] S1064 uses a feature matching algorithm to calculate the similarity with the quality standard template for deep feature representation, and generates a multi-dimensional quality score and a production process quality feature set.

[0087] In an embodiment of the present invention, a feature matching algorithm is used to compare the deep feature representation with the quality standard template through the cosine similarity method to generate a feature matching score table. If the matching score of the fruit platter batch is lower than 0.8, the cause of the deviation is analyzed from the three aspects of color, shape, and freshness, such as recording whether the browning of the edge of the apple slices exceeds the standard or the size of the slices is uneven. The quality assessment module generates a multi-dimensional score, and the comprehensive judgment module outputs a detailed monitoring report, which includes the scores of various indicators and descriptions of deviations. The production process quality feature set records complete feature data to support subsequent process optimization and quality traceability. The embodiment of the present invention improves the accuracy of quality monitoring through multi-dimensional analysis.

[0088] In an embodiment of the present invention, high-risk batch management for airline catering requires efficient task adjustment and real-time monitoring. Taking fruit platters as an example, risk analysis is used to quickly identify batches with abnormal quality, prioritize production schedules, and ensure that first-class tasks are completed first. Production line allocation takes into account differences in production capacity, and delicate tasks are assigned to high-precision equipment to improve processing quality. High-frequency image acquisition is combined with convolutional neural networks to extract key features in real time. For example, if the browning area of a batch of apple slices exceeds the standard, the deviation will be recorded and process adjustment suggestions will be triggered, such as optimizing the frequency of tool cleaning or adjusting the pre-cooling time. Monitoring reports provide detailed data support to production management personnel, guide process improvements, ensure that meals meet the strict standards of different cabins, and enhance the quality assurance capabilities of airline services.

[0089] S107 extracts deviation data based on standardized features, analyzes the correlation between the deviation data and holding time and raw material properties, generates new production control instructions and issues them to the production line, adjusts the holding time and raw material processing process, and obtains food quality consistency data that meets airline catering standards.

[0090] In this embodiment of the present invention, multi-dimensional feature analysis and machine learning algorithms are used to accurately identify the causes of food quality deviations and optimize production parameters. Through real-time monitoring and instructional adjustments, food quality is ensured to meet airline catering standards. This approach effectively addresses the complex dynamics of the production process, improving quality control accuracy and production efficiency, and providing airline passengers with a high-quality dining experience.

[0091] S1071 uses the feature analysis module to extract the deviation values of color uniformity, shape integrity, and texture uniformity of meal samples, and uses the association analysis algorithm to construct a mapping relationship between the deviation and the insulation time and raw material properties to generate a quality deviation mapping table.

[0092] In an embodiment of the present invention, a feature analysis module is used to extract key quality indicators from standardized feature data. Taking Western-style steak as an example, the color uniformity deviation shows that the browning area accounts for about 15%, the shape integrity deviation shows that the edge defect rate is about 12%, and the texture uniformity deviation shows that the hardness fluctuation exceeds the standard deviation by 0.3. The association analysis algorithm quantifies the relationship between the deviation and the holding time and the raw material properties through regression analysis. For example, for every 30 minutes of extension of the holding time, the browning area increases by about 5%, and the looseness of the raw material fiber increases by about 8%. The generated quality deviation mapping table clearly shows the relationship between each deviation indicator and the production conditions, providing a data basis for parameter optimization. The embodiment of the present invention ensures the accurate identification of the cause of the deviation through association analysis.

[0093] S1072 uses a neural network algorithm to build a temperature, time, and raw material impact model based on the quality deviation mapping table, and uses a parameter optimization algorithm to generate new production control parameters and formulate a production adjustment plan.

[0094] In an embodiment of the present invention, a three-dimensional influence model is constructed based on historical data and mapping tables through a deep neural network algorithm: the temperature influence model shows that for every 5 degrees Celsius increase in the holding temperature, the browning rate accelerates by about 20%; the time influence model shows that a holding time of more than 120 minutes will cause meat fiber denaturation; the raw material influence model points out that insufficient pickling time increases water loss by about 10%. The parameter optimization algorithm uses a gradient descent method to iteratively calculate the optimal parameter combination and generate new control parameters. For example, the holding temperature is lowered to 65 degrees Celsius, the holding time is controlled at 100 minutes, and the pickling time is extended to 4 hours. The production adjustment plan clearly defines the target value of each parameter and provides guidance for equipment control. The embodiment of the present invention improves the scientific nature of parameter optimization through multi-model analysis.

[0095] S1073 converts the optimization parameters into equipment control instructions through the instruction generation module according to the production adjustment plan, and uses a verification mechanism to ensure the compliance of the instructions and generate a standardized control instruction set.

[0096] In an embodiment of the present invention, the instruction generation module converts the optimization parameters into specific equipment instructions, such as adjusting the heating power of the temperature control equipment to 65%, the ventilation frequency of the insulated box to 6 times per hour, and the raw material pretreatment time to 240 minutes. The verification mechanism ensures that the temperature control accuracy is within plus or minus 1 degree Celsius and the time error is less than 5 minutes by comparing the instructions with the equipment operating range. If the instruction exceeds the range, it is automatically corrected to the compliant value. The generated standardized control instruction set ensures the executability and security of the instructions. The embodiment of the present invention reduces execution risk through strict verification.

[0097] S1074 generates equipment execution task sequences based on the standardized control instruction set through the task allocation module, and uses real-time monitoring technology to obtain production process data and generate process monitoring data sets.

[0098] In an embodiment of the present invention, the task allocation module breaks down instructions into tasks such as temperature regulation, timing control, and raw material processing. The task scheduling algorithm assigns tasks sequentially, ensuring that the insulation program is initiated after raw material processing is completed. The parameter acquisition module records temperature changes, insulation time, and raw material status data every 5 minutes, generating a temperature curve and status log. For example, the temperature curve shows that the insulation process is stable at 65 degrees Celsius, and the raw material status sensor records that the fiber looseness has decreased to 5%. The parameter verification module compares this data with preset thresholds to generate a process monitoring data set to support quality assessment. This embodiment of the present invention ensures the controllability of the production process through real-time monitoring.

[0099] S1075 uses the process monitoring dataset to extract meal characteristics through the quality analysis module, and uses the random forest algorithm for classification and evaluation to generate a meal quality consistency evaluation table that meets airline catering standards.

[0100] In an embodiment of the present invention, the quality analysis module extracts color, shape and texture characteristics from the monitoring data. For example, the browning area of the steak is reduced to 8%, the edge integrity is increased to 0.95, and the hardness standard deviation is reduced to 0.15. The random forest algorithm classifies the feature data based on 100 decision trees and divides them into three levels: excellent, qualified, and unqualified. Through feature importance analysis, the algorithm confirms that color uniformity has the greatest impact on quality, with a weight of approximately 0.4. The evaluation results show that after optimization, the qualified rate of the batch increased from 85% to 95%, and the excellent rate increased from 40% to 60%. The food quality consistency evaluation form that meets the airline catering standards records the scores and classification results of each indicator, providing a reference for process improvement. The embodiment of the present invention achieves refined quality management through classification evaluation.

[0101] In an embodiment of the present invention, quality optimization for airline catering relies on a closed-loop mechanism of deviation analysis and parameter adjustment. For example, using steak as an example, correlation analysis identifies browning issues caused by excessive holding time. A neural network model further quantifies the impact of temperature and time, guiding parameter optimization. Optimized instructions undergo rigorous verification and task allocation to ensure efficient production line execution. Real-time monitoring data demonstrates precise temperature control, improved raw material condition, and significantly enhanced quality characteristics. Evaluation forms provide production managers with detailed data, such as noting that a batch has been upgraded to an excellent grade due to reduced browning, guiding adjustments to the marinating process or holding equipment settings. This intelligent, data-driven optimization solution effectively reduces quality fluctuations, ensures that meals meet the high standards required by different cabin classes, and enhances the overall competitiveness of airline services.

[0102] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An artificial intelligence-based method for monitoring the quality of airline catering, characterized in that: The method comprises: Obtain flight delay information, cabin configuration data, and real-time production line task sequencing. Combined with preset meal standards, evaluate the impact of delays on task sequencing and meal insulation time, and generate task reorganization requirements and insulation time extension prediction values. Perform real-time image acquisition on semi-finished meals with task reorganization requirements, perform edge detection and texture analysis, obtain raw material appearance quality indicators, and identify raw material attribute change trends based on the obtained raw material appearance quality indicators. Perform cluster analysis on raw material attribute change trends to identify key characteristic parameters that affect meal quality. Combined with cabin meal standardization requirements and insulation time extension prediction values, determine the quality risk level of the current production batch. For high-risk meals with a quality risk level greater than the preset level, adjust the image acquisition parameters based on the appearance deviation of the high-risk meals and the trend of raw material attribute changes. A new set of detection parameters is obtained; a machine learning algorithm is used to optimize the meal detection parameters for the new set of detection parameters, a dynamic threshold table adapted to current production conditions is generated, and real-time detection data is compared with the threshold table to obtain a meal quality consistency assessment result, which is used to evaluate whether the meal quality meets the airline catering standards; if the meal quality does not meet the airline catering standards, the production line task sequence is adjusted according to the quality risk level, high-risk meals are processed first, the production process of the adjusted high-risk meals is monitored, and standardized features of the production images are extracted; deviation data is extracted based on the standardized features, the correlation between the deviation data and the holding time and raw material properties is analyzed, new production control instructions are generated and issued to the production line, the holding time and raw material processing process are adjusted, and meal quality consistency data that meets the airline catering standards is obtained.

2. The method for monitoring airline catering quality based on artificial intelligence according to claim 1, characterized in that: The acquisition of flight delay information, cabin configuration data, and real-time production line task sequencing, combined with preset meal standards, evaluates the impact of delays on task sequencing and meal holding time, and generates task reorganization requirements and holding time extension predictions, including: Based on the flight delay time data and cabin configuration data set, the corresponding meal production specification parameters of the flight are obtained from the preset meal specification library through data comparison association rules to obtain the delay task sorting change dataset; For the delayed task sorting change dataset, the airport catering loading time limit parameters are obtained from the preset task window database through the time window matching algorithm to obtain a candidate task sequence group that meets the loading time limit constraint; For the candidate task sequence group, the process execution threshold is obtained from the production line capacity constraint database to obtain the process insulation constraint data set; According to the process insulation constraint data set, candidate task sequence groups are screened to obtain task reorganization requirements and insulation time extension prediction values.

3. The method according to claim 1, characterized in that The method further includes: obtaining flight delay data and cabin configuration information, calculating the impact coefficient of the delay on the production line task sorting, if the task sorting impact coefficient is greater than a target value, adjusting the task priority according to the production line task sorting and preset meal standards, and using a linear regression algorithm to predict the meal delivery scheduling time based on the adjusted task priority.

4. The method for monitoring airline catering quality based on artificial intelligence according to claim 1, characterized in that: The real-time image acquisition of the semi-finished meal required for task reorganization, edge detection and texture analysis to obtain raw material appearance quality indicators, and identification of raw material attribute change trends based on the obtained raw material appearance quality indicators, include: Collect images of semi-finished food in four directions: up, down, left, and right. Perform Gaussian noise removal and brightness equalization to obtain a pre-processed image group. Extracting surface feature vectors of the food according to the preprocessed image group, and obtaining a feature description matrix by calculating the Sobel operator and the gray-level co-occurrence matrix; For the feature description matrix, preset food quality parameters are obtained from the quality standard database, and a standard matching operation is performed through a feature similarity calculation unit to obtain a quantitative index of semi-finished product quality; Based on the continuous monitoring data sequence constructed by the semi-finished product quality quantitative indicators and the time window splitter, a quality change prediction model is established using the random forest algorithm. Real-time data updates are performed through a sliding time window to obtain the trend of semi-finished product attribute changes.

5. The method for monitoring airline catering quality based on artificial intelligence according to claim 1, characterized in that: The cluster analysis of the changing trends of raw material properties identifies key characteristic parameters that affect meal quality. Combined with the standardization requirements for cabin meals and the predicted value of extended holding time, the quality risk level of the current production batch is determined, including: Collecting raw material surface feature data, grouping the raw material surface feature data, and obtaining a raw material feature clustering table; Based on the raw material feature clustering table, cabin meal specification requirements are obtained from the meal standard database, and a mapping relationship between raw material features and meal specifications is established through a feature mapping unit to obtain a meal quality key feature table; According to the food quality key feature table, feature standard thresholds are read from the quality assessment rule library, and the color, shape, and smell indicators are scored using a feature quantification unit to obtain a raw material quality feature score matrix; A hierarchical clustering algorithm is used to divide the risk levels into the raw material quality feature score matrix, and the preset risk level threshold is read from the risk warning rule library to obtain the quality risk level of the current production batch.

6. The method for monitoring airline catering quality based on artificial intelligence according to claim 1, characterized in that: For high-risk meals with a quality risk level greater than a preset level, the image acquisition parameters are adjusted according to the appearance deviation of the high-risk meals and the change trend of the raw material properties to obtain a new detection parameter set, including: The risk scoring unit obtains test data of appearance integrity, color uniformity, and shape regularity. If the risk score of the test data is higher than a preset warning value, new test benchmark data is calculated; Based on the detection benchmark data, a support vector machine algorithm is used to compare the standard sample with the measured image to obtain a quantitative table of feature deviations; Processing the feature deviation quantitative table through a time window calculation module to generate an image acquisition frequency sequence, setting detection sensitivity parameters, and obtaining a feature extraction rule set; With respect to the feature extraction rule set, the feature recognition module performs a real-time detection task to obtain a new detection parameter set.

7. The method according to claim 1, characterized in that The method also includes: obtaining food appearance deviation data detected by computer vision through a real-time production line monitoring system, extracting raw material attribute change data from a raw material quality monitoring system, standardizing the food appearance deviation data and the raw material attribute change data, and dynamically adjusting exposure parameters of an image acquisition device based on the standardized deviation data.

8. The method for monitoring airline catering quality based on artificial intelligence according to claim 1, characterized in that: The new set of test parameters is optimized using a machine learning algorithm to generate a dynamic threshold table adapted to current production conditions. Real-time test data is compared with the threshold table to obtain a meal quality consistency assessment result. The consistency assessment result is used to evaluate whether the meal quality meets the airline catering standards, including: Obtain the shape feature data, color feature data, and texture feature data of the detection parameters, use the random forest algorithm to calculate the weight distribution of each feature data, and obtain the detection parameter optimization matrix; For the detection parameter optimization matrix, a detection threshold sequence is constructed by a threshold generation unit, and the detection threshold sequence is updated by a time window sliding method to obtain a reference threshold table; According to the reference threshold table, the threshold is corrected for temperature and calibrated for humidity through an environmental compensator, and a parameter adaptor is used to generate a dynamic threshold interval to obtain a real-time detection threshold set; For the real-time detection threshold set, meal detection data is obtained through a data collector, and a multi-dimensional comparator is used to calculate the matching degree between the detection data and the real-time detection threshold set to obtain a meal quality consistency assessment result.

9. The method for monitoring airline catering quality based on artificial intelligence according to claim 1, characterized in that: If the food quality does not meet the airline catering standards, the production line task sequence will be adjusted according to the quality risk level, with high-risk meals being processed first. The production process of the adjusted high-risk meals will be monitored, and standardized features of the production images will be extracted, including: Calculate the quality risk value of each batch through the risk scoring unit according to the quality consistency assessment data, and use the priority calculator to generate a task priority sequence based on the quality risk value to obtain a high-risk task priority table; A task timing planner is used to generate a production task sequence for the high-risk task priority table, and a capacity allocator is used to allocate the production task sequence according to the processing capacity of the production line to obtain a task execution plan; According to the task execution plan, an image acquisition device is used to acquire a production process image sequence, and a feature extractor is used to obtain color distribution, shape contour and texture structure data from the image sequence to obtain a standardized feature vector; A deep feature representation is extracted from the standardized feature vector through a convolutional neural network to obtain standardized features of the production image.

10. The method for monitoring airline catering quality based on artificial intelligence according to claim 1, characterized in that: The method extracts deviation data based on standardized features, analyzes the correlation between the deviation data and holding time and raw material properties, generates new production control instructions and issues them to the production line, adjusts the holding time and raw material processing flow, and obtains food quality consistency data that meets airline catering standards, including: Obtain the color uniformity, shape integrity, and texture uniformity deviation values of meal samples, and use the correlation analyzer to construct a quality deviation mapping table; For the quality deviation mapping table, a temperature influence model and a time influence model are established by a neural network algorithm, and a parameter optimizer is used to generate a temperature control value and a holding time value according to the influence model; Generate an equipment control instruction according to the temperature control value and the holding time value, use an instruction verifier to verify the compliance of the equipment control instruction, and generate a new production control instruction; In response to the production control instructions, the temperature change curve and the holding time record are obtained through the parameter collector, and the temperature change curve and the holding time record are classified and evaluated using the random forest algorithm to obtain the final food quality consistency data.

Images