A multi-dimensional parameter monitoring method and system for food production quality control

Abstract

This invention proposes a multi-dimensional parameter monitoring method and system for food product quality control, belonging to the field of parameter monitoring technology. It collects various types of information from food at the serving station, performs image analysis on the food, and compares the images with preset plating standards. By combining the multi-dimensional sensing data with the plating quality judgment results, the food's serving quality status is comprehensively analyzed and evaluated to obtain a serving condition satisfaction evaluation result. A serving instruction is generated based on the serving condition satisfaction evaluation result. The invention also dynamically analyzes and adjusts food parameters by combining the multi-dimensional sensing data with the serving instruction to obtain parameter monitoring optimization information. This invention enables simultaneous acquisition and fusion analysis of multi-source data (images, temperature, weight) during the serving process, and comprehensively controls food plating, serving timing, and food portion sizes accordingly. It also provides a dynamic calorie prediction-based food product quality control scheme.

Description

Technical Field

[0001] This invention proposes a multi-dimensional parameter monitoring method and system for food product quality control, which relates to the field of parameter monitoring technology, specifically to the field of multi-dimensional parameter monitoring technology for food product quality control. Background Technology

[0002] In the food preparation process, some solutions use cameras to photograph and record food, relying on manual sampling to check plating standards afterward. Other solutions use temperature probes to measure food temperature during preparation, only determining if minimum temperature requirements are met. Still others use weighing equipment to sample food weight for cost accounting. However, these solutions operate independently, with image, temperature, and weight data collected, recorded, and judged separately, lacking temporal synchronization and correlation analysis. This makes it difficult to integrate and utilize multi-dimensional quality information of the same food. Furthermore, current plating inspections rely heavily on manual visual judgment, resulting in inconsistent standards and a lack of quantification. The timing of food preparation is determined solely by process node times rather than the actual temperature of the food. Calorie labeling is based on fixed calculations from standard recipes, failing to reflect actual calorie variations due to differences in portion sizes, ingredient ratios, and cooking methods. Moreover, existing technologies lack a mechanism for comprehensive decision-making across different quality indicators. When one indicator is severely substandard, it may be masked by the compliance of other indicators, leading to distorted overall judgment and impacting the reliability of food quality control and food safety assurance. Summary of the Invention

[0003] This invention provides a multi-dimensional parameter monitoring method and system for food product quality control, to solve the above-mentioned problems: This invention proposes a multi-dimensional parameter monitoring method and system for food product quality control, the method comprising: S1. Collect various types of information from the food at the food preparation station using information collection equipment to obtain multi-dimensional sensory data of the food. S2. The food is image analyzed using the multi-dimensional perception data and compared with a preset plating standard to obtain a plating quality judgment result. S3. By combining the multi-dimensional perception data with the plating quality judgment results, the food serving quality status is comprehensively analyzed and evaluated to obtain the serving condition satisfaction evaluation result, and a serving instruction is generated based on the serving condition satisfaction evaluation result. S4. By combining the multi-dimensional sensing data with the meal preparation instructions, food parameters are monitored, dynamically analyzed, and adjusted to obtain parameter monitoring optimization information.

[0004] Furthermore, the system includes: The multi-dimensional perception module is used to collect various types of information from the food at the food preparation station through information acquisition devices to obtain multi-dimensional perception data of the food. The quality judgment module is used to perform image analysis on the food through the multi-dimensional perception data and compare it with the preset plating standards to obtain the plating quality judgment result. The comprehensive analysis module is used to comprehensively analyze and evaluate the food's serving quality status by combining the multi-dimensional perception data with the plating quality judgment results, obtain the serving condition satisfaction evaluation results, and generate serving instructions based on the serving condition satisfaction evaluation results. The analysis and adjustment module is used to dynamically analyze and adjust food parameters by combining the multi-dimensional sensing data with the meal preparation instructions, and to obtain parameter monitoring optimization information.

[0005] The beneficial effects of this invention are as follows: This method solves the technical problem in traditional catering management where image, temperature, and weight data are collected, recorded, and difficult to correlate separately by automatically triggering simultaneous multi-sensor acquisition at the moment of food preparation. By integrating an RGB camera and an infrared temperature probe into the same module and arranging them in a co-positioned manner facing the weighing sensor platform, the technical problems of inconsistent field of view and ambiguous data correspondence among multiple sensors are solved. By attaching a precise timestamp to each data point and performing time-series binding, the technical problem of inconsistent data time bases caused by different sensor acquisition frequencies and response speeds is solved. This enables the simultaneous acquisition of complete, correlated sensory data of image, temperature, and weight at the same time and workstation during food preparation, improving the usability and reliability of multi-source data in food quality control scenarios. Attached Figure Description

[0006] Figure 1 This is a schematic diagram of a multi-dimensional parameter monitoring method for food product quality control. Detailed Implementation

[0007] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0008] In one embodiment of the present invention, a multi-dimensional parameter monitoring method and system for food product quality control is proposed, the method comprising: S1. Collect various types of information from the food at the food preparation station using information collection equipment to obtain multi-dimensional sensory data of the food. S2. The food is image analyzed using the multi-dimensional perception data and compared with a preset plating standard to obtain a plating quality judgment result. S3. By combining the multi-dimensional perception data with the plating quality judgment results, the food serving quality status is comprehensively analyzed and evaluated to obtain the serving condition satisfaction evaluation result, and a serving instruction is generated based on the serving condition satisfaction evaluation result. S4. By combining the multi-dimensional sensing data with the meal preparation instructions, food parameters are dynamically analyzed and adjusted to obtain parameter monitoring optimization information, such as... Figure 1 As shown.

[0009] Multidimensional perception data refers to a collection of digital information about different attributes of food (such as visual appearance, surface temperature, overall weight and collection time) acquired simultaneously at the same workstation through various types of information collection devices. Its core characteristics are the heterogeneity of data sources and temporal correlation.

[0010] Time stamping is a process of uniformly comparing and correcting the timestamps (such as image acquisition time, temperature response time, and weight stability reading time) recorded independently by different acquisition devices. When the time difference between devices exceeds the preset synchronization tolerance, the image and temperature data are corrected by using the weight timestamp as a reference, so that all data are aligned to the same time reference point, thereby ensuring the comparability and consistency of multi-source data in the time dimension.

[0011] Geometric features of food plating refer to the position coordinates, contour boundary curves, coverage area and shape parameters of food extracted from the standardized plating image within the plane of the plate. These geometric quantification indicators provide a spatial positioning basis for subsequent food identification and plating standardization determination.

[0012] The information on food categories and distribution relationships includes two types of content: first, the names of various food items and their independent spatial distribution parameters (centroid coordinates, distribution area, coverage range) on the plate, identified through pixel-by-pixel color comparison and regional connectivity analysis; and second, the records of the relative positional relationships between food items obtained through pairwise comparisons (centroid distance, adjacent boundary length, overlapping area determination). This information fully describes the spatial arrangement pattern of each component of the food.

[0013] The plating quality assessment result is a conclusive evaluation of whether the food appearance conforms to the design specifications by comparing the extracted plating geometric features and food distribution with the position requirements, area ratio requirements and shape requirements in the preset plating standard library, combined with various deviation records and non-conformity lists, and through comprehensive judgment rules.

[0014] The evaluation result of food service condition satisfaction is a comprehensive judgment on whether food can be served. The evaluation is based on three sub-results: plating quality judgment, temperature compliance judgment, and weight compliance judgment. By identifying the highest risk level among the indicators and implementing a veto logic (when a certain risk reaches the preset unacceptable threshold, it is directly judged as unqualified), the comprehensive evaluation result is finally divided into different levels (such as unqualified, conditionally permissible, and fully permissible) as an overall conclusion.

[0015] The meal dispensing instruction is a corresponding control command generated based on the level of the comprehensive evaluation result of the meal dispensing conditions. It includes three types: prohibiting meal dispensing, allowing direct meal dispensing, and allowing meal dispensing with conditions. This instruction directly drives the execution path selection of subsequent operation processes (such as abnormal recording, direct release, or delayed review).

[0016] Parameter monitoring and optimization information is supplementary and adjustment information obtained through dynamic analysis of collected data during the execution of the meal preparation instruction (such as correcting the initial heat reference value by combining temperature values ​​and cooking methods, or initiating delayed temperature verification when preparing meals under certain conditions). This information, along with the original sensing data and judgment results, is packaged and uploaded to the back-end management system for monitoring continuous parameter optimization and quality traceability.

[0017] The "one-vote veto signal" refers to a mandatory non-compliance judgment mechanism triggered when any of the following factors—the severity of plating deviation, the risk level of temperature deviation, or the impact level of portion deviation—reaches a preset unacceptable threshold: non-compliance of a single indicator. This mechanism allows serious non-compliance of a single indicator to directly result in an overall non-compliance evaluation of the meal preparation conditions, without the need for weighted summation of other indicators.

[0018] Time axis interpolation correction is used when asynchronous acquisition of multi-source data is detected. Using a reference timestamp (weight timestamp), the values ​​corresponding to the reference timestamp are estimated by linear interpolation or spline interpolation between the respective acquisition times of image data and temperature data. This achieves data alignment on the time axis and eliminates time deviations caused by equipment response delays or differences in acquisition frequency.

[0019] The preset tolerable deviation upper limit is the allowable fluctuation range for various deviation indicators (such as position offset, area ratio error, and degree of shape distortion) in the judgment of plating quality. Deviations within this range are considered acceptable, while deviations exceeding this range trigger a severity classification record, which is used to distinguish between plating quality problems of the flaw level and the defect level.

[0020] The calorie correction factor is a numerical factor matched from a preset lookup table based on the food cooking method (such as frying, steaming, baking, etc.). This factor is used to multiply the calorie reference value determined based on the type and weight of the ingredients to reflect the impact of different cooking processes on the final calorie content of the food.

[0021] The temperature correction factor is an additional adjustment parameter determined based on the current temperature category of the food (low temperature, normal temperature, high temperature). It is used to perform secondary compensation on top of the basic calorie value to correct the calorie estimation deviation caused by the difference in heat conduction and dissipation of food at different consumption temperatures, and finally obtain calorie information that is closer to the actual consumption state.

[0022] The working principle and technical effect of the above technical solution are as follows: This method places the plate containing food on a weighing sensor on the detection platform. After the weighing sensor detects a change in the weight of the plate, it automatically triggers the system to start the data acquisition process. After the system is activated, three information acquisition devices are simultaneously activated: an RGB camera, an infrared temperature probe, and a weighing sensor. The RGB camera captures a top-down image of the food from directly above the plate, obtaining information on the food's appearance, color distribution, and ingredient layout. The infrared temperature probe collects the surface temperature of the food and the edge temperature of the plate at the same station, obtaining information on the food's thermodynamic state. The weighing sensor collects the total weight of the food and records a timestamp at the moment of acquisition, obtaining the food's mass information. The three sensors are physically arranged in a co-positioned manner by integrating the RGB camera and the infrared temperature probe into the same module, all facing the weighing sensor platform, ensuring that the data collected by the three sensors comes from different physical dimensions of the same food and the same placement state. After the data collection is completed, the system binds the image data, temperature data, and weight data to their respective collection timestamps to form multi-dimensional raw data with time stamps. Then, by using time-series markers, all the data are unified to the same time coordinate system, ultimately forming a complete multi-dimensional sensing dataset with time stamps.

[0023] This method solves the technical problem of separate collection, recording, and correlation of image, temperature, and weight data in traditional catering management by automatically triggering simultaneous multi-sensor acquisition at the moment of food preparation. By integrating an RGB camera and an infrared temperature probe into the same module and arranging them in a co-positioned manner facing the weighing sensor platform, it resolves the technical issues of inconsistent field of view and ambiguous data correspondence among multiple sensors. Furthermore, by attaching precise timestamps to each data point and performing time-series binding, it addresses the technical problem of inconsistent data time bases caused by differences in acquisition frequency and response speed among different sensors. This enables the simultaneous acquisition of complete, correlated sensor data across image, temperature, and weight dimensions at the same time and workstation during food preparation, improving the usability and reliability of multi-source data in food quality control scenarios.

[0024] In one embodiment of the present invention, S1 includes: The system is activated by placing the plate on the weighing sensor on the detection platform, thus obtaining the system activation command. The system activates the RGB camera to capture a top-down image of the food, obtaining raw image data. The system activation command activates the infrared temperature probe to collect the surface temperature of the food and the edge temperature of the plate at the same workstation, thereby obtaining the temperature data of the food. By integrating the RGB camera and the infrared temperature probe into the same module and arranging both lenses toward the weighing sensor platform, a co-position acquisition module that can simultaneously acquire food image information and temperature information is obtained. The system activation command is used to start the weighing sensor to collect the total weight of the food and record the timestamp of the collection time to obtain the weight value and timestamp. By binding the original image data, the temperature data, the weight data, and the timestamp, the multi-dimensional sensing data is time-stamped to obtain a complete multi-dimensional sensing dataset with time stamps.

[0025] The working principle and technical effect of the above technical solution are as follows: This method performs refined time-series synchronization processing on multi-source sensing data. First, the system divides the acquisition time corresponding to each frame of the image from the original image data stream according to the acquisition frame rate of the RGB camera and records it as an image timestamp; it records each response time of the infrared temperature probe as a temperature timestamp and binds it to the temperature value; it records the moment when the weighing sensor reading reaches stability as a weight timestamp and binds it to the weight value. The system compares the image timestamp, temperature timestamp, and weight timestamp and calculates the time difference between the three. When the time difference is within the preset synchronization tolerance range, it is determined to be in a synchronized state, and the original data is used directly; when the time difference exceeds the synchronization tolerance, it is determined to be in an asynchronous state and correction is required. During the correction process, the system uses the weight timestamp as a unified reference benchmark. The reason for choosing the weight timestamp as the benchmark is that the moment when the weighing sensor reaches a stable reading after the plate is placed can most accurately reflect the instantaneous state of the food being served. For image data, the system retrieves images from two adjacent frames before and after the image's timestamp, interpolates based on the time difference between the image timestamp and the weight timestamp, and generates a virtual frame image aligned with the weight timestamp. For temperature data, the system retrieves temperature values ​​from two adjacent temperature samples before and after the temperature timestamp, interpolates based on the time difference between the temperature timestamp and the weight timestamp, and generates an interpolated temperature value aligned with the weight timestamp. Data that is already in a synchronized state is directly used as alignment data. Finally, all aligned image data, temperature data, weight data, and weight timestamps are collected and packaged to form three-source sensing data with the weight timestamp as a unified time reference point.

[0026] This method addresses the technical issue of inconsistencies in the timeline of multi-source data from RGB cameras, infrared temperature probes, and weighing sensors due to their different acquisition frequencies (frame-rate sampling for images, intermittent response for temperatures, and single, stable readings for weight). By using weight timestamps as a unified benchmark instead of simply discarding asynchronous data, it solves the problem of low data utilization caused by the forced discarding of valid samples due to data asynchrony. Furthermore, by employing frame-wise and sample-wise interpolation correction methods, it addresses the issue of asynchronous data being unable to participate in subsequent comprehensive analysis. This achieves precise alignment of data from the three heterogeneous sensors in the time dimension, ensuring that subsequent comprehensive evaluations are based on data truly reflecting the food's state at the same moment. This avoids the risk of misjudging the current food quality based on outdated data due to time sequence misalignment, thus improving the accuracy and reliability of multi-source data fusion evaluation.

[0027] In one embodiment of the present invention, the step of binding the original image data, the temperature data, the weight data, and the timestamp to perform time-series marking on the multi-dimensional sensing data to obtain a complete multi-dimensional sensing dataset with time-series marking includes: By dividing the original image data into image acquisition times according to the acquisition frame rate and recording the image acquisition times as image timestamps, an image dataset with image acquisition time identifiers is obtained; By recording the response time of the infrared temperature probe corresponding to the temperature numerical data as a temperature timestamp, and binding the temperature timestamp with the temperature numerical data, a temperature dataset with a temperature acquisition time identifier is obtained. By recording the stable reading time of the weighing sensor corresponding to the weight value as a weight timestamp, and binding the weight timestamp with the weight value, a weight dataset with a weight acquisition time identifier is obtained. By comparing the image timestamp, the temperature timestamp, and the weight timestamp, the time difference between the acquisition times of the three is calculated. When the time difference exceeds the preset synchronization tolerance, it is marked as an acquisition asynchronous state, and the data synchronization verification result is obtained. When the data synchronization verification result is that the acquisition is out of sync, the image data and temperature data are corrected by time axis interpolation based on the weight timestamp. The corrected data and weight data are aligned to the same time reference point to obtain the time-aligned three-source sensing data. By packaging and binding the three-source sensing data with the weight timestamp, a unified time-series label is applied to the multi-dimensional sensing data, resulting in a complete multi-dimensional sensing dataset with time-series stamps.

[0028] Specifically, when the data synchronization verification result indicates an asynchronous acquisition state, time-axis interpolation correction is performed on the image data and temperature data based on the weight timestamp. The corrected data is then aligned with the weight data to the same time reference point to obtain time-aligned three-source sensing data, including: When the data synchronization verification result is in the state of asynchronous acquisition, the original image data of the two adjacent frames before and after the image timestamp are retrieved, and the time axis interpolation processing of the two adjacent frames of image data is performed according to the time difference between the image timestamp and the weight timestamp to obtain interpolated image data aligned with the weight timestamp. When the data synchronization verification result is that the acquisition is out of sync, the temperature data of two adjacent samples before and after the temperature timestamp are retrieved, and the time axis interpolation processing is performed on the two adjacent temperature data according to the time difference between the temperature timestamp and the weight timestamp to obtain the interpolated temperature data aligned with the weight timestamp. When the data synchronization verification result is in the acquisition synchronization state, the original image data is directly used as image data aligned with the weight timestamp to obtain the original image aligned data; When the data synchronization verification result is in the acquisition synchronization state, the temperature value data is directly used as the temperature data aligned with the weight timestamp to obtain the original temperature aligned data. By binding the interpolated image data or the original image alignment data with the weight timestamp, an image dataset aligned with the weight timestamp is obtained; By binding the interpolated temperature data or the original temperature aligned data with the weight timestamp, a temperature dataset aligned with the weight timestamp is obtained; By aggregating and integrating the image dataset aligned with the weight timestamp, the temperature dataset aligned with the weight timestamp, and the weight values ​​and weight timestamps, three-source sensing data with the weight timestamp as a unified time reference point is obtained.

[0029] The working principle and technical effects of the above-mentioned technical solution are as follows: This method utilizes image data from a complete multi-dimensional perceptual dataset that has undergone time-series alignment for plating quality analysis. First, the original images are standardized, including distortion correction, illumination normalization, and background segmentation, eliminating image inconsistencies caused by differences in shooting angle, lighting conditions, and plate shape, resulting in standardized plating images. The system extracts the position, outline boundaries, and coverage area of ​​food on the plate from the standardized images, forming a geometric feature description of the overall spatial layout of the food. Subsequently, the system uses the plating geometric feature information to assist image analysis, automatically identifying different food categories, such as main dishes, side dishes, and garnishes, and further determining the distribution areas of various food items on the plate and the relative positional relationships between different food items. After obtaining the food category and spatial distribution information, the system calls a preset plating standard library, which sets positional requirements, area ratio requirements, and morphological integrity requirements for different food types. The system compares the actually extracted geometric features and food distribution relationships with each requirement in the standard library, marking all non-compliant items and their degree of deviation. Ultimately, the system comprehensively determines whether the plating is up to standard based on the comparison results, and outputs three judgment statuses: qualified, requires manual adjustment, or unqualified and rejected, along with a detailed record of each deviation.

[0030] This method solves the technical problems of inconsistent standards, strong subjectivity, and difficulty in quantification in traditional manual visual inspection of food plating by extracting geometric features and identifying ingredient categories from standardized images of prepared food. By combining the geometric features of the plating with information on the distribution of ingredient categories, it addresses the issue of judging plating quality solely based on overall outlines without evaluating the rationality of the arrangement between different ingredients. Furthermore, by quantitatively comparing the actual plating state with a preset plating standard library item by item, it solves the technical problems of lacking objective basis for plating quality judgment and difficulty in traceability and reproducibility. This achieves automated, standardized, and quantitative evaluation of plating quality, transforming plating control from an experience-based behavior relying on the chef's personal aesthetics to a systematic quality control method with preset standards, automatic judgment, and traceable deviations, thus improving the consistency of food plating across different stores and among different operators in chain restaurant settings.

[0031] In one embodiment of the present invention, S2 includes: Standardized plating images are obtained by standardizing the image data in the complete multi-dimensional perception dataset. The position, outline, and coverage area of ​​food on the plate are extracted using the standardized plating image to obtain the geometric feature information of the plating. By combining the geometric feature information of the plating with the standardized plating image, the categories of different ingredients are identified, and the distribution area of ​​each type of ingredient on the plate and the positional relationship between them are determined, thereby obtaining information on the categories and distribution relationships of ingredients. The geometric features of the plating and the information on the categories and distribution of the ingredients are compared item by item with the position requirements, area ratio requirements and shape requirements in the preset plating standard library to obtain the comparison results of each indicator and a list of non-compliance items. The plating quality is comprehensively judged by comparing the results of various indicators and the list of non-conformities. The plating quality judgment result is divided into three states: qualified, requiring manual adjustment, or unqualified and rejected. Various deviation records are also obtained.

[0032] The working principle and technical effect of the above technical solution are as follows: This method system compares the food areas in the standardized plating image pixel by pixel with a preset food color feature library. Each food corresponds to a specific set of color feature ranges. By matching pixel colors, the system achieves preliminary classification and labeling of the pixel areas of various food items, forming a color-based food area classification map. However, since different food items may have similar colors, the preliminary classification may lead to misclassification. Therefore, the system further utilizes the contour boundaries in the extracted plating geometric feature information to perform regional connectivity analysis. Spatially connected pixel areas belonging to the same preliminary classification are merged into complete areas, while isolated discrete pixels that are inconsistent with the color features of the surrounding areas are identified as possible misclassifications and reclassified to obtain a corrected food area distribution map. After determining the pixel areas of various food items, the system maps these pixel areas to the physical coordinate system of the plate. During the mapping process, the system calls the preset plate coordinate calibration parameters to convert the pixel coordinate origin to the origin of the plate's physical coordinate system and converts the pixel size to the physical size, establishing a pixel-physical coordinate transformation matrix. All pixel coordinates of various ingredients are batch-converted into physical coordinates on the plate using this matrix, forming a set of physical coordinate points for each ingredient. Based on these sets, the system calculates the centroid physical coordinates (determined by the average of the horizontal and vertical coordinates of all points), distribution area (calculated by the area of ​​the closed contour enclosed by the boundary points and normalized to the total area of ​​the plate to obtain the area percentage), and coverage area (defined by the maximum and minimum values ​​along the horizontal and vertical axes to define a rectangular region) for each ingredient. Finally, the system summarizes the centroid physical coordinates, area percentage, and coverage area of ​​each ingredient to form independent spatial distribution parameters for each ingredient, fully expressed using the plate's physical coordinate system.

[0033] This method employs a two-stage recognition approach combining pixel-by-pixel color comparison and contour boundary connectivity analysis. This addresses the technical problem of classification confusion and misjudgment caused by similar food colors when relying solely on color for food identification, thus improving the accuracy of food category recognition. By establishing a pixel-to-physical coordinate transformation matrix and mapping food pixel regions to the physical coordinate system of the plate, it solves the technical problem of unclear correspondence between image coordinates and the actual spatial position of the plate, and the inability to accurately locate deviations in plating evaluation. Furthermore, by calculating quantitative spatial parameters such as the centroid coordinates, area proportion, and coverage of each type of food, it solves the technical problem of the inability to objectively describe and quantify the subjective feeling of whether a plating is aesthetically pleasing. This achieves precise spatial positioning and quantitative description of each food item in the plating, deepening the evaluation of plating quality from an overall impression to a fine-grained analysis of the rationality of the layout of each food item, significantly improving the refinement of plating quality control and the accuracy of deviation location.

[0034] In one embodiment of the present invention, the step of identifying different food categories by combining the plating geometric feature information with the standardized plating image, and determining the distribution area of ​​each type of food on the plate and the positional relationship between them, to obtain food category and distribution relationship information, includes: By comparing the food areas in the standardized plating image with a preset food color feature library pixel by pixel, the pixel areas belonging to various types of food are initially classified and marked to obtain a color-based food area classification map. By combining the food region classification map with the contour boundaries in the plating geometric feature information to perform region connectivity analysis, discrete pixels that are misclassified due to similar colors are merged or reclassified to obtain a corrected food region distribution map. By mapping the pixel regions of various ingredients in the food distribution map to the plate coordinate system, the centroid coordinates, distribution area and coverage of each type of ingredient on the plate are determined, and the independent spatial distribution parameters of each type of ingredient are obtained. By comparing the independent spatial distribution parameters of each type of food in pairs, the centroid distance, adjacent boundary length, and whether there is an overlapping area between any two types of food are determined, and the relative positional relationship between the food is recorded. By comparing the recorded relative positional relationships between the ingredients with a preset food category standard ingredient distribution template, the system identifies whether there are any abnormal layout deviations between different types of ingredients, and obtains the ingredient layout deviation identification results. By summarizing and integrating the independent spatial distribution parameters with the results of the food layout deviation identification, a complete description is given of the name, area ratio, spatial location, and relationship with adjacent food items for each type of food, thus obtaining information on food categories and distribution relationships.

[0035] Specifically, by mapping the pixel regions of various ingredients in the food distribution map to the plate coordinate system, the centroid coordinates, distribution area, and coverage range of each type of ingredient on the plate are determined, thereby obtaining independent spatial distribution parameters for each type of ingredient, including: By calling preset plate coordinate calibration parameters, the pixel coordinate origin in the food area distribution map is mapped to the origin of the plate physical coordinate system, and the pixel size and physical size are proportionally converted to obtain the pixel-physical coordinate transformation matrix. By substituting all pixel coordinates of each type of food in the food distribution map into the pixel-physical coordinate transformation matrix, the pixel coordinates of each type of food are converted into physical coordinates of the plate in batches, thus obtaining the physical coordinate point set of each type of food. By calculating the centroid of the physical coordinate point set of each type of food, the average of the abscissas of all coordinates in the point set is used as the abscissa of the centroid, and the average of the ordinates of all coordinates is used as the ordinate of the centroid, thus obtaining the physical coordinates of the centroid of each type of food. By connecting the boundary points of the physical coordinate point set of each type of food to form a closed contour, the area of ​​the region enclosed by the closed contour is calculated, and the area is normalized with the total area of ​​the plate to obtain the area ratio of each type of food in the plate. By determining the maximum and minimum values ​​of the physical coordinate point set of each type of food in the horizontal and vertical directions, the rectangular area defined by the difference between the maximum and minimum values ​​of the horizontal axis and the difference between the maximum and minimum values ​​of the vertical axis is used as the coverage area of ​​that type of food, thus obtaining the plate coverage range of each type of food. By summarizing the centroid physical coordinates of the various ingredients, their area proportions, and the coverage area of ​​the plate, independent spatial distribution parameters of the various ingredients expressed in the plate physical coordinate system are obtained.

[0036] The working principle and technical effect of the above technical solution are as follows: This method utilizes the plating quality judgment results obtained in the preceding steps, combined with temperature and weight information from multi-dimensional sensing data, to comprehensively evaluate the overall serving quality of the food. The system first checks the plating quality judgment results to confirm whether the plating meets basic standards; simultaneously, it compares the temperature values ​​from the multi-dimensional sensing data with the preset suitable serving temperature range to determine if the food temperature is within the ideal serving temperature range; and it compares the weight values ​​with the preset standard weight range to determine if the food portion meets the standard requirements. Through these three independent judgments, the system obtains three evaluation information items: plating quality status, temperature compliance and deviation direction, and weight compliance and deviation direction. Subsequently, the system integrates these three pieces of information, not through simple arithmetic addition, but by performing an overall evaluation of the serving condition satisfaction based on the compliance status and deviation degree of each indicator, resulting in a comprehensive evaluation result and detailed evaluation records for each indicator. Finally, the system generates corresponding meal preparation instructions based on the overall evaluation result (three levels: excellent, qualified, and unqualified): for excellent overall evaluation, meals can be prepared directly; for qualified overall evaluation but with some minor deviations, an instruction is generated to allow meals to be prepared with conditions attached; for unqualified overall evaluation, an instruction is generated to prohibit meals from being prepared.

[0037] This method comprehensively evaluates the results of independent assessments of plating quality, temperature, and weight, addressing the technical problem of unacceptable overall food quality despite passing a single-dimensional indicator (e.g., perfectly plated food that is already cold, or food that meets temperature standards but is severely underweight). By setting three evaluation levels—excellent, acceptable, and unacceptable—and generating corresponding serving instructions, it overcomes the technical problem of traditional serving control's binary "pass" or "fail" results lacking refined decision support. Furthermore, by introducing intermediate states with accompanying conditions, it addresses the lack of flexible handling mechanisms for serving scenarios bordering on acceptable quality, ensuring a minimum quality standard while avoiding excessively low yields due to extreme stringency. This achieves multi-dimensional comprehensive judgment and tiered decision-making for food quality, upgrading serving control from single-point threshold judgment to multi-factor comprehensive evaluation, thus improving the scientific rigor and precision of serving decisions.

[0038] In one embodiment of the present invention, S3 includes: By judging the plating quality assessment results for the condition compliance, temperature compliance, and weight compliance, the plating quality assessment results, temperature compliance information, and weight compliance information are obtained. By comprehensively considering the results of the plating quality assessment, the temperature compliance information, and the weight compliance information, the overall evaluation of the food's service conditions is conducted, resulting in a comprehensive evaluation result of the service conditions and detailed evaluation records of each indicator. Based on the three levels of the comprehensive evaluation results of the meal preparation conditions, corresponding meal preparation instructions are generated, including instructions that allow direct meal preparation, allow meal preparation with conditions, or prohibit meal preparation, and evaluation records of various indicators are obtained.

[0039] The working principle and technical effects of the above-mentioned technical solution are as follows: This method system compares each deviation record in the plating quality judgment result with the preset tolerable deviation upper limit. Based on the degree to which the deviation exceeds the tolerable range, the plating deviation is classified into severity levels: slight deviation, moderate deviation, and severe deviation, forming a plating deviation severity classification record. Simultaneously, the system compares the direction of temperature deviation (too high or too low) with a preset deviation consequence comparison table. For example, excessively high temperatures may lead to burns or deterioration of taste, while excessively low temperatures may lead to food safety hazards or loss of flavor. Based on this, the potential consequences of temperature deviation are qualitatively classified, forming a temperature deviation risk level record. For weight deviation, the system compares the degree of overweight or underweight with a preset weight deviation impact standard. Overweight may lead to cost overruns, while underweight may lead to consumer dissatisfaction. Based on this, the impact level of weight deviation is assessed, forming a weight deviation impact level record. After obtaining the risk levels of the three indicators, the system identifies the highest risk level among the three: when the risk level of any one of the plate placement deviation, temperature deviation, or weight deviation reaches the preset unacceptable threshold, the system immediately triggers a veto signal, directly determining the comprehensive evaluation result as unqualified, and recording the specific indicator that triggered the veto and its level as the reason for the veto; only when the risk levels of all three indicators are below the unacceptable threshold will the system output the comprehensive evaluation result in the conventional manner.

[0040] This method addresses the technical problem of traditional weighted comprehensive evaluation methods where a serious failure in one dimension can be masked by high scores in other dimensions, leading to the erroneous release of substandard food. By establishing a veto mechanism based on the highest risk level, it solves the technical problem of incomparable and unified decision-making between indicators of different dimensions and natures. Through an asymmetric evaluation logic where any indicator reaching an unacceptable threshold results in total rejection, it resolves the "weakest link" problem in food safety and consumer experience. Food quality is determined by the worst-performing dimension, not the average of all dimensions. This achieves a baseline guarantee for food quality evaluation, ensuring that food will not be released to consumers if any indicator reaches a dangerous or unacceptable level, significantly improving the safety assurance capability of food quality control and the level of consumer rights protection.

[0041] In one embodiment of the present invention, the overall evaluation of the food's serving conditions is conducted by comprehensively considering the plating quality judgment results, the temperature compliance information, and the weight compliance information, resulting in a comprehensive evaluation result of the serving conditions and detailed evaluation records of each indicator, including: By comparing the various deviation records in the plating quality judgment results with the preset tolerable deviation upper limit, the plating deviation is confirmed item by item to determine whether it is within the allowable range, and a classification record of the severity of plating deviation is obtained. By comparing the temperature deviation direction in the temperature compliance and deviation direction records with a preset deviation consequence comparison table, the food safety risk or taste deterioration risk is qualitatively assessed, and a temperature deviation risk level record is obtained. By comparing the degree of overweight or underweight in the weight compliance and deviation direction records with the preset component deviation impact standard, the impact of cost deviation or consumer satisfaction is evaluated, and a component deviation impact level record is obtained. By aggregating the severity grading records of the plate placement deviation, the risk level records of the temperature deviation, and the impact level records of the component deviation, the highest risk level among the various indicators is identified. When the highest risk level reaches a preset unacceptable threshold, a veto signal is triggered, and the veto trigger signal and the corresponding veto reason record are obtained. When the rejection trigger signal is in the triggered state, the comprehensive evaluation result is set to the unqualified level, and the rejection reason is recorded as the main output of the comprehensive evaluation result of the meal preparation conditions, so as to obtain the rejection-type comprehensive evaluation result of the meal preparation conditions. When the rejection trigger signal is in an untriggered state, the compliance status, deviation degree, and risk level of the three indicators of plating, temperature, and weight are summarized, and the overall qualification degree of each indicator is described to obtain a comprehensive evaluation result of the comprehensive meal preparation conditions and a detailed evaluation record of each indicator.

[0042] Specifically, by aggregating the severity grading records of the plate-setting deviation, the risk level records of the temperature deviation, and the impact level records of the component deviation, the highest risk level among these indicators is identified. When the highest risk level reaches a preset unacceptable threshold, a veto signal is triggered, obtaining the veto trigger signal and the corresponding veto reason record, including: By aggregating the severity level values ​​from the severity grading record of the plate placement deviation, the risk level values ​​from the risk level record of the temperature deviation, and the impact level values ​​from the impact level record of the component deviation, a three-dimensional index level vector composed of plate placement level values, temperature level values, and weight level values ​​is formed, thus obtaining the three-dimensional index level vector. By comparing the plating level value, temperature level value, and weight level value in the three-dimensional indicator level vector with a preset unacceptable threshold one by one, when any level value is greater than or equal to the unacceptable threshold, the indicator is marked as a trigger item, and when all level values ​​are less than the unacceptable threshold, all indicators are marked as safe items, thus obtaining the trigger status marking of each indicator. When any one of the trigger status flags of each indicator exists, a veto signal is triggered, and the indicator name and level value corresponding to the trigger item are recorded as the veto reason; when all the trigger status flags of each indicator are safe items, no veto signal is triggered and the veto reason record is empty, thus obtaining the veto trigger signal and the corresponding veto reason record.

[0043] The working principle and technical effect of the above technical solution are as follows: This method determines whether the food dispensing instruction is "prohibit dispensing". If so, it directly triggers the anomaly recording process, completely archiving and saving all sensory data and evaluation results of the food as raw materials for quality traceability and anomaly analysis. If the food dispensing instruction allows direct dispensing or allows dispensing under certain conditions, the system enters the calorie assessment process: it retrieves the actual weight value from the multi-dimensional sensory data and the ingredient distribution information obtained from image analysis, compares it with the preset correspondence between ingredient categories and calories, and determines the basic calorie reference value of the food. The system further retrieves the temperature value and combines it with the cooking method of the food to dynamically correct the calorie reference value. Different cooking methods (such as frying, steaming, and braising) will significantly affect the final calorie of the food, and the system adjusts the calorie value according to the temperature status and cooking method. For the case where the food dispensing instruction allows dispensing under certain conditions, the system starts a delayed review process after the food is served: after waiting for a preset time interval, the temperature of the food is re-collected, and the re-collected temperature value is compared with the safe temperature range to observe whether the food temperature remains in a safe state within a reasonable time after dispensing and to record the temperature change. Finally, the system integrates all data, including multi-dimensional perception data, plating quality judgment results and deviation records, meal preparation instructions, adjusted calorie information, and delayed review records, into a complete data package and uploads it to the backend management system.

[0044] This method addresses the technical problem of lacking corresponding follow-up processing mechanisms for different meal service decisions by implementing differentiated processing flows based on the type of meal service instruction. Foods prohibited from service require complete record-keeping of the reasons for their prohibition for traceability and improvement, while foods permitted for service require supplemental calorie information to inform consumers. By comprehensively applying actual weight, ingredient distribution information, and temperature status to calorie calculation, it solves the technical problem that traditional calorie labeling, based on fixed recipes, cannot reflect actual calorie fluctuations in the same dish caused by differences in portion size, ingredient ratios, and cooking conditions. By introducing a delayed verification mechanism for scenarios where meal service is permitted under certain conditions, it addresses the technical problem of unmonitored potential risks from food that is at the moment of service but rapidly cools to an unsafe temperature after service. By uploading complete data packages to the backend management system, it solves the technical problem of isolated quality data for each meal, which cannot form a quality traceability chain or a basis for process optimization. This achieves a complete closed-loop management from meal service decision-making to calorie output to status verification to data archiving, improving the completeness of meal service quality control and the ability to deeply mine data value.

[0045] In one embodiment of the present invention, S4 includes: By determining whether the food delivery instruction is to prohibit food delivery, when the food delivery instruction is to prohibit food delivery, an abnormal recording process is triggered, marking all sensory data and evaluation results of the food and archiving and saving them to obtain an abnormal data archive record. When the food preparation instruction allows direct food preparation or allows food preparation with conditions, the weight values ​​and food categories and distribution information in the complete multi-dimensional perception dataset with time stamps are retrieved, and the calorie reference values ​​are determined by comparing them with the preset food category and calorie correspondence, thus obtaining the initial calorie reference information. By retrieving the temperature values ​​from the complete multi-dimensional sensing dataset with time stamps and combining them with the cooking method of the food, the initial heat reference information is adjusted, and the heat information is dynamically corrected to obtain the adjusted heat information. When the food delivery instruction is conditionally permitted to be delivered, the food temperature is re-collected after a set time interval by starting a delayed verification process. The re-collected temperature value is compared with the safe temperature range to observe whether the temperature is within the safe range and record the temperature changes, thus obtaining a temperature verification record after food delivery. By integrating the complete multi-dimensional sensing dataset with time stamps, the plating quality judgment results and deviation records, the meal serving instructions, the adjusted calorie information, and the post-meal temperature verification records into a complete data package and uploading it to the backend management system, parameter monitoring and optimization information is obtained.

[0046] The working principle and technical effect of the above technical solution are as follows: This method compares the temperature values ​​in the multi-dimensional sensing data with preset low-temperature, normal-temperature, and high-temperature ranges to determine which range the current temperature state of the food belongs to, obtaining a food temperature state category identifier. This identifier reflects whether the food is in a freshly cooked high-temperature state, a suitable normal-temperature state, or a cooled low-temperature state. Simultaneously, the system identifies the oily sheen areas on the food surface from the image data in the multi-dimensional sensing data. By analyzing the area and distribution of characteristic light spots formed by oil reflection in the image, the system determines the degree of oil adhesion on the food surface. When the oil sheen coverage exceeds a preset threshold, it is marked as a high-oil state, obtaining a food surface oil state identifier. Subsequently, the system combines the temperature state category identifier and the oil state identifier to infer the cooking method of the food: high temperature and high oil correspond to frying or deep-frying; high temperature and low oil correspond to steaming or boiling; normal temperature and high oil correspond to braising or stewing; and normal temperature and low oil correspond to cold dishes or blanching. After determining the cooking method, the system matches this method with a preset table of calorie correction coefficients for different cooking methods to obtain the corresponding coefficient. For example, fried foods require a significant increase in calorie value, steamed foods require maintaining or decreasing calorie value, and braised foods fall somewhere in between. The system combines the initial calorie reference information with this correction coefficient to dynamically adjust the calorie reference value based on unit weight × standard calories, resulting in a calorie baseline value corrected for the cooking method. Finally, the system combines the calorie baseline value with a temperature correction factor in the temperature status category. The actual calories ingested for the same food differ at different consumption temperatures due to variations in heat conduction and loss. This difference is compensated for by the temperature correction factor, ultimately yielding the adjusted calorie information.

[0047] This method automatically infers the cooking method from two objectively perceived dimensions: temperature and oil coverage. This solves the technical problem of traditional calorie estimation, which requires manual selection and input of the cooking method, leading to the risk of human error or false alarms. It automates the calorie correction process. By introducing a cooking method correction coefficient to dynamically adjust the initial calorie reference value, it addresses the issue of significant calorie differences in the same ingredient due to different cooking methods (e.g., steamed fish vs. fried fish), where fixed calorie values ​​cannot reflect these differences. Furthermore, by introducing a temperature correction factor for secondary compensation, it solves the problem of differences in actual calorie intake due to variations in heat conduction and dissipation at different consumption temperatures for the same food. This represents an upgrade from fixed recipe calorie values ​​to dynamic calorie assessment based on actual conditions. The calorie output reflects the actual calorie fluctuations of the same dish due to differences in weight, ingredient ratios, cooking methods, and consumption temperatures, significantly improving the accuracy and personalization of calorie estimation.

[0048] In one embodiment of the present invention, the step of adjusting the initial calorific reference information by retrieving temperature values ​​from the complete multi-dimensional sensing dataset with time stamps and combining them with the cooking method of the food, and dynamically correcting the calorific information to obtain the adjusted calorific information includes: By comparing the temperature values ​​in the complete multi-dimensional sensing dataset with time stamps with preset low temperature range, normal temperature range and high temperature range, the temperature state category to which the current temperature of the food belongs is determined, and the food temperature state category identifier is obtained. By identifying the glossy areas on the food surface using image data from the complete multi-dimensional perception dataset, the degree of grease adhesion on the food surface is determined. When the gloss coverage exceeds a preset threshold, it is marked as a high-grease state, thus obtaining a grease state identifier for the food surface. By combining the food temperature state category identifier and the food surface oil state identifier, the cooking method of the food is inferred in reverse to obtain the inferred food cooking method category; By matching the inferred food cooking method category with a preset cooking method calorie correction coefficient lookup table, the calorie correction coefficient corresponding to the cooking method is obtained, and the cooking method correction coefficient is obtained. By combining the initial calorie reference information with the cooking method correction coefficient, the calorie reference value based on unit weight × standard calories is dynamically adjusted to obtain the basic calorie value; By combining the baseline calorie value with the temperature correction factor in the food temperature state category identifier, the calorie differences caused by heat conduction and dissipation at different consumption temperatures of the same food are compensated and adjusted to obtain the adjusted calorie information.

[0049] Specifically, by combining the food temperature state category identifier and the food surface oil state identifier, the cooking method of the food is inferred in reverse to obtain the inferred food cooking method category, including: By comparing the temperature status in the food temperature status category identifier with the preset high-temperature cooking temperature threshold, when the temperature status is in the high-temperature range, it is marked as a high-temperature cooking candidate, and when the temperature status is in the non-high-temperature range, it is marked as a room-temperature cooking candidate, thus obtaining a temperature-cooking method mapping identifier. By comparing the degree of oil adhesion in the oil state identifier on the food surface with a preset high oil threshold, when the degree of oil adhesion exceeds the preset high oil threshold, it is marked as a high oil cooking candidate, and when the degree of oil adhesion does not exceed the preset high oil threshold, it is marked as a low oil cooking candidate, thus obtaining an oil-cooking method mapping identifier; By combining the temperature-cooking method mapping identifier and the oil-cooking method mapping identifier, a two-dimensional temperature-oil combination identifier is formed. The two-dimensional combination identifier includes four combination states: high temperature and high oil, high temperature and low oil, room temperature and high oil, and room temperature and low oil, thus obtaining the temperature-oil combination identifier. By identifying the temperature-oil combination as high temperature and high oil content, and matching the cooking method as deep-frying or pan-frying, the candidate result of the first cooking method is obtained. By identifying the temperature-oil combination as high temperature and low oil, matching the cooking method as steaming or boiling, a second cooking method candidate result is obtained; By identifying the temperature-oil combination as room temperature and high oil content, and matching the cooking method as braising or stewing, a third cooking method candidate result is obtained; By identifying the temperature-oil combination as room temperature and low oil, and matching the cooking method as cold dish or blanching, a fourth cooking method candidate result is obtained.

[0050] The working principle and technical effect of the above technical solution are as follows: Automatic reverse inference of food cooking methods is achieved through a combination of two dimensions: temperature and oil status. Specifically, the system first compares the temperature status in the food temperature status category with a preset high-temperature cooking temperature threshold. When the temperature is in the high-temperature range, it is marked as a high-temperature cooking candidate; when the temperature is in the non-high-temperature range, it is marked as a room-temperature cooking candidate, thus obtaining a temperature-cooking method mapping identifier. Simultaneously, the system compares the degree of oil adhesion in the food surface oil status identifier with a preset high-oil threshold. When the degree of oil adhesion exceeds the preset high-oil threshold, it is marked as a high-oil cooking candidate; when the degree of oil adhesion does not exceed the preset high-oil threshold, it is marked as a low-oil cooking candidate, thus obtaining an oil-cooking method mapping identifier. The system combines these two mapping identifiers to form four temperature-oil combination identifiers: high temperature and high oil, high temperature and low oil, room temperature and high oil, and room temperature and low oil. For each combination of identifiers, the system matches the corresponding cooking method category: when the temperature-oil combination identifier is high temperature and high oil, it matches deep-frying or pan-frying; when the combination identifier is high temperature and low oil, it matches steaming or boiling; when the combination identifier is room temperature and high oil, it matches braising or stewing; and when the combination identifier is room temperature and low oil, it matches cold salad or blanching. Through this series of comparisons, mappings, and matches, the system can automatically infer the cooking method category of food without requiring any manual input of cooking information, relying entirely on objectively perceived data.

[0051] This method uses a combination of two objectively perceived dimensions—temperature and oil condition—to infer cooking methods, thus solving the technical problems of manual input and human error inherent in traditional calorie estimation. By comparing continuous values ​​of temperature and oil condition with preset thresholds and converting them into discrete combination identifiers, it addresses the difficulty of directly using continuous perceived data for category matching and decision-making. Furthermore, by establishing a deterministic mapping between four temperature-oil combination identifiers and four cooking methods, it overcomes the challenge of automatically identifying cooking methods for the same ingredient based on appearance characteristics. This achieves complete automation of cooking method identification, enabling calorie correction without any manual intervention. This reduces operator workload and eliminates calorie estimation biases caused by human error, false alarms, or inconsistent subjective judgment, further improving the objectivity and stability of calorie output.

[0052] In one embodiment of the present invention, the system includes: The multi-dimensional perception module is used to collect various types of information from the food at the food preparation station through information acquisition devices to obtain multi-dimensional perception data of the food. The quality judgment module is used to perform image analysis on the food through the multi-dimensional perception data and compare it with the preset plating standards to obtain the plating quality judgment result. The comprehensive analysis module is used to comprehensively analyze and evaluate the food's serving quality status by combining the multi-dimensional perception data with the plating quality judgment results, obtain the serving condition satisfaction evaluation results, and generate serving instructions based on the serving condition satisfaction evaluation results. The analysis and adjustment module is used to dynamically analyze and adjust food parameters by combining the multi-dimensional sensing data with the meal preparation instructions, and to obtain parameter monitoring optimization information.

[0053] The working principle and technical effect of the above technical solution are as follows: This method places the plate containing food on a weighing sensor on the detection platform. After the weighing sensor detects a change in the weight of the plate, it automatically triggers the system to start the data acquisition process. After the system is activated, three information acquisition devices are simultaneously activated: an RGB camera, an infrared temperature probe, and a weighing sensor. The RGB camera captures a top-down image of the food from directly above the plate, obtaining information on the food's appearance, color distribution, and ingredient layout. The infrared temperature probe collects the surface temperature of the food and the edge temperature of the plate at the same station, obtaining information on the food's thermodynamic state. The weighing sensor collects the total weight of the food and records a timestamp at the moment of acquisition, obtaining the food's mass information. The three sensors are physically arranged in a co-positioned manner by integrating the RGB camera and the infrared temperature probe into the same module, all facing the weighing sensor platform, ensuring that the data collected by the three sensors comes from different physical dimensions of the same food and the same placement state. After the data collection is completed, the system binds the image data, temperature data, and weight data to their respective collection timestamps to form multi-dimensional raw data with time stamps. Then, by using time-series markers, all the data are unified to the same time coordinate system, ultimately forming a complete multi-dimensional sensing dataset with time stamps.

[0054] This method solves the technical problem of separate collection, recording, and correlation of image, temperature, and weight data in traditional catering management by automatically triggering simultaneous multi-sensor acquisition at the moment of food preparation. By integrating an RGB camera and an infrared temperature probe into the same module and arranging them in a co-positioned manner facing the weighing sensor platform, it resolves the technical issues of inconsistent field of view and ambiguous data correspondence among multiple sensors. Furthermore, by attaching precise timestamps to each data point and performing time-series binding, it addresses the technical problem of inconsistent data time bases caused by differences in acquisition frequency and response speed among different sensors. This enables the simultaneous acquisition of complete, correlated sensor data across image, temperature, and weight dimensions at the same time and workstation during food preparation, improving the usability and reliability of multi-source data in food quality control scenarios.

[0055] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for monitoring multi-dimensional parameters for food product quality control, characterized in that, The method includes: S1. Collect various types of information from the food at the food preparation station using information collection equipment to obtain multi-dimensional sensory data of the food. S2. The food is image analyzed using the multi-dimensional perception data and compared with a preset plating standard to obtain a plating quality judgment result. S3. By combining the multi-dimensional perception data with the plating quality judgment results, the food serving quality status is comprehensively analyzed and evaluated to obtain the serving condition satisfaction evaluation result, and a serving instruction is generated based on the serving condition satisfaction evaluation result. S4. By combining the multi-dimensional sensing data with the meal preparation instructions, food parameters are monitored, dynamically analyzed, and adjusted to obtain parameter monitoring optimization information.

2. The multi-dimensional parameter monitoring method for food product quality control according to claim 1, characterized in that, S1 includes: The system is activated by placing the plate on the weighing sensor on the detection platform, thus obtaining the system activation command. The system activates the RGB camera to capture a top-down image of the food, obtaining raw image data. The system activation command activates the infrared temperature probe to collect the surface temperature of the food and the edge temperature of the plate at the same workstation, thereby obtaining the temperature data of the food. The system activation command is used to start the weighing sensor to collect the total weight of the food and record the timestamp of the collection time to obtain the weight value and timestamp. By binding the original image data, the temperature data, the weight data, and the timestamp, the multi-dimensional sensing data is time-stamped to obtain a complete multi-dimensional sensing dataset with time stamps.

3. The multi-dimensional parameter monitoring method for food product quality control according to claim 2, characterized in that, The process involves binding the original image data, temperature data, weight data, and timestamp to perform time-series tagging on the multi-dimensional sensing data, resulting in a complete multi-dimensional sensing dataset with time-series tags, including: By dividing the original image data into image acquisition times according to the acquisition frame rate and recording the image acquisition times as image timestamps, an image dataset is obtained; By recording the response time of the infrared thermometer corresponding to the temperature numerical data as a temperature timestamp, and binding the temperature timestamp with the temperature numerical data, a temperature dataset is obtained; By recording the stable reading time of the weighing sensor corresponding to the weight value as a weight timestamp, and binding the weight timestamp with the weight value, a weight dataset is obtained; By comparing the image timestamp, the temperature timestamp, and the weight timestamp, the time difference between the acquisition times of the three is calculated. When the time difference exceeds the preset synchronization tolerance, it is marked as an acquisition asynchronous state, and the data synchronization verification result is obtained. When the data synchronization verification result is that the acquisition is out of sync, the image data and temperature data are corrected by time axis interpolation based on the weight timestamp. The corrected data and weight data are aligned to the same time reference point to obtain the time-aligned three-source sensing data. By packaging and binding the three-source sensing data with the weight timestamp, a unified time-series label is applied to the multi-dimensional sensing data to obtain a complete multi-dimensional sensing dataset.

4. The multi-dimensional parameter monitoring method for food product quality control according to claim 1, characterized in that, S2 includes: Standardized plating images are obtained by standardizing the image data in the complete multi-dimensional perception dataset. The position, outline, and coverage area of ​​food on the plate are extracted using the standardized plating image to obtain the geometric feature information of the plating. By combining the geometric feature information of the plating with the standardized plating image, the categories of different ingredients are identified, and the distribution area of ​​each type of ingredient on the plate and the positional relationship between them are determined, thereby obtaining information on the categories and distribution relationships of ingredients. The geometric features of the plating and the information on the categories and distribution of the ingredients are compared item by item with the position requirements, area ratio requirements and shape requirements in the preset plating standard library to obtain the comparison results of each indicator and a list of non-compliance items. The plating quality is determined by comparing the results of various indicators and reviewing the list of non-conformities.

5. The multi-dimensional parameter monitoring method for food product quality control according to claim 4, characterized in that, The process of identifying different food categories by combining the geometric feature information of the plating with the standardized plating image, and determining the distribution area of ​​each type of food on the plate and the positional relationship between them, to obtain information on food categories and distribution relationships, includes: By comparing the food areas in the standardized plating image with a preset food color feature library pixel by pixel, the pixel areas belonging to various types of food are initially classified and marked to obtain a color-based food area classification map. By combining the food region classification map with the contour boundaries in the plating geometric feature information to perform region connectivity analysis, discrete pixels are merged or reclassified to obtain a corrected food region distribution map. By mapping the pixel regions of various ingredients in the food distribution map to the plate coordinate system, the centroid coordinates, distribution area and coverage of each type of ingredient on the plate are determined, and the independent spatial distribution parameters of each type of ingredient are obtained. By comparing the independent spatial distribution parameters of each type of food in pairs, the centroid distance, adjacent boundary length, and whether there is an overlapping area between any two types of food are determined, and the relative positional relationship between the food is recorded. By comparing the recorded relative positional relationships between the ingredients with a preset food category standard ingredient distribution template, the system identifies whether there are any abnormal layout deviations between different types of ingredients, and obtains the ingredient layout deviation identification results. By summarizing and integrating the independent spatial distribution parameters with the results of the food layout deviation identification, a complete description is given of the name, area ratio, spatial location, and relationship with adjacent food items for each type of food, thus obtaining information on food categories and distribution relationships.

6. The multi-dimensional parameter monitoring method for food product quality control according to claim 1, characterized in that, S3 includes: By judging the plating quality assessment results for the condition compliance, temperature compliance, and weight compliance, the plating quality assessment results, temperature compliance information, and weight compliance information are obtained. By comprehensively considering the results of the plating quality assessment, the temperature compliance information, and the weight compliance information, the overall evaluation of the food's service conditions is conducted, resulting in a comprehensive evaluation result of the service conditions and detailed evaluation records of each indicator. Based on the comprehensive evaluation results of the meal preparation conditions, corresponding meal preparation instructions are generated according to the three levels.

7. The multi-dimensional parameter monitoring method for food product quality control according to claim 6, characterized in that, The process involves comprehensively evaluating the food's performance against the conditions for serving, based on the results of plating quality assessment, temperature compliance, and weight compliance. This yields a comprehensive evaluation result and detailed records of each indicator, including: By comparing the various deviation records in the plating quality judgment results with the preset tolerable deviation upper limit, the plating deviation is confirmed item by item to determine whether it is within the allowable range, and a classification record of the severity of plating deviation is obtained. By comparing the temperature deviation direction in the temperature compliance and deviation direction records with a preset deviation consequence comparison table, the food safety risk or taste deterioration risk is qualitatively assessed, and a temperature deviation risk level record is obtained. By comparing the degree of overweight or underweight in the weight compliance and deviation direction records with the preset component deviation impact standard, the impact of cost deviation or consumer satisfaction is evaluated, and a component deviation impact level record is obtained. By aggregating the severity grading records of the plate placement deviation, the risk level records of the temperature deviation, and the impact level records of the component deviation, the highest risk level among the various indicators is identified. When the highest risk level reaches a preset unacceptable threshold, a veto signal is triggered, and the veto trigger signal and the corresponding veto reason record are obtained. When the rejection trigger signal is in the triggered state, the comprehensive evaluation result is set to the unqualified level, and the rejection reason is recorded as the main output of the comprehensive evaluation result of the meal preparation conditions, so as to obtain the rejection-type comprehensive evaluation result of the meal preparation conditions. When the rejection trigger signal is in an untriggered state, the compliance status, deviation degree, and risk level of the three indicators of plating, temperature, and weight are summarized, and the overall qualification degree of each indicator is described to obtain a comprehensive evaluation result of the comprehensive meal preparation conditions and a detailed evaluation record of each indicator.

8. The multi-dimensional parameter monitoring method for food product quality control according to claim 1, characterized in that, S4 includes: By determining whether the food delivery instruction is to prohibit food delivery, when the food delivery instruction is to prohibit food delivery, an abnormal recording process is triggered, marking all sensory data and evaluation results of the food and archiving and saving them to obtain an abnormal data archive record. When the food preparation instruction allows direct food preparation or allows food preparation with conditions, the weight values ​​and food categories and distribution information in the complete multi-dimensional perception dataset with time stamps are retrieved, and the calorie reference values ​​are determined by comparing them with the preset food category and calorie correspondence, thus obtaining the initial calorie reference information. By retrieving the temperature values ​​from the complete multi-dimensional sensing dataset with time stamps and combining them with the cooking method of the food, the initial heat reference information is adjusted, and the heat information is dynamically corrected to obtain the adjusted heat information. When the food delivery instruction is conditionally permitted to be delivered, the food temperature is re-collected after a set time interval by starting a delayed verification process. The re-collected temperature value is compared with the safe temperature range to observe whether the temperature is within the safe range and record the temperature changes, thus obtaining a temperature verification record after food delivery. By integrating the complete multi-dimensional sensing dataset with time stamps, the plating quality judgment results and deviation records, the meal serving instructions, the adjusted calorie information, and the post-meal temperature verification records into a complete data package and uploading it to the backend management system, parameter monitoring and optimization information is obtained.

9. The multi-dimensional parameter monitoring method for food product quality control according to claim 8, characterized in that, The process involves retrieving temperature values ​​from the complete multi-dimensional sensing dataset with time stamps and combining this with the food's cooking method to adjust the initial calorie reference information, dynamically correcting the calorie information, and obtaining the adjusted calorie information. This includes: By comparing the temperature values ​​in the complete multi-dimensional sensing dataset with time stamps with preset low temperature range, normal temperature range and high temperature range, the temperature state category to which the current temperature of the food belongs is determined, and the food temperature state category identifier is obtained. By identifying the glossy areas on the food surface using image data from the complete multi-dimensional perception dataset, the degree of grease adhesion on the food surface is determined. When the gloss coverage exceeds a preset threshold, it is marked as a high-grease state, thus obtaining a grease state identifier for the food surface. By combining the food temperature state category identifier and the food surface oil state identifier, the cooking method of the food is inferred in reverse to obtain the inferred food cooking method category; By matching the inferred food cooking method category with a preset cooking method calorie correction coefficient lookup table, the calorie correction coefficient corresponding to the cooking method is obtained, and the cooking method correction coefficient is obtained. By combining the initial calorie reference information with the cooking method correction coefficient, the calorie reference value is dynamically adjusted to obtain the basic calorie value; By combining the baseline calorie value with the temperature correction factor in the food temperature state category identifier, the calorie differences caused by heat conduction and dissipation at different consumption temperatures of the same food are compensated and adjusted to obtain the adjusted calorie information.

10. A multi-dimensional parameter monitoring system for food product quality control, characterized in that, The system includes: The multi-dimensional perception module is used to collect various types of information from the food at the food preparation station through information acquisition devices to obtain multi-dimensional perception data of the food. The quality judgment module is used to perform image analysis on the food through the multi-dimensional perception data and compare it with the preset plating standards to obtain the plating quality judgment result. The comprehensive analysis module is used to comprehensively analyze and evaluate the food's serving quality status by combining the multi-dimensional perception data with the plating quality judgment results, obtain the serving condition satisfaction evaluation results, and generate serving instructions based on the serving condition satisfaction evaluation results. The analysis and adjustment module is used to dynamically analyze and adjust food parameters by combining the multi-dimensional sensing data with the meal preparation instructions, and to obtain parameter monitoring optimization information.

Images