A method for matching preservation parameters based on visual recognition of fruits and vegetables
By acquiring images of fruits and vegetables before packaging using fruit and vegetable visual recognition technology, identifying and segmenting fruit and vegetable areas, and correcting gas and water vapor exchange conditions, the deviation problem of matching packaging parameters for fresh-cut fruits and vegetables in existing technologies is solved, enabling more accurate selection of packaging parameters and ensuring gas and water vapor requirements during storage.
Patent Information
- Application Number
- CN202611263694.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-19
- Publication Date
- 2026-11-17
AI Technical Summary
Existing methods for matching preservation packaging parameters are insufficient to effectively address the differences in cut surface, morphology, and maturity of fresh-cut fruits and vegetables. This leads to discrepancies between the membrane material's permeability conditions and the actual oxygen consumption, carbon dioxide production, and water vapor release, making it impossible to accurately select suitable packaging parameters.
By combining fruit and vegetable visual recognition technology, object recognition and region segmentation are performed by acquiring images before packaging to generate visual preservation status records. Based on the cut surface load, abnormal cut surfaces, and leakage level, gas and water vapor exchange conditions are corrected to form gas exchange constraints and water vapor exchange constraints, and candidate packaging records that meet storage requirements are selected.
It enables matching appropriate packaging parameters according to the specific condition of fruits and vegetables, ensuring simultaneous coverage of gas exchange and moisture release needs within the packaging, avoiding condensation or leakage problems caused by insufficient moisture release capacity, and improving the accuracy of packaging effect.
Smart Images

Figure CN123117625A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of food preservation and packaging technology, and in particular to a method for matching preservation parameters by combining visual recognition of fruits and vegetables. Background Technology
[0002] Fresh-cut fruits and vegetables are typically washed, peeled, deseeded, cut into chunks, slices, shreds, or mixed cuts in central kitchens, pre-processed vegetable workshops, and supermarkets before being sold under refrigeration in trays covered with film or bags. This fresh-cutting process exposes the cut surface, and the respiration, moisture loss, browning, and mechanical damage of the cut tissue affect the oxygen, carbon dioxide, and moisture environment inside the packaging. Therefore, the packaging film material, film specifications, permeability level, and packaging size must match the pre-packaging condition of the product.
[0003] Existing methods for matching preservation packaging parameters typically use fruit and vegetable type, packaging quality, storage temperature, target storage duration, and packaging form as inputs, and then query a parameter database for film material, permeability rating, tray specifications, or bag specifications. Some methods further incorporate maturity level, preset respiration rate, or empirical correction parameters to determine the gas exchange capacity required for modified atmosphere packaging. While these methods can select packaging parameters based on product type and storage conditions, under the same product type and packaging quality, the cut surface area, number of cut surfaces, thickness distribution, and skin retention ratio of block, sheet, shredded, irregularly cut, and mixed cut products vary significantly. When simply using the same packaging parameters based on product type and weight, discrepancies can easily arise between the film material permeability conditions and the actual oxygen consumption, carbon dioxide production, and water vapor release status of fresh-cut products.
[0004] Another approach uses visual recognition to identify the type, maturity, appearance, browning, or mechanical damage of fresh-cut fruits and vegetables, outputting category, grade, or abnormal label. While this type of recognition reflects the product's appearance, it typically focuses on quality classification or rejection, and struggles to convert fresh, abnormal, and seepage areas into criteria for gas permeability, moisture permeability, and effective exchange area. When packaging parameter libraries primarily focus on oxygen and carbon dioxide permeability, candidate packaging with insufficient moisture release capacity may still be retained, leading to condensation inside the bag, surface seepage accumulation, or misjudgment of appearance. Post-packaging temperature, humidity, or gas monitoring can detect internal packaging anomalies, but by then the film material and packaging specifications are already determined. The process usually involves alarms, rework, or repackaging, lacking a continuous matching relationship between pre-packaging identification results and preservation parameter selection. Therefore, a preservation parameter matching method combining visual recognition of fruits and vegetables is urgently needed to address these issues. Summary of the Invention
[0005] This invention provides a method for matching preservation parameters by combining visual recognition of fruits and vegetables, solving the problem that the visual state of fresh-cut fruits and vegetables is difficult to convert into a basis for screening preservation packaging parameters.
[0006] This invention provides a method for matching preservation parameters based on visual recognition of fruits and vegetables, comprising: Acquire images of fresh-cut fruits and vegetables before packaging and packaging input information for a batch of fresh-cut fruits and vegetables to be packaged. The packaging input information includes packaging quality, storage temperature, target storage time and packaging form. The packaging quality is the net content of fresh-cut fruits and vegetables in a unit package. The image before packaging is subjected to object recognition and region segmentation to obtain the fruit and vegetable category, cutting shape, maturity level, and region recognition results of non-cut surface tissue area, fresh cut surface area, abnormal cut surface area and seepage area; Based on the region identification results and packaging quality, a visual preservation status record is generated, which includes the cut surface load level, abnormal cut surface level, cut object density level, and leakage status level. Based on the fruit and vegetable category, cutting shape, maturity level, visual preservation status record and packaging input information, the preservation parameter library is called to obtain candidate packaging records with gas permeability level, water vapor permeability level, packaging size conditions and effective exchange area range. Based on the fruit and vegetable category and storage temperature, the basic gas exchange conditions are read. According to the increasing state of the cut surface load level, maturity level and abnormal cut surface level, the basic gas exchange conditions are corrected to a higher gas exchange capacity level to obtain gas exchange constraints. The applicable storage time recorded in the candidate packaging is used to cover the target storage time as the retention condition. Based on the fruit and vegetable category, packaging form and storage temperature, the basic water vapor exchange conditions are read. According to the increasing state of the cross-sectional load level and the seepage state level, the basic water vapor exchange conditions are corrected to a higher water vapor exchange capacity level to obtain the water vapor exchange constraint. The effective gas exchange area and effective water vapor exchange area are determined according to the packaging size and packaging form of the candidate packaging records, and the gas exchange capacity level and water vapor exchange capacity level of the candidate packaging records are formed in combination with the packaging quality. The gas exchange capacity level and water vapor exchange capacity level are compared with the gas exchange constraint and water vapor exchange constraint, respectively, to form a common feasible candidate set. Output matching results when the common feasible candidate set is not empty, and output packaging condition adjustment information when the common feasible candidate set is empty.
[0007] In some embodiments, the fresh-cut fruit and vegetable object area is the image area occupied by the fresh-cut fruit and vegetable object identified in the pre-packaging image; the cut surface load level is determined by the ratio of the area of the fresh cut surface area to the area of the fresh-cut fruit and vegetable object area; the abnormal cut surface level is determined by the ratio of the area of the abnormal cut surface area to the area of the fresh cut surface area or the area of the fresh-cut fruit and vegetable object area; the cut object density level is determined by the number of cut objects, the edge length density, or the image connectivity density corresponding to the packaging quality; and the leakage state level is determined by the ratio of the leakage area to the area of the fresh-cut fruit and vegetable object area or the area of the packaging-bearing area.
[0008] In some embodiments, object recognition and region segmentation are performed on the pre-packaging image, including: Identify each fresh-cut fruit and vegetable object in the pre-packaging image and establish object identifiers. Create object outlines for the fresh-cut fruit and vegetable objects in the pre-packaging image and use the area enclosed by the object outlines as the fresh-cut fruit and vegetable object area. Within the area of the fresh-cut fruits and vegetables, image units are divided according to pixel blocks or superpixel blocks, and the continuous texture of the epidermis, wax reflection, outer surface texture of leaf veins, cutting edge, tissue fiber texture, internal color distribution and color gradient change are extracted from the image units. For each fresh-cut fruit and vegetable object, areas with continuous epidermal texture, leaf vein outer surface texture, waxy reflection, or continuous epidermal color bands are selected as non-cut surface tissue candidate areas. Regions with at least two of the following characteristics—cutting edge, internal tissue fiber texture, non-epidermal color distribution, and internal color gradient—and whose epidermal texture response value is below the corresponding threshold are considered as fresh cut surface candidate regions. The fresh cut surface discrimination value is calculated by weighting the inverse values of the cutting edge response value, tissue fiber texture response value, internal color gradient response value and epidermal texture response value. Based on the fresh cut surface discrimination value and epidermal texture response value, the image regions belonging to the same fresh cut fruit and vegetable object are divided into non-cut surface tissue regions and fresh cut surface regions.
[0009] In some embodiments, object recognition and region segmentation of the pre-packaging image further includes: Continuous image units whose fresh cut discrimination value reaches the fresh cut discrimination threshold and whose epidermal texture response value is lower than the epidermal texture preservation threshold are merged into fresh cut regions, and continuous image units whose epidermal texture response value reaches the epidermal texture preservation threshold and whose fresh cut discrimination value is lower than the fresh cut threshold are merged into non-cut tissue regions. For a boundary image unit that simultaneously presents cutting edge features and epidermal texture features, its affiliation is determined based on the distance between the boundary image unit and the outer contour of the object, the category of adjacent regions, and the direction of edge continuity. In the fresh cut area, browning and mechanical damage areas are identified, and the union of browning and mechanical damage areas is taken as the abnormal cut area. When the two overlap, the overlapping image unit is counted once. The seepage area can be identified by collecting changes in the reflective liquid area, the continuous liquid coverage area, the connection relationship between the liquid area and the edge of the fresh cut area, or by intervals. Generate a record linking the region category, region range, and object identifier.
[0010] In some embodiments, the candidate packaging records in the preservation parameter library include fruit and vegetable category identifiers, cutting morphology identifiers, maturity level applicable conditions, state applicable conditions, temperature applicable conditions, packaging form conditions, packaging quality range, applicable storage duration range, packaging size conditions, film material, film specifications, gas permeability level, water vapor permeability level, and effective exchange area range; the gas permeability level includes oxygen permeability level and carbon dioxide permeability level; the state applicable conditions include cut surface load level applicable conditions, abnormal cut surface level applicable conditions, cut object density level applicable conditions, and seepage state level applicable conditions.
[0011] In some embodiments, determining the effective gas exchange area and effective water vapor exchange area according to the packaging size conditions and packaging form recorded in the candidate packaging includes: When the packaging is a tray with film, the effective gas exchange area and the effective water vapor exchange area are determined according to the film-covered area of the tray opening, the ventilation window area, and the overlapping area of the two. When the packaging is a bag, the effective gas exchange area and the effective water vapor exchange area are determined according to the area of the bag that participates in the exchange within the unfolded area. The effective gas exchange area is related to the gas permeability level and the packaging quality to form the gas exchange capacity level, and the effective water vapor exchange area is related to the water vapor permeability level and the packaging quality to form the water vapor exchange capacity level.
[0012] In some embodiments, a state monotonic consistency check is performed on the candidate packaging record group in the preservation parameter library before a common feasible candidate set is formed. In the candidate packaging record group, except for the state level to be verified, the other fruit and vegetable categories, cutting shapes, maturity levels, packaging forms, packaging size conditions, packaging quality ranges, and applicable temperature conditions are the same; the state monotonicity consistency verification is carried out by comparing discrete state levels, and candidate packaging records with increased cut surface load level and decreased gas exchange capacity level, increased leaching state level and decreased water vapor exchange capacity level, and increased abnormal cut surface level and extended applicable storage time are excluded.
[0013] In some embodiments, when the common feasible candidate set is not empty, the sorting result is determined according to the applicable storage duration coverage margin, packaging size adaptation relationship and effective exchange area margin, and the candidate packaging records in the sorting result are determined as the matching result; the matching result includes membrane material, membrane specifications, gas permeability grade, water vapor permeability grade, packaging size and storage parameters.
[0014] In some embodiments, when the common feasible candidate set is empty, packaging condition adjustment information is generated based on the constraint categories that are not met by the candidate packaging records; When gas exchange or water vapor exchange conditions are not met, adjustment items are generated to reduce packaging quality, increase effective exchange area, or change packaging form; when packaging size conditions are not met, adjustment items are generated to adjust the range of processable packaging specifications or change packaging form; when target storage time conditions are not met, adjustment items are generated to shorten target storage time, package in batches according to abnormal cut surface grade, or remove fresh-cut fruits and vegetables whose abnormal cut surface grade does not meet the conditions; the preservation parameter library is called back based on the adjustment items, and if the common feasible candidate set is still empty after the recall, manual verification information is output.
[0015] In some embodiments, the method further includes determining the reliability level of the region identification result and establishing a parameter matching traceability record; when the region identification confidence information does not reach the preset reliability level of the corresponding region category, the state level of the region involved in the generation and the adjacent state levels are jointly recorded in the visual preservation state record to form a combined state applicable interval, and the state applicable conditions of the candidate packaging record cover the combined state applicable interval as the retention condition of the candidate packaging record; the parameter matching traceability record includes the image identifier of the image before packaging, packaging input information, region identification result, visual preservation state record, preservation parameter library version, candidate packaging record identifier, constraint comparison result, exclusion reason, matching result or packaging condition adjustment information.
[0016] Through the above technical solution, the present invention can achieve at least the following beneficial effects: This invention acquires pre-packaging images and packaging input information of batches of fresh-cut fruits and vegetables to be packaged, and generates a visual preservation status record from non-cut surface tissue areas, fresh cut surface areas, abnormal cut surface areas, and seepage areas. The cut surface load level, abnormal cut surface level, cut piece quantity level corresponding to unit packaging weight, and seepage status level serve as the criteria for screening preservation parameter libraries. This allows differences in block, slice, fibrous, irregular, and mixed cut pieces under the same category and packaging weight to be converted into packaging parameter matching conditions. Therefore, the determination of film material, film specifications, and packaging dimensions can correspond to the cut surface exposure state and surface seepage state of fresh-cut fruits and vegetables before packaging.
[0017] This invention establishes gas exchange constraints and water vapor exchange constraints based on visual preservation status records, and correlates the effective exchange area of candidate packaging records with gas permeability and water vapor permeability levels to form gas exchange capacity levels and water vapor exchange capacity levels. Candidate packaging records must simultaneously meet both gas exchange and water vapor exchange constraints to enter the common feasible candidate set, thus excluding packaging records that only meet oxygen and carbon dioxide exchange conditions but have insufficient water vapor release capacity. This ensures that the screening results simultaneously cover the gas exchange and water vapor release requirements within the packaging.
[0018] In the process of generating gas exchange constraints, this invention converts the increased states of cut surface load level, maturity level, and abnormal cut surface level into higher gas exchange capacity levels, and limits the target storage duration that candidate packaging records can cover based on the abnormal cut surface level. Therefore, gaseous environment disturbances caused by browning, mechanical damage, or tissue breakage are incorporated into the candidate packaging record screening process. When the target storage duration exceeds the coverage range of the abnormal cut surface state, a packaging condition adjustment process is initiated, ensuring that the packaging parameter matching results correspond to the current state of the fresh-cut fruits and vegetables.
[0019] This invention divides non-cut tissue regions and fresh-cut regions using continuous epidermal texture, cutting edge, tissue fiber texture, internal color gradient, and epidermal texture inverse response, and forms a combined state applicable range for low-reliability regions. Candidate packaging records are retained based on coverage of the combined state applicable range. Even when there is partial occlusion, water droplet reflection, or undetermined areas in the image, the region identification result can still enter the screening process with a definite state range. When fresh-cut regions and seepage areas cannot be distinguished, the processing information of the image before re-acquiring packaging is output, ensuring that erroneous region identification results do not participate in the determination of membrane material and membrane specifications. Attached Figure Description
[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope of the present invention.
[0021] Figure 1 This is a flowchart of the preservation parameter matching method that combines visual recognition of fruits and vegetables in the embodiment. Detailed Implementation
[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0023] All terms used in this invention (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined.
[0024] Example 1: like Figure 1 As shown, this embodiment employs a preservation parameter matching method that combines visual recognition of fruits and vegetables, including: Step S1: Obtain the pre-packaging image and packaging input information of the batch of fresh-cut fruits and vegetables to be packaged. The packaging input information includes packaging quality, storage temperature, target storage time and packaging form. Specifically, the batch of fresh-cut fruits and vegetables to be packaged refers to a group of fresh-cut fruits and vegetables that have been cut, drained, and weighed before entering the packaging station. The image before packaging is a visible light image of this batch taken before packaging, and the image identifier is linked to the batch identifier, weighing time, and acquisition time.
[0025] When multiple images exist for the same batch, the image with the acquisition time closest to the weighing time and that completely covers the packaging area is used as the main image, and the remaining images are used as supplementary images. For example, the image before packaging is acquired within 5 minutes after weighing, at a acquisition distance of 30 cm to 80 cm, with an image resolution of no less than 1280×720 and an illumination intensity of 300 lx to 1200 lx.
[0026] The packaging weight is the net content of fresh-cut fruits and vegetables within a unit package, read by an electronic scale at the packaging station, recorded in grams with a resolution of 1 g to 5 g; the storage temperature is recorded by a set value at the refrigeration station or by a near-end temperature sensor, recorded in degrees Celsius; the target storage time is calculated from the time of packaging completion and recorded in hours; the packaging form is either a tray with film or a bag, and is associated with the corresponding range of processable packaging sizes.
[0027] Step S2: Perform object recognition and region segmentation on the pre-packaging image to obtain the region recognition results of fruit and vegetable category, cutting shape, maturity level, non-cut surface tissue area, fresh cut surface area, abnormal cut surface area and seepage area; Among them, fresh-cut fruits and vegetables are fruit and vegetable cutting units with independent outer contours in the pre-packaging image and participating in the same packaging batch counting; the packaging carrying area is the image area used to carry fresh-cut fruits and vegetables in the bottom surface of the tray, the carrying surface of the conveyor belt, or the unfolded carrying surface of the packaging bag.
[0028] Fruit and vegetable categories are determined by the color distribution, texture distribution, and outline features of fresh-cut fruits and vegetables; cutting morphology is determined by the outline regularity, aspect ratio, number density, and cutting edge distribution of fresh-cut fruits and vegetables; maturity level is determined by the surface color, cut surface color distribution, and texture variation under the corresponding fruit and vegetable category.
[0029] The region identification result includes at least the region category, region extent, object identifier, and region identification confidence information. The region extent is recorded using pixel area, circumscribed contour, and connected component number. The region identification confidence information ranges from 0 to 1 and is obtained by weighting the object identification confidence, region segmentation boundary confidence, and region category discrimination confidence. The reliability level threshold and unusable level threshold are determined by a validation set of labeled samples under the same acquisition illumination conditions. For example, a region identification confidence information of 0.75 is recorded as a reliable level, below 0.75 but not below 0.50 is recorded as a low reliability level, and below 0.50 is recorded as an unusable level. In other embodiments, the reliability level threshold can be 0.70–0.85, and the unusable level threshold can be 0.40–0.60. If the threshold is too high, more normal images will enter the low reliability processing; if the threshold is too low, the probability of missegmentation results of the cross-sectional region and the seepage region participating in subsequent parameter selection increases.
[0030] Based on the regional identification results and packaging quality, a visual preservation status record is generated, which includes the cut surface load level, abnormal cut surface level, cut object density level, and liquid seepage status level. Based on the fruit and vegetable category, cutting shape, maturity level, visual preservation status record and packaging input information, the preservation parameter library is called to obtain candidate packaging records with gas permeability level, water vapor permeability level, packaging size conditions and effective exchange area range. Step S3: Read the basic gas exchange conditions based on the fruit and vegetable category and storage temperature, and correct the basic gas exchange conditions according to the cross-sectional load level, maturity level and abnormal cross-sectional level to obtain the gas exchange constraints. In one example, the standard storage duration is the upper limit of the applicable storage duration under the corresponding fruit and vegetable category, storage temperature, baseline cutting morphology, and baseline maturity level in the preservation parameter library. When the abnormal cut level is low, the target storage duration is allowed to be the standard duration; when the abnormal cut level is medium, the target storage duration is allowed to not exceed 80% of the standard duration; when the abnormal cut level is high, the target storage duration is allowed to not exceed 50% of the standard duration. In other embodiments, the proportion corresponding to medium abnormal cuts can be 60% to 90%, and the proportion corresponding to high abnormal cuts can be 30% to 70%. If the proportion is too high, the gas disturbance inside the packaging caused by mechanical damage and browning cannot be absorbed by the storage duration limit; if the proportion is too low, a large number of batches that can still be packaged and sold will enter the adjustment process.
[0031] Based on the fruit and vegetable category, packaging form and storage temperature, the basic water vapor exchange conditions are read, and the basic water vapor exchange conditions are corrected according to the cross section load level and the seepage state level to obtain the water vapor exchange constraint. Among them, when the cross section load level and the seepage state level increase, the water vapor exchange constraint moves to a higher water vapor exchange capacity level. Step S4: Determine the effective gas exchange area and effective water vapor exchange area according to the packaging size conditions and packaging form of the candidate packaging records, and combine them with the packaging quality to form the gas exchange capacity level and water vapor exchange capacity level of the candidate packaging records. Step S5: Compare the gas exchange capacity level and water vapor exchange capacity level with the gas exchange constraint and water vapor exchange constraint respectively to form a common feasible candidate set; output the matching result when the common feasible candidate set is not empty, and output the packaging condition adjustment information when the common feasible candidate set is empty.
[0032] Furthermore, the fresh-cut fruit and vegetable object area is the image area occupied by the fresh-cut fruit and vegetable objects identified in the pre-packaging image; the visual preservation status record includes the cut surface load level, abnormal cut surface level, cut object density level, and leakage status level; the cut surface load level is determined based on the occupancy relationship between the fresh cut surface area and the fresh-cut fruit and vegetable object area, the abnormal cut surface level is determined based on the occupancy relationship between the abnormal cut surface area and the fresh cut surface area and the fresh-cut fruit and vegetable object area, the cut object density level is determined by the number of cut objects corresponding to the packaging quality, the cut edge length density, or the density of the image connected region, and the leakage status level is determined based on the occupancy relationship between the leakage area and the fresh-cut fruit and vegetable object area and the packaging-bearing area.
[0033] The segmentation thresholds for each status level are determined by historical packaging records or calibration samples under the corresponding fruit and vegetable category, cutting form, and packaging form. The following values are example values under the same calibration batch.
[0034] Furthermore, the cut surface load level, abnormal cut surface level, cut object density level, and effusion status level are all generated by summarizing the region identification results of each fresh-cut fruit and vegetable object within the same batch. The cut surface load level is determined by the ratio of the area of the fresh cut surface region to the area of the fresh-cut fruit and vegetable object; for example, a ratio below 20% indicates a low cut surface load, 20%–45% indicates a medium cut surface load, and above 45% indicates a high cut surface load. The abnormal cut surface level is determined by the ratio of the area of the abnormal cut surface region to the area of the fresh cut surface region; for example, a ratio below 5% indicates a low abnormal cut surface, 5%–15% indicates a medium abnormal cut surface, and above 15% indicates a high abnormal cut surface. The cut object density level is determined according to the cutting morphology; for block and slice products, it is determined by the ratio of the number of independent fresh-cut fruit and vegetable objects to the packaging weight; for filamentous products, it is determined by the ratio of the number of skeletal branches or the length of the cut edge to the packaging weight; and for leafy vegetables, it is determined by the density of the number of connected regions and the length of the cut edge. The leakage status level is determined by the coverage ratio of the leakage area in the fresh-cut fruit and vegetable object area and the packaging area; for example, when the leakage area coverage ratio is less than 2%, it is low leakage, 2% to 8% is medium leakage, and more than 8% is high leakage.
[0035] When a region is missing but the object outline is complete, the state level corresponding to the missing region is generated based on the median state of adjacent reliable regions and similar objects, and this state level is recorded as a low reliability level; the candidate record retention method corresponding to the low reliability level is handled according to the combined state applicable interval described later.
[0036] Furthermore, the candidate packaging records in the preservation parameter library include fruit and vegetable category identifiers, cutting morphology identifiers, applicable conditions for maturity level, applicable conditions for state, applicable temperature, packaging form conditions, packaging quality range, applicable storage duration range, packaging size conditions, film material, film specifications, gas permeability level, water vapor permeability level, and effective exchange area range. Gas permeability level includes oxygen permeability level and carbon dioxide permeability level; applicable conditions for state include applicable conditions for cut surface load level, applicable conditions for abnormal cut surface level, applicable conditions for cut object density level, and applicable conditions for leakage state level. When accessing the preservation parameter library, candidate packaging records with mismatched cutting morphology identifiers are first excluded based on fruit and vegetable category and cutting morphology. Then, candidate packaging records are filtered based on applicable conditions for maturity level, applicable conditions for state, applicable temperature, packaging form conditions, packaging quality range, packaging size conditions, and applicable storage duration range.
[0037] Specifically, candidate packaging records use the candidate packaging record identifier as a unique index and record the preservation parameter library version, calibration date, and applicable processing line identifier. Applicable conditions include cross-sectional load level, abnormal cross-sectional level, cut object density level, and leakage state level. Temperature applicable conditions, packaging weight range, applicable storage duration range, packaging size conditions, and effective exchange area range are all recorded using closed intervals, with boundary values included in the applicable range. Oxygen permeability, carbon dioxide permeability, and water vapor permeability levels use integer levels from low to high; for example, levels range from 1 to 5, with higher levels indicating higher exchange capacity per unit area. The unit for effective exchange area is cm², the unit for packaging weight is g, and the unit for applicable storage duration is h.
[0038] If a candidate packaging record is missing or exceeds the allowable range of the preservation parameter library for the applicable conditions of maturity level, applicable conditions of state, gas permeability level, water vapor permeability level, packaging size conditions, effective exchange area range, packaging quality range, or applicable storage duration range, the candidate packaging record will be marked as an invalid record and excluded during screening.
[0039] Furthermore, the effective gas exchange area and effective water vapor exchange area are determined according to the packaging size and form recorded in the candidate packaging records. When the packaging form is a tray with film, the tray opening size, ventilation window size, and sealing area size are read from the candidate packaging records. The union of the tray opening film area and the ventilation window area is used as the candidate area for exchange. If the tray opening film area and the ventilation window area overlap, the overlapping area is counted once. When the packaging form is a bag, the bag length and width, sealing width, and ventilation area size are read from the candidate packaging records. The area in the unfolded area of the bag that participates in exchange is used as the candidate area. The effective gas exchange area is used to form the gas exchange capacity level, and the effective water vapor exchange area is used to form the water vapor exchange capacity level. When the two are the same, they are recorded using the same area; when they are different, they are recorded separately.
[0040] Furthermore, before forming a common feasible candidate set, a state monotonicity consistency check is performed on the candidate packaging record groups in the preservation parameter library. Except for the state level to be checked, the candidate packaging record groups share the same fruit and vegetable category, cutting shape, maturity level, packaging form, packaging size conditions, packaging quality range, and applicable temperature conditions. The state monotonicity consistency check is performed by comparing discrete state levels, excluding candidate packaging records with increased cut surface load level and decreased gas exchange capacity level, increased leaching state level and decreased water vapor exchange capacity level, and increased abnormal cut surface level and extended applicable storage time. Candidate packaging records that pass the state monotonicity consistency check proceed to subsequent screening.
[0041] Furthermore, when the common feasible candidate set is empty, packaging condition adjustment information is generated based on the constraint categories that are not met in the candidate packaging records. If the gas exchange condition or water vapor exchange condition is not met, adjustment items are generated to reduce packaging quality, increase the effective exchange area, or change the packaging form; if the packaging size condition is not met, adjustment items are generated to adjust the range of processable packaging specifications or change the packaging form; if the target storage time condition is not met, adjustment items are generated to shorten the target storage time, package in batches according to the abnormal cut surface level, or remove fresh-cut fruits and vegetables whose abnormal cut surface level does not meet the condition. The preservation parameter library is called back based on the adjustment items; if the common feasible candidate set is still empty after the recall, manual review information is output.
[0042] Furthermore, this embodiment also includes determining the reliability level of the region identification results; when the region identification confidence information of non-cut surface tissue region, fresh cut surface region, abnormal cut surface region, or oozing region does not reach the preset reliability level of the corresponding region category, the state level of the region involved in the generation and the adjacent state levels are jointly recorded in the visual preservation state record to form a combined state applicable interval, and the state applicable conditions of the candidate packaging record cover the combined state applicable interval as the retention condition of the candidate packaging record; when the fresh cut surface region and the oozing region cannot be distinguished due to water droplet coverage, object overlap, or reflection of the packaging bearing area, the processing information of re-acquiring the image before packaging is output.
[0043] Furthermore, this embodiment also includes establishing a parameter matching traceability record, which includes image identifiers of the pre-packaging image, packaging input information, fruit and vegetable category, cutting shape, maturity level, region identification results, visual preservation status record, preservation parameter library version, candidate packaging record identifier, constraint comparison results corresponding to each candidate packaging record, exclusion reasons, matching results or packaging condition adjustment information; the parameter matching traceability record is stored in association according to the batch and processing order of the fresh-cut fruits and vegetables to be packaged.
[0044] In a preferred embodiment, the basic gas exchange conditions are determined by the type of fruit and vegetable and the storage temperature, and the gas exchange capacity level recorded in the candidate packaging is used as the comparison criterion.
[0045] Specifically, the basic gas exchange conditions are determined by calibration samples of the same fruit and vegetable variety at the baseline cutting morphology, baseline maturity level, and baseline storage temperature. The calibration samples include historical packaging records with known packaging quality, packaging form, film material, gas permeability rating, water vapor permeability rating, effective exchange area, and actual storage results; the actual storage results include at least one of the following at the end of the target storage period: condensation inside the packaging, browning on the outside, leakage expansion, and packaging bloating.
[0046] The number of calibration samples for each fruit and vegetable category at each storage temperature range is determined by the amount of historical data and the storage duration. For example, 30 sets of calibration samples are used to establish the basic gas exchange conditions for each fruit and vegetable category at each storage temperature range; in other embodiments, the number of calibration samples can be 20 to 100 sets. If the number of samples is too small, the basic gas exchange conditions are easily affected by differences in single batches of raw materials; if the number of samples is too large but not grouped according to the cutting morphology, the differences between block, sheet, and fibrous samples will be averaged out.
[0047] When the preservation parameter database does not contain basic gas exchange conditions corresponding to the fruit and vegetable category and storage temperature, the gas exchange conditions are marked as unmet, and the reason for not meeting them is written into the exclusion reasons; when the preservation parameter database does not contain basic water vapor exchange conditions corresponding to the fruit and vegetable category, packaging form and storage temperature, the water vapor exchange conditions are marked as unmet, and the reason for not meeting them is written into the exclusion reasons.
[0048] The gas exchange capacity level is calculated based on the gas permeability level, effective gas exchange area, and packaging weight according to the preset level mapping relationship in the preservation parameter library. The higher the level number, the higher the gas exchange capacity per unit packaging weight.
[0049] In one implementation, gas exchange capacity levels include oxygen exchange capacity levels and carbon dioxide exchange capacity levels. Oxygen exchange capacity is calculated by converting the oxygen permeability per unit area, effective gas exchange area, and packaging weight corresponding to the oxygen permeability level; carbon dioxide exchange capacity is calculated by converting the carbon dioxide permeability per unit area, effective gas exchange area, and packaging weight corresponding to the carbon dioxide permeability level; and water vapor exchange capacity is calculated by converting the water vapor permeability per unit area, effective water vapor exchange area, and packaging weight corresponding to the water vapor permeability level. The conversion relationships are as follows: , , , in, For oxygen exchange capacity, For carbon dioxide exchange capacity, For water vapor exchange capacity, This represents the oxygen permeability value per unit area corresponding to the oxygen permeability level. This represents the carbon dioxide permeability value per unit area corresponding to the carbon dioxide permeability rating. This represents the water vapor permeability value per unit area corresponding to the water vapor permeability level. For the effective gas exchange area, For the effective water vapor exchange area, For packaging quality, the preservation parameter library is divided according to a preset segmentation table. , and These are respectively mapped to oxygen exchange capacity level, carbon dioxide exchange capacity level, and water vapor exchange capacity level.
[0050] The basic gas exchange conditions include oxygen exchange constraints and carbon dioxide exchange constraints. The gas exchange dimension in the following formula represents either the oxygen exchange dimension or the carbon dioxide exchange dimension. Both use the same correction structure and are compared with their corresponding exchange capacity levels. The basic gas exchange conditions are expressed as follows: , in, For fruits and vegetables And the storage temperature is The corresponding basic gas exchange conditions; The types of fresh-cut fruits and vegetables to be packaged in batches; The storage temperature for batches of fresh-cut fruits and vegetables to be packaged; For fruits and vegetables And the storage temperature is The lower limit of the basic gas exchange capacity level corresponding to that time; For fruits and vegetables And the storage temperature is The corresponding upper limit of the basic gas exchange capacity level; subscript This indicates the gas exchange dimension. The superscript 0 indicates the basic state without level correction. The lower limit of the basic gas exchange capacity level is less than or equal to the upper limit of the basic gas exchange capacity level, and both are within the range of gas exchange capacity levels allowed by the preservation parameter library.
[0051] The cut surface load level, maturity level, and abnormal cut surface level correspond to preset gas exchange correction levels. These correction levels are determined by historical packaging records or calibration samples of the same fruit and vegetable category, cut shape, and storage temperature. An increase in the cut surface load level indicates an increase in the proportion of fresh cut surface exposed per unit package weight, shifting the basic gas exchange conditions to a higher gas exchange capacity level. An increase in the maturity level indicates enhanced respiration and metabolic release from fruit and vegetable tissues, shifting the basic gas exchange conditions to a higher gas exchange capacity level according to the preset maturity level. An increase in the abnormal cut surface level indicates increased disturbance to the gas environment within the packaging due to browning, mechanical damage, or tissue breakage, shifting the basic gas exchange conditions to a higher gas exchange capacity level and limiting the applicable target storage duration.
[0052] The correction for gas exchange level is determined by the following formula: , in, This is a correction amount for the gas exchange level; For section load rating; In the fruit and vegetable category The segmentation form is And the storage temperature is At that time, the gas exchange correction level is determined by the section load level; Maturity level; In the fruit and vegetable category The segmentation form is And the storage temperature is At that time, the gas exchange correction level is determined by the maturity level; The level of abnormal section; In the fruit and vegetable category The segmentation form is And the storage temperature is At that time, the gas exchange correction level is determined by the abnormal section level; The cutting form of fresh-cut fruits and vegetables to be packaged in batches; and The aforementioned definition will be used.
[0053] Each gas exchange correction level is an integer level greater than or equal to 0. The upper and lower limits of the corrected gas exchange constraints are saturated and limited by the highest gas exchange capacity level in the preservation parameter library.
[0054] The gas exchange correction level and the water vapor exchange constraint correction caliber listed alongside it can be determined according to Table 1.
[0055] Table 1: The basic water vapor exchange conditions are determined by the fruit and vegetable variety, packaging form, and storage temperature, and the water vapor exchange capacity level recorded in the candidate packaging is used as the comparison criterion. The range of basic water vapor exchange capacity levels is determined by historical packaging records or calibration samples of the same fruit and vegetable variety, the same packaging form, and the same storage temperature.
[0056] In one implementation, the basic water vapor exchange conditions are represented as a range of water vapor exchange capacity levels. The cross-sectional load level and the seepage state level correspond to water vapor exchange correction levels, and the sum of these two levels yields the water vapor exchange capacity correction amount. The upper and lower limits of the water vapor exchange constraint are obtained by adding the upper and lower limits of the basic water vapor exchange conditions to the water vapor exchange capacity correction amount, and are further saturated by the highest water vapor exchange capacity level in the preservation parameter library. When the water vapor exchange capacity level of a candidate packaging record falls within the water vapor exchange constraint, the candidate packaging record is determined to satisfy the water vapor exchange constraint.
[0057] As the cross-sectional load level increases, the basic water vapor exchange conditions shift to a higher water vapor exchange capacity level; as the seepage state level increases, the basic water vapor exchange conditions shift further to a higher water vapor exchange capacity level. The revised upper and lower limits of water vapor exchange constraints are saturated and limited by the highest water vapor exchange capacity level in the preservation parameter library.
[0058] The water vapor exchange capacity level is calculated by combining the water vapor permeability level, effective water vapor exchange area, and packaging mass of the candidate packaging record. Under the same water vapor permeability level, the larger the effective water vapor exchange area and the smaller the packaging mass, the higher the water vapor exchange capacity level corresponding to the unit packaging mass.
[0059] When a candidate packaging record simultaneously meets the constraints of gas exchange, water vapor exchange, applicable state conditions, applicable temperature conditions, packaging form conditions, packaging quality range, packaging size conditions, and applicable storage duration range, it is included in the common feasible candidate set; if any condition is not met, the category of the unmet condition is written as the exclusion reason.
[0060] After completing the level correction, the gas exchange constraint is obtained as follows: , in, For the corrected gas exchange constraint; This represents the highest permissible gas exchange capacity level for the preservation parameter library; , , , and The aforementioned definition will be used.
[0061] The anomaly profile level is also used to limit the target storage duration that candidate packaging records can cover. When Less than or equal to At that time, with Storage duration as a screening criterion for candidate packaging records; when Greater than When the target storage time condition is not met, the system will mark the target storage time condition as not met and output adjustment information such as shortening the target storage time, packaging in batches according to the abnormal cut surface grade, or removing fresh-cut fruits and vegetables whose abnormal cut surface grade does not meet the condition; reducing packaging quality will not be used as an adjustment item to lift the upper limit of abnormal cut surface storage time.
[0062] in, Enter the target storage duration in the packaging information; The abnormal section level is Fruits and vegetables are And the storage temperature is The upper limit of the allowable calibrated storage time; , and The aforementioned definition will be used.
[0063] When screening candidate packaging records, the gas exchange capacity level of the candidate packaging records is compared with... In comparison, the gas exchange capacity level falls into The candidate packaging records satisfy the gas exchange constraints; At the same time, Less than or equal to At that time, the applicable storage duration recorded in the candidate packaging should be compared with... Comparison, cannot cover Candidate packaging records were excluded. After screening for gas exchange constraints and water vapor exchange constraints, their intersection was used as the common feasible candidate set. When the common feasible candidate set contained multiple candidate packaging records, they were first sorted in ascending order of applicable storage duration coverage margin, then in ascending order of packaging size and packaging form compatibility, and finally in ascending order of effective exchange area margin. The candidate packaging records in the sorting results were output as matching results, which included membrane material, membrane specifications, gas permeability rating, water vapor permeability rating, packaging size, and storage parameters.
[0064] For batches whose section load level, abnormal section level, or leakage state level is generated from the region identification results of low reliability level, the low reliability level processing includes three stages: state generation, candidate record retention, and exchange capability comparison. In the state generation stage, the state level corresponding to the low reliability level and its adjacent state levels are combined to form a combined state applicable range. In the candidate record retention stage, the state applicable conditions of the candidate packaging records covering the combined state applicable range are used as the retention condition. In the exchange capability comparison stage, when multiple state levels in the combined state applicable range are covered by the same candidate packaging record, the correction level corresponding to the state level with the higher exchange capability requirement is used for comparison. For batches with a maturity level at the low reliability level, the maturity level and its adjacent maturity levels are combined to form a combined maturity applicable range, and the maturity applicable conditions of the candidate packaging records covering the combined maturity applicable range are used as the retention condition. When the region identification result is an unusable level, it is not included in the combined state applicable range calculation, and processing information for re-acquiring the image before packaging is output; if it is still unusable after re-acquisition, packaging condition adjustment information is output, and the reason for exclusion is recorded as unusable region identification result.
[0065] Example 2: Based on Example 1, this example provides a specific method for object recognition and region segmentation of pre-packaging images in the preservation parameter matching method.
[0066] Specifically, each fresh-cut fruit and vegetable object in the pre-packaging image is identified and an object identifier is established. The object outline of the fresh-cut fruit and vegetable object in the pre-packaging image is established, and the area enclosed by the object outline is taken as the fresh-cut fruit and vegetable object area.
[0067] Within the fresh-cut fruit and vegetable object area, image units are divided according to pixel blocks or superpixel blocks, and the continuous texture of the epidermis, wax reflection, outer surface texture of leaf veins, cutting edge, tissue fiber texture, internal color distribution and color gradient changes are extracted from the image units.
[0068] For each fresh-cut fruit and vegetable object, regions with continuous epidermal texture, leaf vein outer surface texture, waxy reflection, or continuous epidermal color bands are selected as non-cut surface tissue candidate regions; regions with at least two of the following, namely cutting edge, internal tissue fiber texture, non-epidermal color distribution, and internal color gradient, and whose epidermal texture response value is lower than the corresponding threshold, are selected as fresh cut surface candidate regions.
[0069] The fresh cut surface discrimination value is obtained by weighting the response values of the cutting edge, tissue fiber texture, internal color gradient, and epidermal texture. Then, based on the fresh cut surface discrimination value and the epidermal texture response value, the image regions belonging to the same fresh-cut fruit and vegetable object are divided into non-cut tissue regions and fresh cut surface regions.
[0070] Object recognition and region segmentation of images before packaging also include: Continuous image units whose fresh cut discrimination value reaches the fresh cut discrimination threshold and whose epidermal texture response value is lower than the epidermal texture preservation threshold are merged into fresh cut regions, and continuous image units whose epidermal texture response value reaches the epidermal texture preservation threshold and whose fresh cut discrimination value is lower than the fresh cut threshold are merged into non-cut tissue regions. For a boundary image unit that simultaneously presents cutting edge features and epidermal texture features, its classification is determined based on the distance between the boundary image unit and the outer contour of the object, the category of adjacent regions, and the direction of edge continuity. Image units that are close to the outer contour of the object and continuous with the epidermal texture are classified into non-cut surface tissue regions, while image units that are located inside the cutting boundary and continuous with the internal tissue texture are classified into fresh cut surface regions. Image units that do not meet either the fresh cut surface discrimination threshold or the epidermal texture preservation threshold are kept as undetermined regions and merged according to the connectivity ratio of adjacent classified regions; when the area of the undetermined region exceeds the preset ratio of the area of the corresponding fresh cut fruit and vegetable object, the region recognition result of the object is marked as low reliability level. For example, the preset ratio is 15%. In other embodiments, the preset ratio can be 8% to 25%. If the preset ratio is too small, local shadows, slight occlusion, or water droplet reflections will frequently trigger a low reliability level; if the preset ratio is too large, more unclassified image units will participate in the subsequent area proportion calculation. When the unclassified area is not adjacent to the classified area, the classified image units within the fresh-cut fruit and vegetable object are merged according to their area proportion; when the area of the classified image units within the fresh-cut fruit and vegetable object is less than 60% of the area of the fresh-cut fruit and vegetable object, the object does not participate in the main statistics of the status level, but only in the low reliability level marking and traceability recording.
[0071] In the fresh cut area, identify regions where the color or texture changes continuously relative to similar fresh cut areas. Regions that meet the browning characteristics are marked as browned regions, and regions with broken edges, missing tissue, or compression deformation are marked as mechanical damage regions. Browned regions and mechanical damage regions are merged into abnormal cut areas. When browned regions and mechanical damage regions overlap, the overlapping image units are counted once in the calculation of the area of abnormal cut areas.
[0072] Identify reflective or continuously covered liquid areas on the surface of fresh-cut fruits and vegetables or in the packaging area, and determine the seepage area by combining the connection relationship between the liquid area and the edge of the fresh cut surface area, the expansion and change of the liquid area before and after interval sampling, or the residual state after re-draining; washing residual water areas located only at the edge of the packaging area and not connected to the fresh-cut fruits and vegetables are not included in the seepage area.
[0073] Generate a record linking the region category, region range, and object identifier.
[0074] In a preferred embodiment, after establishing the object outline of the fresh-cut fruit and vegetable object in the image before packaging, the area enclosed by the object outline is taken as the fresh-cut fruit and vegetable object area, and the image unit is divided into pixel blocks or superpixel blocks within the fresh-cut fruit and vegetable object area, and the image unit is used as the calculation object of the fresh cut surface discrimination value.
[0075] The fresh cut surface discrimination value is determined according to the following formula: , in, For the first Among the fresh-cut fruits and vegetables, the first Fresh section discrimination value of each image unit; The serial number of the fresh-cut fruits and vegetables; For the first Image unit number within a fresh-cut fruit and vegetable object; For the first The cutting edge weight corresponding to each fresh-cut fruit and vegetable object; For the first Among the fresh-cut fruits and vegetables, the first The cutting edge response value of each image unit; For the first The tissue fiber weights corresponding to each fresh-cut fruit and vegetable object; For the first Among the fresh-cut fruits and vegetables, the first The tissue fiber texture response value of each image unit; For the first The internal color gradient weights corresponding to each fresh-cut fruit and vegetable object. For the first Among the fresh-cut fruits and vegetables, the first Internal color gradient response value of each image unit; For the first Inverse weight of the skin texture corresponding to each fresh-cut fruit and vegetable object; For the first Among the fresh-cut fruits and vegetables, the first The epidermal texture response value of each image unit.
[0076] The above , , and All values are normalized to a range greater than or equal to 0 and less than or equal to 1. , , and All values are greater than or equal to 0 and less than or equal to 1, and the sum of all four values equals 1. A higher cutting edge response value indicates that the image unit is closer to the boundary abrupt change formed by tool cutting; a higher tissue fiber texture response value indicates that the image unit is closer to the texture of the pulp, potato pulp, melon pulp, or petiole; a higher internal color gradient response value indicates that the image unit is closer to the color distribution of the cut surface that transitions continuously from the outer edge to the inside; a higher epidermal texture response value indicates that the image unit is closer to the texture of the outer skin, epidermal wax layer, or outer surface of the leaf.
[0077] Each weight, fresh cut surface discrimination threshold, and epidermal texture preservation threshold are calibrated using a validation set of labeled samples of the same fruit and vegetable category, the same cut morphology, and the same collection lighting conditions. The overlap between the non-cut surface tissue area and the fresh cut surface area is used as the selection criterion. When changes in collection lighting cause an overall brightness shift, the background brightness of the whiteboard or packaging area is used as a reference for normalization compensation. The compensated response value is still limited to a range greater than or equal to 0 and less than or equal to 1.
[0078] For example, on the labeled sample validation set, the overlap between the non-cut tissue region and the fresh cut region is used as the parameter selection criterion. The cutting edge weight, tissue fiber weight, internal color gradient weight, and epidermal texture reverse weight are set to 0.30, 0.25, 0.25, and 0.20, respectively. The fresh cut discrimination threshold is set to 0.60, and the epidermal texture preservation threshold is set to 0.55.
[0079] In other embodiments, a single weight can be 0.10–0.50, and each weight is normalized before being used for discrimination; the fresh cut surface discrimination threshold can be 0.50–0.75, and the epidermal texture retention threshold can be 0.45–0.70. When the fresh cut surface discrimination threshold is too low, bright spots or natural color difference areas on the epidermis are easily merged into fresh cut surface areas; when the fresh cut surface discrimination threshold is too high, the edges of thin slices or the cut edges of leafy vegetables are easily kept as undetermined areas. When the epidermal texture retention threshold is too low, the proportion of residual epidermal areas is amplified; when the epidermal texture retention threshold is too high, rough flesh texture may be mistakenly classified as non-cut surface tissue areas.
[0080] For different types of fruits and vegetables, identification examples matching their appearance and structure can be used. For fresh-cut pears, the outer skin area typically has a continuous peel color, few lenticels, and a surface texture with minimal exposed internal fibers, making it a suitable non-cut surface area. The pear flesh typically has light-colored flesh, fine fibrous texture, and a color gradient extending from the cut edge inwards, making it a suitable fresh cut surface area. For potatoes, the skin-on area typically has a brown or yellow peel, rough speckled texture, and a continuous peel boundary, making it a suitable non-cut surface area. The potato flesh typically has a relatively uniform light-colored starchy texture, cut edges, and blocky color distribution inside, making it a suitable fresh cut surface area. For pumpkins, the outer skin area typically has a dark, hard peel, striped skin texture, or waxy sheen, making it a suitable non-cut surface area. The pumpkin flesh typically has orange-yellow internal tissue, fibrous bundle texture, and smooth edges formed by cutting, making it a suitable fresh cut surface area. For leafy vegetables, the outer surface of the leaf typically features continuous veins, surface wrinkles, and a natural sheen, which can be considered as the non-cut surface tissue area. The cut petiole or leaf margin usually exhibits fracture boundaries, severed veins, and exposed tissue fibers, which can be considered as the fresh cut surface area. For different fruit and vegetable categories, visual features corresponding to their outer surface tissue and internal cut tissue are selected to distinguish between the non-cut surface tissue area and the fresh cut surface area.
[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
[0082] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are meant to be within the scope of the invention and form different embodiments. For example, all the embodiments above can be used in any combination. The information disclosed in the Background section is intended only to enhance the understanding of the general background of the invention and should not be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art.
Claims
1. A method for matching preservation parameters based on visual recognition of fruits and vegetables, characterized in that, include: Acquire images of fresh-cut fruits and vegetables before packaging and packaging input information for a batch of fresh-cut fruits and vegetables to be packaged. The packaging input information includes packaging quality, storage temperature, target storage time and packaging form. The packaging quality is the net content of fresh-cut fruits and vegetables in a unit package. The image before packaging is subjected to object recognition and region segmentation to obtain the fruit and vegetable category, cutting shape, maturity level, and region recognition results of non-cut surface tissue area, fresh cut surface area, abnormal cut surface area and seepage area; Based on the region identification results and packaging quality, a visual preservation status record is generated, which includes the cut surface load level, abnormal cut surface level, cut object density level, and leakage status level. Based on the fruit and vegetable category, cutting shape, maturity level, visual preservation status record and packaging input information, the preservation parameter library is called to obtain candidate packaging records with gas permeability level, water vapor permeability level, packaging size conditions and effective exchange area range. Based on the fruit and vegetable category and storage temperature, the basic gas exchange conditions are read. According to the increasing state of the cut surface load level, maturity level and abnormal cut surface level, the basic gas exchange conditions are corrected to a higher gas exchange capacity level to obtain gas exchange constraints. The applicable storage time recorded in the candidate packaging is used to cover the target storage time as the retention condition. Based on the fruit and vegetable category, packaging form and storage temperature, the basic water vapor exchange conditions are read. According to the increasing state of the cross-sectional load level and the seepage state level, the basic water vapor exchange conditions are corrected to a higher water vapor exchange capacity level to obtain the water vapor exchange constraint. The effective gas exchange area and effective water vapor exchange area are determined according to the packaging size and packaging form of the candidate packaging records, and the gas exchange capacity level and water vapor exchange capacity level of the candidate packaging records are formed in combination with the packaging quality. The gas exchange capacity level and water vapor exchange capacity level are compared with the gas exchange constraint and water vapor exchange constraint, respectively, to form a common feasible candidate set. Output matching results when the common feasible candidate set is not empty, and output packaging condition adjustment information when the common feasible candidate set is empty.
2. The method for matching preservation parameters based on visual recognition of fruits and vegetables according to claim 1, characterized in that, The fresh-cut fruit and vegetable object area is the image area occupied by the fresh-cut fruit and vegetable objects identified in the pre-packaging image; the cut surface load level is determined by the ratio of the area of the fresh cut surface area to the area of the fresh-cut fruit and vegetable object area; the abnormal cut surface level is determined by the ratio of the area of the abnormal cut surface area to the area of the fresh cut surface area or the area of the fresh-cut fruit and vegetable object area; the cut object density level is determined by the number of cut objects, the cut edge length density, or the image connectivity density corresponding to the packaging quality; and the leakage state level is determined by the ratio of the leakage area area to the area of the fresh-cut fruit and vegetable object area or the area of the packaging-bearing area.
3. The method for matching preservation parameters based on visual recognition of fruits and vegetables according to claim 1, characterized in that, Performing object recognition and region segmentation on the pre-packaging image includes: Identify each fresh-cut fruit and vegetable object in the pre-packaging image and establish object identifiers. Create object outlines for the fresh-cut fruit and vegetable objects in the pre-packaging image and use the area enclosed by the object outlines as the fresh-cut fruit and vegetable object area. Within the area of the fresh-cut fruits and vegetables, image units are divided according to pixel blocks or superpixel blocks, and the continuous texture of the epidermis, wax reflection, outer surface texture of leaf veins, cutting edge, tissue fiber texture, internal color distribution and color gradient change are extracted from the image units. For each fresh-cut fruit and vegetable object, areas with continuous epidermal texture, leaf vein outer surface texture, waxy reflection, or continuous epidermal color bands are selected as non-cut surface tissue candidate areas. Regions with at least two of the following characteristics—cutting edge, internal tissue fiber texture, non-epidermal color distribution, and internal color gradient—and whose epidermal texture response value is below the corresponding threshold are considered as fresh cut surface candidate regions. The fresh cut surface discrimination value is calculated by weighting the inverse values of the cutting edge response value, tissue fiber texture response value, internal color gradient response value and epidermal texture response value. Based on the fresh cut surface discrimination value and epidermal texture response value, the image regions belonging to the same fresh cut fruit and vegetable object are divided into non-cut surface tissue regions and fresh cut surface regions.
4. The method for matching preservation parameters by combining visual recognition of fruits and vegetables according to claim 3, characterized in that, Performing object recognition and region segmentation on the pre-packaging image further includes: Continuous image units whose fresh cut discrimination value reaches the fresh cut discrimination threshold and whose epidermal texture response value is lower than the epidermal texture preservation threshold are merged into fresh cut regions, and continuous image units whose epidermal texture response value reaches the epidermal texture preservation threshold and whose fresh cut discrimination value is lower than the fresh cut threshold are merged into non-cut tissue regions. For a boundary image unit that simultaneously presents cutting edge features and epidermal texture features, its affiliation is determined based on the distance between the boundary image unit and the outer contour of the object, the category of adjacent regions, and the direction of edge continuity. In the fresh cut area, browning and mechanical damage areas are identified, and the union of browning and mechanical damage areas is taken as the abnormal cut area. When the two overlap, the overlapping image unit is counted once. The seepage area can be identified by collecting changes in the reflective liquid area, the continuous liquid coverage area, the connection relationship between the liquid area and the edge of the fresh cut area, or by intervals. Generate association records between region category, region range, and object identifier.
5. The method for matching preservation parameters by combining visual recognition of fruits and vegetables according to claim 1, characterized in that, The candidate packaging records in the preservation parameter database include fruit and vegetable category identifiers, cutting morphology identifiers, maturity level applicable conditions, state applicable conditions, temperature applicable conditions, packaging form conditions, packaging quality range, applicable storage duration range, packaging size conditions, film material, film specifications, gas permeability level, water vapor permeability level, and effective exchange area range; the gas permeability level includes oxygen permeability level and carbon dioxide permeability level; the state applicable conditions include cut surface load level applicable conditions, abnormal cut surface level applicable conditions, cut object density level applicable conditions, and leakage state level applicable conditions.
6. The method for matching preservation parameters by combining visual recognition of fruits and vegetables according to claim 1, characterized in that, The effective gas exchange area and effective water vapor exchange area are determined according to the packaging size and form recorded in the candidate packaging records, including: When the packaging is a tray with film, the effective gas exchange area and the effective water vapor exchange area are determined according to the film-covered area of the tray opening, the ventilation window area, and the overlapping area of the two. When the packaging is a bag, the effective gas exchange area and the effective water vapor exchange area are determined according to the area of the bag that participates in the exchange within the unfolded area. The effective gas exchange area is related to the gas permeability level and the packaging quality to form the gas exchange capacity level, and the effective water vapor exchange area is related to the water vapor permeability level and the packaging quality to form the water vapor exchange capacity level.
7. The method for matching preservation parameters by combining visual recognition of fruits and vegetables according to claim 1, characterized in that, Before forming a common feasible candidate set, perform a state monotonic consistency check on the candidate packaging record group in the preservation parameter library; In the candidate packaging record group, except for the state level to be verified, the other fruit and vegetable categories, cutting shapes, maturity levels, packaging forms, packaging size conditions, packaging quality ranges, and applicable temperature conditions are the same; the state monotonicity consistency verification is carried out by comparing discrete state levels, and candidate packaging records with increased cut surface load level and decreased gas exchange capacity level, increased leaching state level and decreased water vapor exchange capacity level, and increased abnormal cut surface level and extended applicable storage time are excluded.
8. The method for matching preservation parameters by combining visual recognition of fruits and vegetables according to claim 1, characterized in that, When the common feasible candidate set is not empty, the sorting result is determined according to the applicable storage duration coverage margin, packaging size adaptation relationship and effective exchange area margin, and the candidate packaging records in the sorting result are determined as the matching result; the matching result includes membrane material, membrane specifications, gas permeability grade, water vapor permeability grade, packaging size and storage parameters.
9. The method for matching preservation parameters by combining visual recognition of fruits and vegetables according to claim 1, characterized in that, When the common feasible candidate set is empty, packaging condition adjustment information is generated based on the constraint categories that are not met by the candidate packaging records; When gas exchange or water vapor exchange conditions are not met, adjustment items are generated to reduce packaging quality, increase effective exchange area, or change packaging form; when packaging size conditions are not met, adjustment items are generated to adjust the range of processable packaging specifications or change packaging form; when target storage time conditions are not met, adjustment items are generated to shorten target storage time, package in batches according to abnormal cut surface grade, or remove fresh-cut fruits and vegetables whose abnormal cut surface grade does not meet the conditions; the preservation parameter library is called back based on the adjustment items, and if the common feasible candidate set is still empty after the recall, manual verification information is output.
10. The method for matching preservation parameters by combining visual recognition of fruits and vegetables according to any one of claims 1 to 9, characterized in that, It also includes determining the reliability level of the area identification results and establishing parameter matching traceability records; when the area identification confidence information does not reach the preset reliability level of the corresponding area category, the status level of the area and the adjacent status levels are jointly recorded in the visual preservation status record to form a combined status applicable range, and the status applicable conditions of the candidate packaging record cover the combined status applicable range as the retention condition of the candidate packaging record. The parameter matching traceability record includes the image identifier of the pre-packaging image, packaging input information, region identification results, visual preservation status record, preservation parameter library version, candidate packaging record identifier, constraint comparison results, exclusion reasons, matching results, or packaging condition adjustment information.
