Intelligent conveying and sorting system and method for cosmetic packaging line

Abstract

The application discloses an intelligent conveying and sorting system and method for a cosmetic packaging line, and relates to the technical field of intelligent manufacturing and packaging automation.The system comprises the following steps: obtaining initial identification data and production sequence data of a cosmetic packaging piece, and determining an identification state according to the initial identification data; when the packaging piece is in a to-be-confirmed identification state, generating to-be-supplemented identification data and sequence maintenance data, and determining review control data according to the to-be-supplemented identification data; obtaining supplementary identification data according to the review control data, updating the to-be-supplemented identification data, and obtaining sequence regression data; and further generating output control data according to the identification confirmation state and the sequence regression state.The application realizes closed-loop control of review identification, sequence regression, sorting output and abnormal compensation of the to-be-confirmed packaging piece.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing and packaging automation technology, and in particular to an intelligent conveying and sorting system and method for a cosmetic packaging line. Background Technology

[0002] As cosmetic products evolve towards multi-category, small-batch, combined, and customized approaches, packaging lines often require continuous identification, conveying, verification, and sorting of different packages such as bottles, boxes, and sets. Existing packaging lines typically rely on visual recognition, barcode reading, or sensor detection to identify the package and control the sorting equipment based on the identification results. This method can meet basic usage requirements in scenarios where the packaging has a regular appearance, the identification area is stable, and the conveying rhythm is fixed.

[0003] However, in cosmetic packaging, packages often have features such as transparent bottles, mirrored caps, gold-stamped labels, cylindrical curved surfaces, small color code markings, or similar appearances. Key identification information may be in a state of pending confirmation due to reflections, obstructions, changes in posture, or partial blindness. Directly sorting based on the initial identification results can easily lead to missorting; conversely, uniformly rejecting or bypassing the confirmation process for unconfirmed packages may disrupt the conveyor cycle and cause shifts in their original batch, set, or order sequence positions. Especially on downstream production lines where boxing, set assembly, and batch traceability are interconnected, simply improving identification accuracy cannot adequately solve the coordination issues between post-confirmation regression, anomaly output, and sequence compensation.

[0004] Therefore, it is necessary to provide an intelligent conveying and sorting solution that can collaboratively generate verification control, sorting control, and anomaly compensation control based on identification data and production sequence data. Summary of the Invention

[0005] This invention overcomes the shortcomings of the prior art and provides an intelligent conveying and sorting system and method for cosmetic packaging lines.

[0006] To achieve the above objectives, the technical solution adopted by this invention is: an intelligent conveying and sorting method for a cosmetic packaging line, comprising the following steps:

[0007] S1. Obtain initial identification data and production sequence data for cosmetic packaging;

[0008] S2. Obtain the identification status of the cosmetic packaging based on the initial identification data. When the identification status is a pending identification status, obtain supplementary identification data and sequence maintenance data based on the initial identification data and the production sequence data.

[0009] S3. Determine the verification control data based on the identification data to be supplemented and the sequence preservation data, and obtain the supplementary identification data according to the verification control data;

[0010] S4. Update the identification data to be supplemented based on the supplementary identification data, and obtain sequence regression data based on the review control data and the production sequence data;

[0011] S5. Obtain the identification confirmation status based on the updated identification data to be supplemented, and obtain the sequence regression status based on the sequence regression data;

[0012] S6. Generate output control data based on the identification confirmation status and the sequence regression status.

[0013] In a preferred embodiment of the present invention, the initial identification data includes at least two of the following: image data, identification code data, batch number data, color code region data, contour data, pose data, and position data.

[0014] The identification status of the cosmetic packaging is obtained based on the initial identification data, including:

[0015] At least two identification confidence values ​​are obtained based on the initial identification data;

[0016] The at least two identification confidence values ​​are fused to obtain comprehensive identification data;

[0017] When the comprehensive identification data is lower than the preset identification threshold, the identification status to be confirmed is obtained.

[0018] In a preferred embodiment of the present invention, obtaining supplementary identification data based on the initial identification data includes:

[0019] Based on the initial identification data, at least one of the following is determined: unreadable identification code data, unreadable batch number data, invisible color number area data, reflective interference data, and abnormal contour matching data.

[0020] Generate corresponding items to be supplemented based on the determined data;

[0021] The items to be supplemented are associated to obtain the identification data to be supplemented.

[0022] In a preferred embodiment of the present invention, the production sequence data includes at least two of the following: batch number, set number, order number, target sorting path, target production location, and target arrival time.

[0023] Based on the production sequence data, sequence preservation data is obtained, including:

[0024] Target sequence location data is obtained based on the target production location and target arrival time;

[0025] Logical sequence constraint data is obtained based on the batch number, set number, and order number;

[0026] The sequence preservation data is obtained based on the target sequence position data and the logical sequence constraint data.

[0027] In a preferred embodiment of the present invention, determining verification control data based on the identification data to be supplemented and the sequence-preserving data includes:

[0028] The required type of review is determined based on the data to be supplemented.

[0029] Determine the allowable verification constraints based on the sequence preservation data;

[0030] Obtain available review resource data;

[0031] Based on the required review type, allowed review constraints, and available review resource data, generate review path data and review action data;

[0032] The review path data and the review action data are associated as the review control data.

[0033] In a preferred embodiment of the present invention, the allowable review constraint includes at least two of the allowable review time, allowable sequence offset, and allowable exposure risk amount;

[0034] The available verification resource data includes at least one of supplementary lighting resource data, secondary imaging resource data, pose adjustment resource data, and weight verification resource data.

[0035] When generating the review control data, candidate review data that exceeds the allowed review time, the allowed sequence offset, or the allowed contact risk exceeds the allowed contact risk are removed.

[0036] In a preferred embodiment of the present invention, obtaining supplementary identification data based on the review control data includes:

[0037] Starting with the identification data to be supplemented, at least one of the following is obtained based on the verification action data: supplementary lighting image data, secondary imaging data, image data after pose adjustment, and weight verification data;

[0038] The acquired data will be used as the supplementary identification data.

[0039] The image data after pose adjustment is data collected after the pose of the cosmetic packaging is adjusted under the condition of restricting lateral friction driving on the appearance surface of the cosmetic packaging.

[0040] In a preferred embodiment of the present invention, updating the identification data to be supplemented based on the supplementary identification data includes:

[0041] The supplementary identification data is obtained based on the supplementary identification data;

[0042] The supplementary identification data is matched with the identification data to be supplemented to obtain the data to be supplemented that can be deducted;

[0043] Based on the deductible data to be supplemented, the corresponding data item in the identification data to be supplemented is deducted, and the updated identification data to be supplemented is obtained.

[0044] When the updated data to be identified meets the preset identification confirmation threshold, an identification confirmation status that meets the preset identification confirmation conditions is obtained.

[0045] When the updated data to be identified does not meet the preset identification confirmation threshold, an identification confirmation status that does not meet the preset identification confirmation conditions is obtained.

[0046] In a preferred embodiment of the present invention, sequence regression data is obtained based on the review control data and the production sequence data, including:

[0047] Based on the aforementioned review control data, review time data, regression candidate position data, and regression candidate time data are obtained;

[0048] Based on the production sequence data, target production location data and target arrival time data are obtained;

[0049] Based on the verification time data, regression candidate location data, regression candidate time data, target production location data, and target arrival time data, sequence deviation data is obtained, and the sequence deviation data is used as the sequence regression data.

[0050] When the sequence deviation data does not exceed the preset sequence deviation threshold, a sequence regression state that satisfies the preset sequence regression condition is obtained;

[0051] When the sequence deviation data exceeds the preset sequence deviation threshold, a sequence regression state that does not meet the preset sequence regression conditions is obtained;

[0052] The process of generating output control data based on the identification confirmation status and the sequence regression status includes:

[0053] When the identification confirmation status meets the preset identification confirmation conditions and the sequence regression status meets the preset sequence regression conditions, sorting control data is generated;

[0054] When at least one of the identification confirmation state and the sequence regression state fails to meet the corresponding conditions, abnormal output control data and sequence compensation data are generated.

[0055] An intelligent conveying and sorting system for a cosmetic packaging line includes: a data acquisition mechanism for acquiring initial identification data and production sequence data of cosmetic packaging components;

[0056] A verification conveying mechanism is used to convey cosmetic packaging pieces whose status needs to be confirmed and identified based on verification control data;

[0057] A verification and identification agency is used to obtain supplementary identification data;

[0058] The sorting execution mechanism is used to control the cosmetic packaging to enter the target sorting path according to the sorting control data;

[0059] An abnormal output mechanism is used to control the cosmetic packaging to enter the abnormal output path based on abnormal output control data.

[0060] The controller is connected to the data acquisition mechanism, the verification and conveying mechanism, the verification and identification mechanism, the sorting execution mechanism, and the abnormal output mechanism, respectively.

[0061] The controller is used to execute any of the methods described herein.

[0062] This invention addresses the shortcomings of the prior art and has the following beneficial effects:

[0063] (1) This invention obtains initial identification data and production sequence data, and generates supplementary identification data, sequence maintenance data and verification control data in the state of pending identification confirmation, so that the verification, transportation and sorting of cosmetic packaging are driven by data status; since identification confirmation and sequence regression jointly participate in output control, the risk of missorting and sequence disorder can be directly reduced; compared with the existing sorting method based only on a single identification result, it further improves the continuous transportation stability and sorting reliability in mixed-line packaging scenarios.

[0064] (2) The present invention characterizes unconfirmed content such as identification code, batch number, color number, reflection or outline through supplementary identification data, and obtains supplementary identification data based on review control data, so that the system can supplement and review the reasons for insufficient identification; since the review action corresponds to the supplementary identification content, it can directly improve the accuracy of identity confirmation of the package to be confirmed; compared with the existing method of directly rejecting or uniformly re-inspecting low-positioned letters, it further reduces the waste of time caused by mis-rejection, mis-sorting and unnecessary review.

[0065] (3) This invention generates sequence retention data through production sequence data and obtains sequence regression data based on review control data and production sequence data, so that the review process not only considers the identification results, but also the regression status of the package in the batch, set or order sequence. Since sorting control data is only generated when the identification confirmation status and the sequence regression status meet the conditions, batch misalignment caused by the return after review can be directly avoided. Compared with the existing bypass review and direct return method, it further ensures the continuity of downstream boxing, set combination and batch traceability.

[0066] (4) The present invention forms a closed loop by verifying control data, supplementing identification data, sequence regression data and output control data, so that identification supplementation, verification conveying, sequence regression, sorting output and abnormal compensation are interconnected; since abnormal output control data and sequence compensation data can intervene in time when identification or sequence conditions are not met, it can directly prevent unconfirmed packages from continuing to enter the normal sorting chain; compared with the existing method of simply relying on manual verification or end rejection, it further improves the automation coordination capability and abnormal handling controllability of cosmetic packaging lines. Attached Figure Description

[0067] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0068] Figure 1 This is a flowchart of a preferred embodiment of the present invention;

[0069] Figure 2 This is a system structure block diagram of a preferred embodiment of the present invention. Detailed Implementation

[0070] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein. Therefore, the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0072] This invention is mainly applied to continuous conveying, online identification, verification and confirmation, path sorting, abnormal output and sequence compensation scenarios in cosmetic packaging production lines. Specifically, it can be applied to bottled cosmetics, boxed cosmetics, tube cosmetics, canned cosmetics, combination sets, packaging with color number markings, packaging with batch number inkjet printing, and packaging with transparent bottles, mirror caps, hot stamping labels, cylindrical curved surfaces or similar appearance features.

[0073] In the application scenario of this invention, cosmetic packaging pieces move continuously along the conveyor line. The packaging pieces flow between the initial identification station, the verification conveying station, the verification identification station, the regression judgment station, the sorting execution station, and the abnormal output station. The process of identity verification, sequence maintenance, verification regression, sorting output, and abnormal compensation needs to be completed without significantly disrupting the production cycle.

[0074] This invention, through the synchronous processing of initial identification data and production sequence data, establishes a correlation between the identification status, supplementary content, verification method, return position, and output path of the package, thereby enabling the package to be verified to be simultaneously constrained by both identity verification conditions and production sequence conditions during the verification process.

[0075] Example 1:

[0076] like Figure 1 As shown, this embodiment provides an intelligent conveying and sorting method for a cosmetic packaging line, including:

[0077] Step S1: Obtain initial identification data and production sequence data of cosmetic packaging;

[0078] Step S2: Obtain the identification status of the cosmetic packaging based on the initial identification data. When the identification status is pending confirmation, obtain the supplementary identification data and sequence maintenance data based on the initial identification data and production sequence data.

[0079] Step S3: Determine the verification control data based on the identification data to be supplemented and the sequence preservation data, and obtain the supplementary identification data based on the verification control data;

[0080] Step S4: Update the identification data to be supplemented based on the supplementary identification data, and obtain the sequence regression data based on the review control data and production sequence data;

[0081] Step S5: Obtain the identification confirmation status based on the updated identification data to be supplemented, and obtain the sequence regression status based on the sequence regression data;

[0082] Step S6: Generate output control data based on the identification confirmation status and sequence regression status.

[0083] In this embodiment, after the cosmetic packaging enters the initial identification station, the controller first receives data provided by visual acquisition, label reading, inkjet printing, position detection and production task interface, and then associates and stores the initial identification data of the same packaging with the production sequence data so that it can be called in subsequent identification judgment, verification control, sequence regression and output control.

[0084] When the initial identification data meets the preset identification conditions, the controller can directly generate the corresponding sorting control data based on the production sequence data and the target sorting path. When the initial identification data does not meet the preset identification conditions, the controller enters the pending identification processing flow and splits the insufficient identification content to generate supplementary identification data.

[0085] In this embodiment, the data to be supplemented for identification is used to characterize the data items that still need to be supplemented and confirmed in the current packaging, the sequence maintenance data is used to characterize the positional relationship that the packaging needs to maintain in the production queue, the verification control data is used to control the verification path, verification action and regression conditions of the packaging in the verification process, and the supplementary identification data is used to supplement the corresponding data items in the data to be supplemented for identification.

[0086] After the verification is completed, the controller updates the data to be identified based on the supplementary identification data, and obtains the sequence regression data based on the verification path, verification time, regression candidate position, target production position and target arrival time. Then, it obtains the identification confirmation status based on the updated data to be identified, and obtains the sequence regression status based on the sequence regression data.

[0087] When both the identification confirmation status and the sequence regression status meet the corresponding conditions, the controller generates sorting control data and controls the package to enter the target sorting path. When the identification confirmation status or the sequence regression status does not meet the corresponding conditions, the controller generates abnormal output control data and sequence compensation data to control the package to enter the abnormal output path and to make compensation records for the production queue.

[0088] The working conditions encountered during the implementation of this embodiment include: unclear edge contours due to transparent bottles, local overexposure due to mirrored bottle caps, interference with character recognition due to hot stamping labels, distortion of identification codes due to cylindrical curved surfaces, color recognition fluctuations due to small areas of color codes, insufficient contour matching due to similar packaging, and misalignment between images and production sequence data due to high-speed transportation.

[0089] In response to the above working conditions, this embodiment identifies the reasons for the deficiency by recording the data to be supplemented, maintains the time and location constraints for the verification of the data records in the sequence, selects the verification action corresponding to the item to be supplemented by the verification control data, and participates in the output control by identifying the confirmation status and the sequence regression status together, so that the verification, regression, sorting and anomaly handling of the package to be confirmed form a closed loop.

[0090] Example 2:

[0091] For step S1, in engineering implementation, this step can be understood as follows: when the cosmetic packaging enters the initial identification station of the packaging line, the controller acquires data that can characterize the appearance, identity, posture, position and production sequence of the packaging through the acquisition device that is triggered synchronously with the packaging, and binds the above data according to the data record of the same packaging.

[0092] The initial identification data includes at least two of the following: image data, identification code data, batch number data, color code area data, contour data, pose data, and location data.

[0093] Image data can be front, side, or top images of the packaging, bottle cap images, label area images, box exterior images, or set frame images. Image data can be acquired by area scan cameras, line scan cameras, industrial cameras, or high dynamic range cameras.

[0094] Identification code data can be barcode data, QR code data, RFID tag reading data, electronic tag data, or virtual identity code data bound to the packaging. Identification code data can be used to directly determine the identity of the packaging, order association, or target sorting path.

[0095] Batch number data can be inkjet printed batch number, laser-printed batch number, embossed batch number, label batch number, or production traceability character data. Batch number data can be used to determine the production batch, filling batch, boxing batch, or traceability batch to which the packaging belongs.

[0096] Color code area data can be color code label images, color block area images, color code number character images, color matrix features, or local color distribution features. Color code area data can be used to distinguish between cosmetic packaging, packaging with different color codes in the same series, or combinations of similar-looking items.

[0097] Contour data can be the edge contour of the packaging, the shape of the bottle, the boundary of the box, the outer frame of the set, the circular contour of the bottle cap, the curve of the bottle shoulder, or the projected area of ​​the packaging. Contour data can be obtained through image edge detection, depth sensing, photoelectric array, or contour scanning equipment.

[0098] Pose data can include the deflection angle, tilt angle, flipping state, orientation state, label visibility state, or identification code area orientation state of the package relative to the conveying direction. Pose data can be used to determine whether the initial recognition failure is caused by posture deviation.

[0099] Location data can be the station coordinates of the package on the conveyor line, encoder pulse position, entry detection time, image acquisition time, arrival time at the detection point, virtual queue position, or pallet number. Location data can be used to match the acquired data with production sequence data.

[0100] The production sequence data includes at least two of the following: batch number, set number, order number, target sorting path, target production location, and target arrival time.

[0101] Batch number is used to identify the production batch to which the package belongs; set number is used to identify the combination position or matching relationship of the package in the set; order number is used to identify the order task or packaging task corresponding to the package; target sorting path is used to identify the sorting channel that the package should enter under normal conditions; target production position is used to identify the target queue position of the package in the predetermined production cycle; target arrival time is used to identify the time when the package should arrive at the target workstation according to the production cycle.

[0102] Specifically, production sequence data can be sent to the controller by the production execution system, packaging line control system, order management system, batch traceability system, sorting task system or host computer scheduling system, or it can be generated by the controller based on the upstream filling sequence, labeling sequence, cartoning sequence or pallet binding relationship.

[0103] In one implementation, after the entrance detection sensor detects that the package has entered the initial identification station, the controller outputs a trigger signal to the image acquisition device, barcode scanning device, inkjet printing device and position detection device, and collects the data returned by each device according to the trigger time, station number and encoder position.

[0104] In one implementation, the controller generates a temporary package number when the package passes through the entrance detection sensor, and then establishes a correspondence between the temporary package number and image data, identification code data, batch number data, color code area data, contour data, pose data, position data, and production sequence data.

[0105] In one implementation, when a package is transported on a pallet or fixture, the controller can use the pallet number, fixture number, or carrier RFID number as the primary key for the package data record and bind the initial identification data and production sequence data to that primary key.

[0106] In one implementation, when packages are transported directly without pallets, the controller can generate a virtual queue number based on the encoder pulse value of the conveyor line and the detection interval of adjacent packages, and use the virtual queue number as the basis for data association.

[0107] Preferably, the trigger position error of the image acquisition device is no greater than half the length of a package, the timestamp synchronization error is no greater than one-third of the time interval between adjacent packages, and when encoder pulse binding is used, the encoder position difference of the same package at the initial recognition station can be limited to a preset position tolerance range.

[0108] Specifically, the preset position tolerance range can be 5mm to 50mm, or 5% to 20% of the center distance between adjacent packages. When the conveying speed is greater than 0.8m per second, the encoder pulse and trigger time can be bound together first. When the conveying speed is less than or equal to 0.8m per second, the trigger time, station number and package position can be bound together.

[0109] Preferably, after acquiring the initial recognition data, the controller can perform brightness equalization, distortion correction, edge enhancement, reflective area suppression, target area cropping, or scale normalization on the image data, and use the processed data as input data for subsequent recognition status judgment.

[0110] Specifically, brightness equalization can be achieved using histogram equalization or adaptive histogram equalization, distortion correction can be performed based on camera calibration parameters, edge enhancement can be achieved using gradient operators or morphological processing, and reflective area suppression can be achieved based on high-brightness area thresholding, polarization image difference, or multi-exposure fusion.

[0111] This step acquires initial identification data and production sequence data simultaneously, enabling the packaged goods to have candidate identity information, appearance feature information, pose information, location tracking information, and production queue information before entering the subsequent review process. This provides a unified data foundation for supplementary identification, sequence preservation, review control, and sequence regression.

[0112] For step S2, in engineering implementation, this step can be understood as the controller performing identification confidence calculation and result consistency judgment on the initial identification data to determine whether the package can be directly confirmed. When the identification result does not meet the preset conditions, the controller further generates supplementary identification data and sequence preservation data.

[0113] The identification status includes a verifiable identification status and a pending identification status. The verifiable identification status indicates that the identity, batch, color number, target sorting path, or production sequence of the package has reached a preset level of confidence. The pending identification status indicates that the package has missing key identification data, insufficient confidence value, conflicting identification results, or identification results that are inconsistent with the production sequence.

[0114] Specifically, the controller can obtain image recognition confidence value based on image data, identification code recognition confidence value based on identification code data, batch number recognition confidence value based on batch number data, color number recognition confidence value based on color number region data, contour matching confidence value based on contour data, and pose recognition confidence value based on pose data.

[0115] Each identification confidence value can be normalized to between 0 and 1 before being weighted and fused to comprehensively identify the data. It can be calculated using the following formula:

[0116] ,

[0117] in, For image recognition confidence values, To identify confidence values ​​for the identifier code, Confidence values ​​are assigned to batch numbers. Identify confidence values ​​for color code regions. For contour matching confidence values, For pose-identifiable confidence values, , , , , and These are the corresponding weights, and each weight satisfies the following formula:

[0118] ,

[0119] Preferred, The weight can be taken as 0.20 to 0.45. The weight can be between 0.10 and 0.30. The weight can be taken as 0.10 to 0.35. The weight can be taken as 0.05 to 0.25. The weight can be set from 0.05 to 0.20, and the specific weight can be configured according to the type of packaging, product similarity, identification code reliability and production traceability requirements.

[0120] In one implementation, when the identification code is successfully recognized and the corresponding order number, batch number, and target sorting path are all consistent with the production sequence data, the controller can determine the package identification status as a verifiable identification status.

[0121] In one implementation, when the identification code is unreadable but the image recognition result, batch number recognition result, color number recognition result, and contour matching result all meet the preset recognition threshold, and the production sequence data correspondence is consistent, the controller can determine the packaging recognition status as a verifiable recognition status.

[0122] In one implementation, when the comprehensive identification data is lower than the preset identification threshold, or when there is a conflict between the identification code identification result and the batch number identification result, or when the color number identification result is inconsistent with the order sequence number, the controller determines the packaging identification status as an identification status pending confirmation.

[0123] Preferably, the preset recognition threshold can be 0.70 to 0.95. When the packaging types are significantly different and the number of target sorting paths is small, the preset recognition threshold can be 0.70 to 0.82. When the packaging appearance is highly similar, the number of color codes is large, or it involves combination sets, the preset recognition threshold can be 0.82 to 0.95.

[0124] The data to be supplemented for identification is used to characterize the data items that still need to be confirmed in the current identification stage of the package. Specifically, it may include at least one of the following: unreadable identification code data, unreadable batch number data, invisible color code area data, reflective interference data, abnormal contour matching data, abnormal pose data, production sequence conflict data, and target path conflict data.

[0125] Specifically, after the controller determines various types of insufficient identification data based on the initial identification data, it generates corresponding supplementary items based on the determined data and establishes associations between the supplementary items and the temporary packaging number, workstation location, production sequence data, target sorting path, required review type, review priority, and abnormal risk level to obtain the identification data to be supplemented.

[0126] Preferably, when the number of failed identification code readings reaches 1 to 3 times, or the identification code recognition confidence value is lower than 0.60 to 0.80, the controller generates an identification code unreadable supplementary item.

[0127] Preferably, when the proportion of missing characters in the batch number recognition is greater than 10% to 40%, or when there are unidentifiable characters, duplicate characters, broken characters, or stuck characters in the character recognition results, the controller generates an unreadable batch number to be supplemented.

[0128] Preferably, when the proportion of color code area being obscured is greater than 20% to 50%, or the color difference value of the color code area is lower than the preset color differentiation threshold, the controller generates a color code area invisible to be supplemented or a color code unconfirmed to be supplemented.

[0129] Preferably, when the area of ​​the bright reflective region in the image accounts for more than 15% to 45% of the area of ​​the key recognition region, or when the reflective region covers the identification code, batch number, color number area or product label area, the controller generates a reflective interference supplement item.

[0130] Preferably, when the contour matching confidence value is lower than 0.65 to 0.85, or when the deviation between the package projection area, length-to-width ratio, bottle cap contour, bottle shoulder curve and the preset product template exceeds the preset range, the controller generates a contour matching abnormality to be supplemented.

[0131] Sequence preservation data is used to characterize the production sequence constraints that the package to be verified needs to maintain during the verification process. Specifically, it may include at least one of the following: target sequence position data, logical sequence constraint data, allowed verification time, allowed sequence offset, allowed regression position range, allowed insertion time window, adjacent package constraints, and package combination constraints.

[0132] Specifically, the controller obtains target sequence position data based on the target production location and target arrival time, obtains logical sequence constraint data based on batch number, set number and order number, and then obtains sequence holding data based on the target sequence position data and logical sequence constraint data.

[0133] The target sequence location data can be represented by the target workstation number, target conveyor coordinates, target encoder pulse value, target queue number, target arrival time, or the expected sorting position before the target sorting path.

[0134] Logical sequence constraint data can be represented by batch continuity constraints, order consistency constraints, package integrity constraints, target path consistency constraints, upstream and downstream workstation connection constraints, or traceability record integrity constraints.

[0135] Preferably, the allowable sequence offset can be determined based on the package center distance, conveying speed and downstream cartoning cycle time, and can be 0 to 3 package positions, or within the range of 0.5 to 10 seconds of the target arrival time.

[0136] Preferably, when the downstream station is a direct boxing station, the allowed sequence offset can be set to 0 to 1 package position; when the downstream station has buffer rearrangement capability, the allowed sequence offset can be set to 1 to 3 package positions.

[0137] This step, through confidence fusion, result consistency judgment, supplementary item generation, and sequence constraint generation, enables the packages to be confirmed to have executable verification requirements and verifiable regression conditions, thereby reducing the impact of single identification fluctuations on sorting operations and lowering the risk of batch misalignment or mismatched sets during the verification process.

[0138] For step S3, in engineering implementation, this step can be understood as follows: the controller determines the required verification type based on the identification data to be supplemented, determines the allowed verification constraints based on the sequence preservation data, and then generates verification control data adapted to the current package by combining the available verification resource data.

[0139] The required verification types may include at least one of the following: supplementary light verification, secondary imaging verification, local magnification verification, polarization imaging verification, pose adjustment verification, rescanning verification, batch number character verification, weight verification, and multi-information fusion verification.

[0140] The verification control data includes verification path data and verification action data. The verification path data is used to indicate the conveying route of the packaged item into the verification conveying section, buffer section, bypass section, return section or abnormal temporary storage section. The verification action data is used to instruct the verification identification mechanism to perform supplementary lighting, imaging, barcode scanning, weighing, rotation, guidance, blocking or return actions.

[0141] Available verification resource data includes at least one of the following: supplementary lighting resource data, secondary imaging resource data, pose adjustment resource data, weight verification resource data, rescanning resource data, cache space data, verification station idle data, and return path occupancy data.

[0142] Specifically, the controller first reads the type and priority of each item to be supplemented in the identification data, then matches candidate review actions according to the type of the item to be supplemented, and determines whether each candidate review action meets the allowable review time, allowable sequence offset and allowable contact risk based on the sequence holding data.

[0143] In one implementation, when the item to be supplemented is a reflective interference item, the controller prioritizes matching candidate verification actions such as supplementary light angle adjustment, polarization imaging, multi-exposure fusion, or high dynamic range imaging.

[0144] In one implementation, when the item to be supplemented is an unreadable identifier, the controller prioritizes matching candidate verification actions such as rescanning, local magnification imaging, secondary imaging, scanning after pose adjustment, or image enhancement of the identifier region.

[0145] In one implementation, when the item to be supplemented is an unreadable batch number, the controller prioritizes matching candidate review actions such as inkjet area positioning, character area cropping, local supplementary lighting, character enhancement, secondary character recognition, or manual review prompt generation.

[0146] In one implementation, when the item to be supplemented is an item that is not visible in the color code area, the controller prioritizes matching candidate verification actions such as pose adjustment, lateral imaging, color code area supplementary lighting, color calibration, color block recognition, or comparison with the order color code table.

[0147] In one implementation, when the item to be supplemented is an item with abnormal contour matching, the controller prioritizes matching candidate verification actions such as multi-angle imaging, contour reconstruction, weight verification, size verification, or secondary matching with the product template library.

[0148] The allowable review constraints include at least two of the following: allowable review time, allowable sequence offset, and allowable contact risk. The allowable review time can be calculated based on the target arrival time, the current time, the estimated time of the review path, and the downstream cycle time margin. The allowable sequence offset can be calculated based on the target queue position, the insertable position, and the sequence relationship of adjacent packages. The allowable contact risk can be calculated based on the appearance and material of the package, the contact method of the review action, and the sensitivity level of the contact area.

[0149] Preferably, the permissible contact risk can be divided into three levels: low risk, medium risk, and high risk. Transparent bottles, mirror caps, hot stamping labels, easily scratched boxes, or labeling areas can be configured as highly sensitive appearance surfaces. When the posture adjustment mechanism performs actions on highly sensitive appearance surfaces, lateral friction drive should be restricted.

[0150] Specifically, when the controller generates review control data, it removes candidate review data that exceeds the allowed review time, the sequence offset exceeds the allowed sequence offset, or the contact risk exceeds the allowed contact risk, and selects data with high review coverage, small regression deviation, and low resource consumption from the remaining candidate review data as review control data.

[0151] Preferably, the candidate review data may include candidate review paths, candidate review actions, expected review time, expected regression location, expected regression time, coverage of items to be supplemented, exposure risk amount, resource occupancy status, and alternative paths for abnormal transfer.

[0152] Specifically, the coverage of items to be supplemented can be determined based on the ratio of the number of items to be supplemented that can be supplemented by the candidate review action to the total number of items to be supplemented, and the resource occupancy status can be determined based on the idle time of the review workstation, the status of the supplementary lighting equipment, the availability status of the camera, the occupancy status of the buffer position, and the occupancy status of the return transport section.

[0153] In one implementation, the controller can establish a review action mapping table, which stores the type of item to be supplemented, recommended review actions, action time range, contact risk level, applicable packaging type, and corresponding review resources.

[0154] In one implementation, the controller can filter candidate review actions from the review action mapping table based on the current package to be supplemented, and then generate review control data based on the current review resource occupancy status and sequence holding data.

[0155] When acquiring supplementary identification data based on the verification control data, the controller takes the identification data to be supplemented as the starting point and acquires at least one of the following based on the verification action data: supplementary light image data, secondary imaging data, image data after pose adjustment, weight verification data, rescanning data, and local character verification data, and uses the acquired data as supplementary identification data.

[0156] Preferably, the supplementary lighting image data can be acquired under different incident angles, different brightness, different color temperatures, or polarized light conditions, and the supplementary lighting brightness can be adjusted according to the average brightness of the initial image, the area of ​​the bright area, and the location of the key recognition area.

[0157] Preferably, the secondary imaging data can be acquired under conditions of short exposure, high dynamic range imaging, local magnification imaging, lateral imaging, or multi-angle imaging, and the secondary imaging area can be determined by the position of the item to be supplemented in the data to be supplemented.

[0158] Preferably, the image data after pose adjustment is the data collected after the pose adjustment of the cosmetic packaging under the condition of limiting lateral friction driving on the appearance surface of the cosmetic packaging. The pose adjustment can be achieved by lifting rotation, flexible limiting, low friction roller, airflow assistance, guide baffle or carrier rotation.

[0159] Preferably, the weight verification data can be obtained through a dynamic weighing module or a static weighing module. The allowable error of the weight verification data can be set to ±0.5% to ±5% of the target weight. When the package weight is light or the conveying speed is high, multiple sampling averages or median filtering can be used to reduce vibration interference.

[0160] This step, through the joint control of items to be supplemented, sequence constraints, and review resources, ensures that the review action corresponds to the identified reasons for the deficiency and that the review path meets the production sequence maintenance conditions, thereby reducing cycle time loss caused by unified review and improving the effectiveness of review and confirmation of packages to be confirmed.

[0161] For step S4, in engineering implementation, this step can be understood as the controller updating the data to be identified based on the supplementary identification data, and obtaining sequence regression data based on the offset caused by the position and time of the package during the verification process.

[0162] Among them, the supplementary identification data is the data extracted from the supplementary identification data that can be used to supplement the items to be supplemented, the deductible supplementary data is the data formed after matching the supplementary identification data with the corresponding items to be supplemented in the supplementary identification data, and the updated supplementary identification data is used to represent data items that have not been confirmed after review.

[0163] Specifically, after receiving the supplementary identification data, the controller can first preprocess the supplementary identification data, then obtain the identification supplement data based on the supplementary identification data, and match the identification supplement data with the items to be supplemented in the identification data to be supplemented item by item.

[0164] In one implementation, when the supplementary identification data is supplementary lighting image data, the controller re-identifies the identification code, batch number, color number area or label area based on the supplementary lighting image data, and matches the re-identification result with the corresponding item to be supplemented.

[0165] In one implementation, when the supplementary identification data is secondary imaging data, the controller performs cropping, magnification, sharpening, character segmentation, or code area positioning on local areas in the secondary imaging data, and obtains identification code supplementary data or batch number supplementary data based on the processing results.

[0166] In one implementation, when the supplementary identification data is image data after pose adjustment, the controller redetermines the label surface, color code area, identification code area or outline boundary of the package based on the adjusted image, and uses the redetermined data as supplementary identification data.

[0167] In one implementation, when the supplementary identification data is weight verification data, the controller compares the weight verification data with the product weight template, order product specifications, or batch specification range, and uses the comparison result as supplementary identification data.

[0168] Preferably, when the identification confidence value corresponding to the supplementary data is not lower than the preset supplementary threshold, the controller allows the corresponding supplementary item to be deducted. The preset supplementary threshold can be 0.70 to 0.95.

[0169] Preferably, for data with clear character results such as identification codes, batch numbers, and order serial numbers, the preset supplementation threshold can be 0.80 to 0.95, and for auxiliary recognition data such as contours, color codes, weights, and poses, the preset supplementation threshold can be 0.70 to 0.88.

[0170] Specifically, the deduction methods may include deleting the corresponding item to be supplemented, reducing the risk weight of the corresponding item to be supplemented, marking the corresponding item to be supplemented as confirmed, adjusting the corresponding item to be supplemented from a high-risk level to a low-risk level, or adding a supplementary source mark to the corresponding item to be supplemented.

[0171] Preferably, when there is a conflict between the supplementary identification data and the production sequence data, the controller can retain the corresponding supplementary item and generate a conflict supplementary item. The conflict supplementary item may include target path conflict, batch conflict, order conflict, set serial number conflict or color number conflict.

[0172] Sequence regression data is used to characterize the deviation of the package from the target production location and target arrival time after verification. Sequence regression data can include verification time data, regression candidate location data, regression candidate time data, target production location data, target arrival time data, and sequence deviation data.

[0173] Specifically, the controller obtains verification time data, regression candidate position data, and regression candidate time data based on the verification control data, obtains target production position data and target arrival time data based on the production sequence data, and then obtains sequence deviation data based on the verification time data, regression candidate position data, regression candidate time data, target production position data, and target arrival time data.

[0174] The time data for verification can include the time it takes for the package to enter the verification path from the main conveyor line, the time it stays at the verification station, the time it takes to perform the verification action, the time it takes to return to the main conveyor line from the verification path, and the time it takes to wait for regression insertion.

[0175] Regression candidate location data can include the station location, queue location, conveyor coordinates, encoder pulse value, or buffer release location of the package that can be regressed to the main conveyor line.

[0176] Regression candidate time data can include the estimated time for the package to reach the main conveyor regression position, the estimated time to reach the downstream cartoning station, the estimated time to reach the sorting execution station, or the estimated time to enter the target sorting path.

[0177] Preferably, to avoid data of different dimensions from directly participating in the calculation, the controller can normalize the positional deviation, time deviation, and logical sequence penalty term before calculating the sequence deviation data.

[0178] Specifically, sequence bias data It can be calculated using the following formula:

[0179] ,

[0180] in, To return to the candidate position, For the target production location, To allow for an upper limit of positional deviation, For the candidate time of regression, For the target arrival time, To set the upper limit for allowable time deviation, For logical sequence penalty terms, For the target path penalty item, , , and Let be the weighting coefficients, and satisfy the following formula:

[0181] ,

[0182] Preferably, when the packaging does not conflict with the sequence of adjacent batches, sets, or orders, Set to 0 when there is a conflict between the package and the sequence of adjacent batches, sets, or orders. Set to 1, when the target sorting path after the package return matches the production sequence data. Set to 0 when there is a conflict in the target sorting path. Take 1.

[0183] Preferably, a can be 0.15 to 0.45, b can be 0.20 to 0.50, c can be 0.10 to 0.45, and d can be 0.05 to 0.30. When the production line has high requirements for the assembly sequence of the sets, c can be a higher value. When the production line has high requirements for the cycle time, b can be a higher value. When there are many target sorting paths and the risk of path switching is high, d can be a higher value.

[0184] In one implementation, the controller can also determine the queue offset based on the number of adjacent packages between the regression candidate location and the target production location, and use the queue offset as part of the sequence regression data.

[0185] In one implementation, the controller can also determine whether the package still has a normal sorting window based on the remaining conveying distance before entering the target sorting path after the package has been checked and the response time of the sorting actuator.

[0186] This step updates the data to be supplemented by supplementing the identification data, so that the data items that have been supplemented can be removed or downgraded from the state to be supplemented. It also quantifies the time and position offset caused by the verification process by using sequence regression data, thereby providing a basis for judging the identification confirmation state and the sequence regression state.

[0187] For step S5, in engineering implementation, this step can be understood as the controller determining the degree of identification of the package based on the updated identification data to be supplemented, and determining whether the package meets the conditions for returning to the production sequence after verification based on the sequence regression data.

[0188] The identification confirmation status includes identification confirmation status that meets the preset identification confirmation conditions and identification confirmation status that does not meet the preset identification confirmation conditions. The sequence regression status includes sequence regression status that meets the preset sequence regression conditions and sequence regression status that does not meet the preset sequence regression conditions.

[0189] Specifically, the controller determines whether the updated data to be supplemented still contains key items to be supplemented, and calculates the identification confirmation result based on the type, quantity, risk level, source of supplementation, and corresponding confidence value of the remaining items to be supplemented.

[0190] Key items to be supplemented may include at least one of the following: unreadable identification code, unreadable batch number, unconfirmable color code area, conflicting target sorting path, conflicting set serial number, conflicting order serial number, conflicting batch, and abnormal contour matching.

[0191] Preferably, when there are no key items to be supplemented in the updated data to be supplemented, or when the comprehensive risk value of the remaining items to be supplemented is lower than the preset identification confirmation threshold, the controller obtains an identification confirmation status that meets the preset identification confirmation conditions.

[0192] Preferably, when there are still key items to be supplemented in the updated data to be supplemented, or when the comprehensive risk value of the remaining items to be supplemented is not lower than the preset identification confirmation threshold, the controller obtains an identification confirmation state that does not meet the preset identification confirmation conditions.

[0193] Specifically, the comprehensive risk value of the remaining items to be supplemented. It can be calculated using the following formula:

[0194] ,

[0195] in, For the first The risk value of the remaining items to be supplemented. For the first The weighting coefficients of the remaining items to be added. This represents the number of remaining items to be added. Number the remaining items to be added.

[0196] Preferably, when the risk values ​​of all remaining items to be supplemented are normalized to between 0 and 1, the weighting coefficients can satisfy the following formula to make the overall risk value... Stay within the normalization risk range:

[0197] ,

[0198] Preferably, the risk value for unreadable identification code supplementary items can be configured to be 0.70 to 1.00, the risk value for unreadable batch number supplementary items can be configured to be 0.50 to 0.90, the risk value for unconfirmable color code area supplementary items can be configured to be 0.50 to 0.95, the risk value for abnormal contour matching supplementary items can be configured to be 0.40 to 0.85, and the risk value for abnormal pose supplementary items can be configured to be 0.20 to 0.60.

[0199] Preferably, the preset identification and confirmation threshold can be a risk threshold of 0.20 to 0.60. When the comprehensive risk value of the remaining items to be supplemented is lower than this threshold, the identification and confirmation status can be considered to meet the preset identification and confirmation conditions. When the comprehensive risk value of the remaining items to be supplemented is not lower than this threshold, the identification and confirmation status can be considered to not meet the preset identification and confirmation conditions.

[0200] For the sequence regression status, the controller can compare the sequence deviation data with the preset sequence deviation threshold, and combine the target path consistency, downstream cache status, package combination integrity and sorting action window to determine the sequence regression status.

[0201] Preferably, when the sequence deviation data does not exceed the preset sequence deviation threshold, and the return position of the package does not disrupt the batch continuity, order consistency, and set integrity, the controller obtains a sequence regression state that meets the preset sequence regression conditions.

[0202] Preferably, when the sequence deviation data exceeds the preset sequence deviation threshold, or when the return position of the package will cause a conflict in the batch, order, set or target sorting path, the controller obtains a sequence regression state that does not meet the preset sequence regression conditions.

[0203] Specifically, the preset sequence deviation threshold can be a normalized value of 0.20 to 0.80. For downstream packaging lines that directly box, directly assemble, or strictly trace batches, the preset sequence deviation threshold can be 0.20 to 0.50. For downstream packaging lines that have the ability to buffer, wait, rearrange, or manually confirm, the preset sequence deviation threshold can be 0.50 to 0.80.

[0204] In one implementation, if the target sorting action window has been missed after the package has been checked, the controller can determine the sequence regression status as not meeting the preset sequence regression condition even if the sequence deviation data does not exceed the preset sequence deviation threshold.

[0205] In one implementation, once the package has completed its verification, it can enter the downstream cache reordering area, and the cache reordering area can restore the package to the target queue position. The controller can then determine the sequence regression status as meeting the preset sequence regression conditions.

[0206] This step, by separately judging the identification confirmation status and the sequence return status, enables the controller to record the two conditions of identity confirmation and queue return respectively, and provides a control basis for subsequent normal sorting, abnormal output, delayed return, sequence compensation or manual review.

[0207] For step S6, in engineering implementation, this step can be understood as the controller generating output control data based on the identification confirmation status and the sequence regression status, and controlling the package to enter the target sorting path, abnormal output path or sequence compensation process through the output control data.

[0208] The output control data includes at least one of sorting control data, abnormal output control data, and sequence compensation data.

[0209] Sorting control data may include target sorting path number, sorting action time, sorting actuator action type, action duration, action speed, action amplitude, sorting confirmation signal, and post-sorting tracking identifier.

[0210] Abnormal output control data may include abnormal output path number, abnormal cause code, abnormal level, manual review prompt information, temporary storage location, abnormal package number, abnormal occurrence workstation, abnormal occurrence time, and abnormal handling strategy.

[0211] Sequence compensation data may include missing identifiers, delayed regression identifiers, substitution insertion identifiers, set waiting identifiers, batch traceability compensation identifiers, order queue correction identifiers, target sorting path rearrangement identifiers, and downstream cache adjustment identifiers.

[0212] Specifically, when the identification confirmation status meets the preset identification confirmation conditions and the sequence regression status meets the preset sequence regression conditions, the controller generates sorting control data and calculates the sorting action time based on the current position of the package, the conveying speed, the target sorting path position, and the response time of the sorting execution mechanism.

[0213] Preferably, the timing of the sorting action can be calculated using the remaining distance before the package reaches the sorting actuator, the current conveying speed, the response time of the sorting actuator, and the action lead time. The action lead time can be 1 to 3 times the response time of the sorting actuator.

[0214] Specifically, when the identification confirmation status does not meet the preset identification confirmation conditions and the sequence regression status meets the preset sequence regression conditions, the controller can generate abnormal output control data and record the abnormality as identification not confirmed. At the same time, it generates sequence compensation data to mark the abnormal status of the package in the production queue.

[0215] Specifically, when the identification confirmation status meets the preset identification confirmation conditions and the sequence regression status does not meet the preset sequence regression conditions, the controller can generate abnormal output control data or delayed regression control data, record the abnormal reason as the sequence cannot be regressed, and generate sequence compensation data to correct the downstream boxing or package combination queue.

[0216] Specifically, when the identification confirmation status does not meet the preset identification confirmation conditions and the sequence regression status does not meet the preset sequence regression conditions, the controller can generate abnormal output control data and send the package to the abnormal output path, manual review station or abnormal temporary storage area, while recording the abnormal reason as identification not confirmed and sequence cannot be regressed.

[0217] In one implementation, after the abnormal output mechanism sends the package into the abnormal output path, the controller can send an abnormal record to the production execution system. The abnormal record may include the temporary package number, collection time, data to be supplemented for identification, supplemented identification data, identification confirmation status, sequence regression status, abnormal cause code, and corresponding production sequence data.

[0218] In one implementation, when the abnormal package belongs to a set assembly, the controller can generate a set waiting identifier and control the downstream set assembly station to suspend the packaging action of the corresponding set until the manual review result or the replacement package result meets the preset conditions.

[0219] In one implementation, when an abnormal package is a critical component for batch traceability, the controller can generate a batch traceability compensation identifier and bind the abnormal output record of the package to the original batch number, order number, and target sorting path.

[0220] This step links the identification and confirmation results, sequence regression results, sorting actions, abnormal outputs, and sequence compensation by outputting control data. This enables the packages to be confirmed to enter the corresponding processing flow according to their actual status after verification, thereby improving the stability of continuous conveying, the reliability of sorting, and the controllability of abnormal handling.

[0221] Example 3:

[0222] The technical solution of this embodiment further refines the core inventive points of the present invention based on the above embodiments.

[0223] The core processing steps of this invention include generating supplementary identification data, generating sequence preservation data, generating verification control data, updating supplementary identification data, calculating sequence regression data, and generating output control data.

[0224] Among them, the data to be supplemented identification is used to transform the identification content that cannot be confirmed in the initial identification stage into data items that can be processed by the review action; the sequence preservation data is used to transform the batch, order, set and target arrival relationship in the production sequence into the constraint conditions in the review process; and the review control data is used to match the data to be supplemented, the review resources and the sequence constraints into the review path and review action.

[0225] In one implementation, the identification data to be supplemented can be stored in the form of a data table, queue, key-value pairs or structured data packets. The data fields of the identification data to be supplemented may include package number, type of item to be supplemented, location of item to be supplemented, confidence value of item to be supplemented, risk level of item to be supplemented, recommended review type, target sorting path and corresponding production sequence identifier.

[0226] In one implementation, the sequence-preserving data can be stored in the form of a queue constraint table, a time window constraint table, a position constraint table, or a logical relation table. The data fields of the sequence-preserving data may include package number, target queue position, target arrival time, allowed review time, allowed regression position range, adjacent package number, batch constraint, order constraint, and set constraint.

[0227] In one implementation, the review control data can be stored in the form of an action control package. The data fields of the review control data may include package number, review path number, review workstation number, review action type, review action parameters, estimated review time, estimated return location, estimated return time, resource usage identifier, and return control identifier.

[0228] Specifically, when generating review control data, the controller can first establish a set of candidate review actions based on the type of the item to be supplemented, then filter the set of candidate review actions based on the sequence holding data, then sort the filtered candidate review actions based on the available review resource data, and take the candidate review actions that meet the preset conditions in the sorting results as review control data.

[0229] Preferably, the candidate verification action set may include supplementary lighting action, secondary imaging action, pose adjustment action, rescanning action, weight verification action, buffer waiting action, bypass delivery action, reflow insertion action, and abnormal temporary storage action.

[0230] Specifically, the action parameters for supplemental lighting can include light source brightness, light source angle, light source duration, light source color temperature, and polarizer angle; the action parameters for secondary imaging can include exposure time, gain, acquisition area, acquisition angle, and image resolution; the action parameters for pose adjustment can include rotation angle, rotation speed, limiting pressure, and contact area; and the action parameters for weight verification can include sampling number, filtering method, and weighing stabilization time.

[0231] Preferably, the review path number can correspond to the mainline bypass review path, short cache review path, long cache review path, backflow review path, or abnormal temporary storage path. The controller can select the corresponding path according to the allowed review time and allowed sequence offset.

[0232] In one implementation, when the allowable verification time is short and the missing items can be quickly supplemented by supplementary lighting or secondary imaging, the controller selects the short path verification method to reduce the time it takes for the package to leave the main line.

[0233] In one implementation, when the item to be supplemented requires pose adjustment or weight verification and the allowable sequence offset is large, the controller selects a bypass verification path and performs backflow insertion at the end of the bypass based on the regression candidate position.

[0234] In one implementation, when the risk level of the item to be supplemented is high and the review resources are insufficient, the controller selects an abnormal temporary storage path and generates abnormal output control data to prevent the package from continuing to enter the normal sorting chain.

[0235] This invention selects the identification confidence value, the risk level of the item to be supplemented, the coverage of the item to be supplemented, the allowed review time, the allowed sequence offset, the allowed exposure risk amount, the review resource occupancy status, and the sequence deviation data as core control indicators.

[0236] The identification confidence value is used to indicate the degree of confirmation of the package identity by a single or multiple identification sources; the risk level of the missing items is used to indicate the degree of impact of the missing identification content on sorting accuracy and traceability integrity; and the coverage of the missing items is used to indicate the ability of the candidate review action to supplement the missing identification data.

[0237] Allowable review time indicates the time margin that can be used for reviewing the package without disrupting the downstream production cycle. Allowable sequence offset indicates the acceptable range of queue position offset after package review. Allowable contact risk indicates the acceptable range of risk of damage to the appearance of the package caused by the review action.

[0238] The resource occupancy status is used to indicate the availability of the current verification station, supplementary lighting equipment, camera, pose adjustment mechanism, weighing module, buffer section and reflow section. The sequence deviation data is used to indicate the comprehensive deviation between the verified package and the target production position and target arrival time.

[0239] Preferred evaluation value of candidate review scheme It can be calculated using the following formula.

[0240] ,

[0241] in, This refers to the coverage of items that need to be supplemented. It can be obtained by subtracting the coverage of the item to be supplemented from 1. For sequence bias data, To the extent of exposure risk, The normalized verification time, To mitigate the risk of resource consumption, , , , and Let be the weighting coefficients, and each weight satisfies the following formula:

[0242] ,

[0243] Preferred, A value of 0.20 to 0.45 is acceptable. A value of 0.15 to 0.40 is acceptable. A value of 0.10 to 0.30 is acceptable. A value of 0.10 to 0.30 is acceptable. A value of 0.05 to 0.25 can be used when the item to be supplemented involves key identity verification. A higher value can be chosen when downstream production cycles are tight. and A higher value can be chosen when the packaging is easily damaged. A higher value can be taken.

[0244] Specifically, the controller can select the candidate review scheme with the lowest evaluation value E that meets the allowable review time, allowable sequence offset, and allowable contact risk as the review control data.

[0245] In one implementation, when the difference in evaluation values ​​among multiple candidate review schemes is less than a preset difference threshold, the controller may prioritize the candidate review scheme with lower resource consumption.

[0246] In one implementation, when multiple candidate verification schemes meet the identification and completion requirements but have different regression candidate positions, the controller can prioritize the candidate verification scheme whose regression candidate position deviates less from the target queue position.

[0247] Preferably, the preset difference threshold can be 0.02 to 0.10, and the resource occupation risk can be calculated based on the remaining available time of the review station, the number of empty spaces in the buffer section, the idle status of the camera, the occupation status of the supplementary lighting equipment, and the occupation status of the return conveyor section.

[0248] In this embodiment, the recognition model can be used for the recognition processing of image data, batch number data, color number region data, contour data and pose data. The recognition model can be at least one of convolutional neural network, character recognition network, object detection network, image segmentation network, template matching model or multi-feature fusion model.

[0249] When using a recognition model, the input data can include the overall image of the packaging, the label area image, the identification code area image, the inkjet printing area image, the color code area image, the contour feature map, the pose angle data, and the position data. The output data can include the product category, the location of the identification code area, the batch number character result, the color code result, the contour matching result, the pose classification result, and the corresponding confidence value.

[0250] Preferably, before the image is input into the model, it can be subjected to size unification, grayscale conversion, color space conversion, brightness normalization, noise filtering, distortion correction, background removal, target area cropping, and reflective area suppression.

[0251] Specifically, size unification can scale the input image to a preset pixel size, brightness normalization can normalize the pixel intensity to the range of 0 to 1, noise filtering can use median filtering, Gaussian filtering or bilateral filtering, and target area cropping can be completed based on the detection box, code area positioning box, inkjet area positioning box or color number area positioning box.

[0252] Preferably, the training data may include legally authorized packaging sample images, packaging images collected and authorized for use in the enterprise's production line, publicly authorized barcode sample images, publicly authorized character sample images, and labeled color code area images.

[0253] Specifically, the annotation information of the training data may include product category annotation, identification code area box, batch number character box, color number area box, contour category, pose category, reflective area mask, and abnormal sample type.

[0254] Preferably, model training may include pre-training, sample augmentation, supervised training, loss calculation, parameter update and validation evaluation. The learning rate may be set to 0.0001 to 0.01, the batch size may be set to 8 to 64, the number of training epochs may be set to 20 to 300, and the optimizer may be SGD, Adam or AdamW.

[0255] Specifically, classification tasks can use the cross-entropy loss function, detection tasks can use a combination of classification loss, bounding box regression loss, and confidence loss, character recognition tasks can use CTC loss or sequence cross-entropy loss, and reflective region segmentation tasks can use Dice loss or cross-entropy loss.

[0256] Preferably, both the model training data and the input data are from compliant sources. The data can be obtained from legally authorized public datasets, production images collected with enterprise authorization, and business data authorized by user anonymization. When order information or traceability information is involved in the data, it has been desensitized. No personal sensitive information or biometric data has been collected in violation of regulations.

[0257] The recognition model in this embodiment is only an optional recognition method. The controller can also use rule algorithms, template matching algorithms, traditional machine vision algorithms, character recognition algorithms or multi-sensor fusion algorithms to realize the calculation of recognition confidence value, generation of supplementary items and processing of supplementary recognition data.

[0258] Through the above settings, the present invention integrates the identification and supplementation of packages to be confirmed, the selection of verification resources, the sequence regression judgment and the output compensation control into the same control process, so that the verification action can directly serve the sorting reliability and queue stability.

[0259] Example 4:

[0260] like Figure 2 As shown, this embodiment provides an intelligent conveying and sorting system for a cosmetic packaging line, which is used to perform the methods in any of the above embodiments.

[0261] Specifically, the system includes a data acquisition mechanism, a verification and conveying mechanism, a verification and identification mechanism, a sorting execution mechanism, an anomaly output mechanism, and a controller. The controller is connected to the data acquisition mechanism, the verification and conveying mechanism, the verification and identification mechanism, the sorting execution mechanism, and the anomaly output mechanism, respectively.

[0262] The data acquisition mechanism is used to acquire initial identification data and production sequence data of cosmetic packaging. The data acquisition mechanism may include at least two of the following: industrial camera, barcode scanner, inkjet reader, color sensor, contour detection sensor, position sensor, conveyor encoder, RFID reader, and interface module for communicating with the production execution system.

[0263] Preferably, the industrial camera can be set above, to the side, diagonally above, or before the sorting station; the barcode scanner can be set to correspond to the packaging identification code area; the inkjet reader can be set to correspond to the packaging batch number area; the color sensor can be set to correspond to the color number area; and the position sensor and conveyor encoder are used to determine the real-time position of the packaging on the conveyor line.

[0264] Specifically, the data acquisition mechanism can be triggered by the entrance detection sensor when the package arrives at the initial identification station, or the controller can generate an acquisition trigger signal based on encoder pulses, conveying speed, and target acquisition position.

[0265] The verification conveying mechanism is used to convey cosmetic packaging pieces whose status needs to be confirmed and identified according to verification control data. The verification conveying mechanism may include at least one of the following: bypass conveying section, buffer conveying section, return conveying section, guiding mechanism, stop mechanism, diversion mechanism, cycle adjustment mechanism, and return insertion mechanism.

[0266] Preferably, the bypass conveying section is located on one side of the main conveyor line. The inlet of the bypass conveying section can be connected to the diversion mechanism, and the outlet of the bypass conveying section can be connected to the return conveying section. The buffer conveying section can be located between the bypass conveying section and the verification and identification mechanism to temporarily store packages waiting for verification or waiting for return.

[0267] Specifically, the guiding mechanism may include a deflector, a guide wheel, a guide belt, or a steering roller; the blocking mechanism may include a cylinder blocking component, an electric blocking component, a lifting baffle, or a servo blocking component; and the return insertion mechanism may include a return belt, a pushing mechanism, a merging guide component, or a cycle release mechanism.

[0268] The verification and identification mechanism is used to acquire supplementary identification data. The verification and identification mechanism may include at least one of the following: a supplementary lighting device, a secondary imaging camera, a polarization imaging component, a local magnification imaging component, a pose adjustment mechanism, a weight verification module, a rescanning module, and a verification station sensor.

[0269] Preferably, the supplementary lighting device may include a ring light source, a strip light source, a surface light source, a coaxial light source, a polarized light source, or a multi-angle combination light source. The supplementary lighting device is connected to the controller and adjusts the brightness, angle, irradiation area, and duration according to the verification action data.

[0270] Preferably, the secondary imaging camera can be set above or to the side of the verification and identification station, and the acquisition area of ​​the secondary imaging camera corresponds to the identification code area, batch number area, color number area, label area or packaging outline area.

[0271] Preferably, the posture adjustment mechanism may include a lifting turntable, a flexible limiting member, a low-friction roller, a rotating carrier, a guide baffle, or an airflow assist member. The posture adjustment mechanism is used to adjust the orientation or angle of the package under the condition of limiting lateral friction drive.

[0272] Preferably, the weight verification module can be set in the weighing section or buffer section of the verification conveying mechanism. The weight verification module may include a weighing sensor, a weighing platform, a dynamic weighing controller and a vibration compensation module.

[0273] Sorting execution mechanisms are used to control cosmetic packaging to enter the target sorting path based on sorting control data. Sorting execution mechanisms may include swing arm sorting mechanisms, push rod sorting mechanisms, guide plate sorting mechanisms, roller sorting mechanisms, belt steering mechanisms, multi-channel sorting conveyor mechanisms, or robotic arm sorting mechanisms.

[0274] Preferably, the sorting execution mechanism is located after the verification return position or at the end of the main conveyor line. The sorting execution mechanism performs sorting actions according to the target sorting path number, the current position of the package, the conveying speed, and the time of the sorting action, and feeds back the action completion signal to the controller.

[0275] An abnormal output mechanism is used to control cosmetic packaging to enter the abnormal output path according to abnormal output control data. The abnormal output mechanism may include an abnormal rejection channel, a manual verification station, a temporary storage buffer area, an abnormal tray area, an isolation conveyor section, or an abnormal recycling container.

[0276] Preferably, the abnormal output mechanism can be set after the verification conveyor, before the sorting execution mechanism, or at the end of the main conveyor line. After the abnormal output mechanism receives the package, the controller records the abnormal cause code, abnormal output time, abnormal output location, data to be supplemented for identification, supplemented identification data, and corresponding production sequence data.

[0277] The controller may include a data receiving module, a data binding module, an identification status judgment module, a supplementary identification generation module, a sequence preservation generation module, a review resource management module, a review control generation module, a supplementary identification processing module, a sequence regression judgment module, an output control module, an exception recording module, and a communication module.

[0278] The data receiving module is used to receive image data, identification code data, batch number data, color code area data, contour data, pose data, position data, and production sequence data uploaded by the data acquisition agency.

[0279] The data binding module is used to bind the initial identification data and production sequence data of the same package into the same package data record based on timestamp, workstation number, encoder pulse, pallet number, RFID number or virtual queue number.

[0280] The recognition status judgment module is used to obtain at least two recognition confidence values ​​based on the initial recognition data, and to fuse the at least two recognition confidence values ​​to obtain comprehensive recognition data and recognition status.

[0281] The supplementary identification generation module is used to determine at least one of the following based on the initial identification data: unreadable identification code data, unreadable batch number data, invisible color number area data, reflective interference data, abnormal contour matching data, abnormal pose data, and production sequence conflict data, and generate supplementary identification data.

[0282] The sequence retention generation module is used to obtain target sequence position data based on the target production location and target arrival time, obtain logical sequence constraint data based on batch number, set number and order number, and obtain sequence retention data based on the target sequence position data and logical sequence constraint data.

[0283] The review resource management module is used to acquire supplementary lighting resource data, secondary imaging resource data, pose adjustment resource data, weight review resource data, cache resource data, reflow path resource data, and review workstation occupancy data.

[0284] The review control generation module is used to determine the required review type based on the data to be supplemented, determine the allowed review constraints based on the sequence preservation data, and generate review path data and review action data based on the required review type, allowed review constraints and available review resource data.

[0285] The supplementary identification processing module is used to receive supplementary light image data, secondary imaging data, pose-adjusted image data, weight verification data, and rescanning data based on the verification control data, and to update the data to be supplemented based on the supplementary identification data.

[0286] The sequence regression judgment module is used to obtain review time data, regression candidate position data and regression candidate time data based on the review control data, to obtain target production position data and target arrival time data based on the production sequence data, and to obtain sequence regression data and sequence regression status based on the above data.

[0287] The output control module is used to generate sorting control data, abnormal output control data and sequence compensation data based on the identification confirmation status and sequence regression status, and to control the sorting execution mechanism, abnormal output mechanism and downstream production queue to perform corresponding actions respectively.

[0288] The anomaly recording module is used to record the cause of the anomaly, the abnormal workstation, the abnormal time, the abnormal path, the identification data to be supplemented, the supplemented identification data, the identification confirmation status, the sequence regression status, and the sequence compensation result of the package to be confirmed.

[0289] The communication module is used to communicate with the production execution system, order management system, batch traceability system, packaging line control system, sorting execution mechanism controller, abnormal output mechanism controller, or host computer.

[0290] In this embodiment, after the packaged item enters the initial identification station, the data acquisition mechanism collects the initial identification data and synchronously obtains the production sequence data through the communication module. The data binding module binds the initial identification data and the production sequence data into the same packaged item data record.

[0291] The identification status judgment module calculates the identification confidence value and obtains the identification status based on the initial identification data. When the identification status is the identification status to be confirmed, the identification supplement generation module generates identification data to be supplemented, and the sequence preservation generation module generates sequence preservation data.

[0292] The verification control generation module generates verification control data based on the identification data to be supplemented, the sequence preservation data, and the available verification resource data provided by the verification resource management module, and controls the verification conveying mechanism to transport the package to the corresponding verification path and verification station.

[0293] The verification and identification agency performs supplementary lighting, secondary imaging, pose adjustment, rescanning or weight verification based on the verification action data, and sends the supplementary identification data to the supplementary identification processing module.

[0294] The supplementary identification processing module updates the data to be supplemented based on the supplementary identification data, and the sequence regression judgment module calculates the sequence regression data based on the review control data and the production sequence data, and obtains the identification confirmation status and the sequence regression status respectively.

[0295] When both the identification confirmation status and the sequence regression status meet the corresponding conditions, the output control module generates sorting control data and controls the sorting execution mechanism to send the package into the target sorting path.

[0296] When at least one of the identification confirmation status and sequence regression status fails to meet the corresponding conditions, the output control module generates abnormal output control data and sequence compensation data, and controls the abnormal output mechanism to send the package into the abnormal output path. At the same time, the sequence compensation result is sent to the production execution system or batch traceability system through the communication module.

[0297] Therefore, this embodiment enables cosmetic packaging to complete targeted verification, sequence maintenance, regression judgment, sorting output, and anomaly compensation in the pending confirmation and identification state through the coordinated cooperation between the data acquisition mechanism, the verification and conveying mechanism, the verification and identification mechanism, the sorting execution mechanism, the anomaly output mechanism, and the controller, and adapts to the automated production needs of multi-category, small-batch, similar-looking, combined sets, and mixed-line conveying.

[0298] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention. Those skilled in the art can combine, replace or modify the technical features in the above embodiments without departing from the concept of the present invention, and such combinations, replacements or modifications should all fall within the scope of protection of the present invention.

Claims

1. An intelligent conveying and sorting method for a cosmetic packaging line, characterized in that, Includes the following steps: S1. Obtain initial identification data and production sequence data for cosmetic packaging; S2. Obtain the identification status of the cosmetic packaging based on the initial identification data. When the identification status is a pending identification status, obtain supplementary identification data and sequence maintenance data based on the initial identification data and the production sequence data. S3. Determine the verification control data based on the identification data to be supplemented and the sequence preservation data, and obtain the supplementary identification data according to the verification control data; S4. Update the identification data to be supplemented based on the supplementary identification data, and obtain sequence regression data based on the review control data and the production sequence data; S5. Obtain the identification confirmation status based on the updated identification data to be supplemented, and obtain the sequence regression status based on the sequence regression data; S6. Generate output control data based on the identification confirmation status and the sequence regression status.

2. The intelligent conveying and sorting method for a cosmetic packaging line according to claim 1, characterized in that, The initial identification data includes at least two of the following: image data, identification code data, batch number data, color code area data, contour data, pose data, and position data. The identification status of the cosmetic packaging is obtained based on the initial identification data, including: At least two identification confidence values ​​are obtained based on the initial identification data; The at least two identification confidence values ​​are fused to obtain comprehensive identification data; When the comprehensive identification data is lower than the preset identification threshold, the identification status to be confirmed is obtained.

3. The intelligent conveying and sorting method for a cosmetic packaging line according to claim 1, characterized in that, Based on the initial identification data, supplementary identification data is obtained, including: Based on the initial identification data, at least one of the following is determined: unreadable identification code data, unreadable batch number data, invisible color number area data, reflective interference data, and abnormal contour matching data. Generate corresponding items to be supplemented based on the determined data; The items to be supplemented are associated to obtain the identification data to be supplemented.

4. The intelligent conveying and sorting method for a cosmetic packaging line according to claim 1, characterized in that, The production sequence data includes at least two of the following: batch number, set number, order number, target sorting path, target production location, and target arrival time. Based on the production sequence data, sequence preservation data is obtained, including: Target sequence location data is obtained based on the target production location and target arrival time; Logical sequence constraint data is obtained based on the batch number, set number, and order number; The sequence preservation data is obtained based on the target sequence position data and the logical sequence constraint data.

5. The intelligent conveying and sorting method for a cosmetic packaging line according to claim 1, characterized in that, Based on the identification data to be supplemented and the sequence-preserving data, the verification control data is determined, including: The required type of review is determined based on the data to be supplemented. Determine the allowable verification constraints based on the sequence preservation data; Obtain available review resource data; Based on the required review type, allowed review constraints, and available review resource data, generate review path data and review action data; The review path data and the review action data are associated as the review control data.

6. The intelligent conveying and sorting method for a cosmetic packaging line according to claim 5, characterized in that, The permitted review constraints include at least two of the following: permitted review time, permitted sequence offset, and permitted exposure risk amount; The available verification resource data includes at least one of supplementary lighting resource data, secondary imaging resource data, pose adjustment resource data, and weight verification resource data. When generating the review control data, candidate review data that exceeds the allowed review time, the allowed sequence offset, or the allowed contact risk exceeds the allowed contact risk are removed.

7. The intelligent conveying and sorting method for a cosmetic packaging line according to claim 5, characterized in that, Supplementary identification data is obtained based on the aforementioned review and control data, including: Starting with the identification data to be supplemented, at least one of the following is obtained based on the verification action data: supplementary lighting image data, secondary imaging data, image data after pose adjustment, and weight verification data; The acquired data will be used as the supplementary identification data. The image data after pose adjustment is data collected after the pose of the cosmetic packaging is adjusted under the condition of restricting lateral friction driving on the appearance surface of the cosmetic packaging.

8. The intelligent conveying and sorting method for a cosmetic packaging line according to claim 1, characterized in that, Updating the identification data to be supplemented based on the supplementary identification data includes: The supplementary identification data is obtained based on the supplementary identification data; The supplementary identification data is matched with the identification data to be supplemented to obtain the data to be supplemented that can be deducted; Based on the deductible data to be supplemented, the corresponding data item in the identification data to be supplemented is deducted, and the updated identification data to be supplemented is obtained. When the updated data to be identified meets the preset identification confirmation threshold, an identification confirmation status that meets the preset identification confirmation conditions is obtained. When the updated data to be identified does not meet the preset identification confirmation threshold, an identification confirmation status that does not meet the preset identification confirmation conditions is obtained.

9. The intelligent conveying and sorting method for a cosmetic packaging line according to claim 1, characterized in that, Based on the review control data and the production sequence data, sequence regression data is obtained, including: Based on the aforementioned review control data, review time data, regression candidate position data, and regression candidate time data are obtained; Based on the production sequence data, target production location data and target arrival time data are obtained; Based on the verification time data, regression candidate location data, regression candidate time data, target production location data, and target arrival time data, sequence deviation data is obtained, and the sequence deviation data is used as the sequence regression data. When the sequence deviation data does not exceed the preset sequence deviation threshold, a sequence regression state that satisfies the preset sequence regression condition is obtained; When the sequence deviation data exceeds the preset sequence deviation threshold, a sequence regression state that does not meet the preset sequence regression conditions is obtained; The process of generating output control data based on the identification confirmation status and the sequence regression status includes: When the identification confirmation status meets the preset identification confirmation conditions and the sequence regression status meets the preset sequence regression conditions, sorting control data is generated; When at least one of the identification confirmation state and the sequence regression state fails to meet the corresponding conditions, abnormal output control data and sequence compensation data are generated.

10. An intelligent conveying and sorting system for a cosmetic packaging line, characterized in that, include: Data acquisition agencies are used to obtain initial identification data and production sequence data for cosmetic packaging. A verification conveying mechanism is used to convey cosmetic packaging pieces whose status needs to be confirmed and identified based on verification control data; A verification and identification agency is used to obtain supplementary identification data; The sorting execution mechanism is used to control the cosmetic packaging to enter the target sorting path according to the sorting control data; An abnormal output mechanism is used to control the cosmetic packaging to enter the abnormal output path based on abnormal output control data. The controller is connected to the data acquisition mechanism, the verification and conveying mechanism, the verification and identification mechanism, the sorting execution mechanism, and the abnormal output mechanism, respectively. The controller is used to execute the method according to any one of claims 1 to 9.

Images