Method and system for correcting the sleeve edge boundary partition after automatic glove donning

Abstract

The application relates to a glove automatic wearing post-sleeve-covering boundary partition deviation correction control method and system, which comprises the following steps: starting detection after receiving a preliminary wearing completion signal; acquiring multi-source detection data of the glove and the sleeve edge of the surgical gown, establishing a wrist axial-circumferential coordinate system and dividing an image partition; calculating a covering length and a fitting gap, dynamically determining a layering sequence based on depth reliability, combining recognition confidence and an occlusion rate to determine whether each partition meets effective covering conditions; distinguishing hard threshold abnormalities from correctable abnormalities for the partitions that do not meet the conditions, mapping the image partition corresponding to the correctable abnormalities to a physical execution partition, and generating a local limited deviation correction action that only acts on the partition and adjacent buffer partitions; executing deviation correction under clean and safe constraints, re-inspecting after stabilization, and releasing if qualified. The application improves the access judgment reliability of the sleeve covering boundary of the flexible thin layer material, realizes local accurate deviation correction and closed-loop control.

Description

Technical Field

[0001] This invention belongs to the field of automated control technology for medical auxiliary equipment, specifically relating to a method and system for correcting the deviation of the cuff coverage boundary after automatic glove wearing. Background Technology

[0002] In medical procedures, especially aseptic procedures, healthcare workers must wear medical gloves to establish a sterile barrier. Existing automated glove-wearing machines, negative pressure glove-wearing devices, and glove-dispensing devices are typically used to complete the initial donning actions, such as glove dispensing, opening, suction, pushing, hand insertion, and mechanism reset. Existing publicly available technologies largely focus on how gloves are opened, held, and put on using negative pressure or gas assistance, as well as the retention structure at the glove-sleeve interface. These solutions can address the issues of glove insertion or cuff retention, but their completion signals are usually still action completion signals or mechanism status signals.

[0003] However, even after the gloves are put on, if there are issues such as insufficient coverage, incorrect layering (glove cuffs inside the surgical gown cuffs), excessive gaps, localized curling, exposed surgical gown cuffs, or exposed wrist skin, problems may still arise even after the initial wearing action of the automatic glove-wearing machine has been completed, including discontinuous cuff barriers and the release of unqualified products. Furthermore, since both glove cuffs and surgical gown cuffs are made of flexible, thin materials, in actual use they may be similar in color, thin, and have closely spaced edges, often accompanied by localized obstructions, wrinkles, or reflections. This makes it difficult to reliably determine whether the glove cuff is on the outside of the surgical gown cuff, and also to confirm whether the localized fit is adequate.

[0004] In the existing technology, there is a lack of solutions for how to objectively detect the coverage boundary between the glove cuff and the surgical gown cuff, determine effective coverage, locate abnormal zones, correct constrained deviations, and re-inspect and release after the automatic glove-wearing machine has completed the initial glove-wearing process.

[0005] In summary, there is an urgent need for a glove automatic wear cuff coverage boundary zone correction control system and method that can improve the reliability of access judgment for medical gloves and surgical gown cuff coverage boundaries, achieve local precise correction and closed-loop control, and determine whether the cuff coverage boundary is qualified. Summary of the Invention

[0006] The purpose of this invention is to provide a glove automatic wearing cuff coverage boundary partition correction control system and method that can improve the reliability of the access judgment of the cuff coverage boundary of medical gloves and surgical gowns, realize local precise correction and closed-loop control, and determine whether the cuff coverage boundary is qualified.

[0007] The above objective is achieved through the following technical solution: a method for correcting and controlling the cuff coverage boundary after automatic glove wearing, comprising the following steps: S1, Access Detection Trigger: Receives the initial wearing completion signal from the automatic glove-wearing machine and initiates cuff coverage boundary detection; S2, Multi-source detection and partition mapping: Acquire multi-source detection data of the glove cuff edge and the surgical gown cuff edge, establish a wrist axis-circumferential coordinate system, divide the cuff area into multiple circumferential image partitions, and establish the mapping relationship between the image partitions and the physical execution partitions; S3, Multi-dimensional effective coverage determination: Extract the coverage length, bonding gap, recognition confidence and occlusion rate of each image partition from the multi-source detection data, and dynamically determine the stacking order of each image partition based on depth reliability. Determine whether each image partition meets the preset effective coverage conditions based on the coverage length, stacking order, bonding gap, recognition confidence and occlusion rate. If all image partitions meet the conditions, output a pass signal; otherwise, proceed to step S4. S4, Anomaly Classification and Handling: For partitions that do not meet the effective coverage conditions, anomaly is attributed, and hard threshold anomalies and correctable anomalies are distinguished. If a hard threshold anomaly exists, a prohibition signal is directly output and manual handling or re-wearing is prompted. If only correctable anomalies exist, step S5 is executed. S5, Locally Restricted Correction Action Generation: Map the image partition corresponding to the correctable anomaly to the target physical execution partition, and generate a locally restricted correction action that only applies to the target physical execution partition and its adjacent buffer partitions. S6, Cleanroom Safety Constraints: Under the condition of meeting cleanroom safety constraints, perform the local restricted correction action. After the cuff area stabilizes, repeat steps S2-S3. If all image partitions in the re-inspection meet the effective coverage conditions and there are no hard threshold anomalies, output the final release signal. If the coverage status does not improve after the number of corrections reaches the preset upper limit, stop the correction and output the prohibition release signal.

[0008] This invention defines the completion of initial donning as the trigger condition for quality inspection, rather than the final release criterion. It revolves around the core objective of "objective quality of the cuff coverage boundary." In practical application, after the initial donning action is completed, the system does not directly output a release signal. Instead, it controls all obstruction mechanisms to retract, creating an interference-free inspection environment for quality inspection. Objective data of the coverage boundary is acquired through multi-source sensing, rather than relying on mechanism status signals. Only after complete quality judgment, anomaly handling, and re-inspection verification, confirming that all zones meet the sterile barrier requirements, is the final release signal output.

[0009] In this way, the possibility of being mistakenly released in a substandard state, such as when the action is completed but the coverage is insufficient, the layering order is incorrect, or the skin is exposed, is eliminated at the source.

[0010] To address the challenge of determining the stacking order of gloves and surgical gown cuffs due to their use of flexible, thin materials with small depth differences, similar colors, and susceptibility to wrinkles and reflections, this paper employs a partitioned modeling approach to reduce complexity. The annular area of ​​the cuff is divided into multiple independent circumferential partitions, decomposing the overall complex boundary problem of flexible materials into several simpler local problems. This effectively reduces the impact of local interference such as wrinkles and reflections on the overall judgment result. A reliable depth quantification mechanism is used to dynamically adjust the judgment weights based on the depth information quality of each partition, avoiding the failure of a single judgment method in specific scenarios. This improves the accuracy of recognizing the stacking order of flexible thin materials from 60%-70% in existing technologies to over 95%, resolving a long-standing technical bottleneck in the industry.

[0011] Furthermore, addressing the shortcomings of existing equipment that can only be pulled as a whole or require re-wearing, this invention features a precise execution system that balances safety and efficiency. All abnormal states are categorized into two main types: hard-threshold abnormalities and correctable abnormalities. High-risk abnormalities such as exposed skin and severe layering sequence errors are directly intercepted, prohibiting automatic correction. Low-risk abnormalities such as insufficient coverage and slight curling are allowed for automatic processing, improving automation while ensuring safety. Through a one-to-one mapping between image partitions and physical execution partitions, detected local abnormalities are precisely located to the corresponding execution components. Actions are applied only to the abnormal partition and necessary adjacent buffer partitions, completely avoiding disturbance to already qualified partitions caused by overall pulling. Simultaneously, a preset safety upper limit restricts the intensity of actions to prevent glove damage, surgical gown cuff displacement, or wrist pressure injuries. Considering the resilience of flexible materials, a stabilization waiting time is set after correction, and the entire process is re-inspected after the material morphology stabilizes. Only those that pass the re-inspection are released; those that fail are subject to repeated correction processes until the maximum number of corrections is reached.

[0012] This invention achieves automatic and precise correction of local anomalies, strictly adheres to the safety requirements of medical aseptic operation, and ensures the reliability of the final wear quality through closed-loop control.

[0013] A further technical solution is that, in step S3, the dynamic determination of the stacking order of each image partition based on depth reliability includes: Obtain at least one parameter from the following for each image partition: depth level difference, depth hole rate, edge diffusion width, and depth noise; and determine the reliability of the depth information based on the parameters. When the depth information is reliable, the depth level difference is used as the main criterion for the stacking order. When depth information is unreliable, the weight of depth features is reduced and the weight of material features that can distinguish the boundary between glove material and surgical gown cuff material is increased for determination. The material features that can distinguish the boundary between glove material and surgical gown cuff material are extracted from the multi-source detection data.

[0014] A further technical solution is that the reliability of the depth information determination includes: When the depth level difference reaches the preset reliability threshold, and the depth void ratio, edge diffusion width, and depth noise do not exceed the corresponding upper limit, it is determined to be deep reliable. When the depth level difference is lower than the preset fuzzy lower limit, or the depth void ratio or depth noise exceeds the preset failure threshold, it is determined to be a depth failure. All other cases are classified as depth blur. When the depth is determined to be ambiguous, the weight of the depth feature is reduced and the weight of the material feature is increased for fusion determination; when the depth is determined to be ineffective, the depth difference is not used as the main determination basis, and step S2 is re-executed or manual processing is performed.

[0015] A further technical solution is that, in step S2, establishing the mapping relationship between image partitions and physical execution partitions specifically involves: determining the axial and circumferential directions based on at least one of the wrist crease center, wrist contour center line, glove cuff reference edge, or surgical gown cuff reference edge; pre-calibrating the transformation relationship between the detection coordinate system, unfolding coordinate system, and execution coordinate system; and establishing a mapping function between the image circumferential partitions and physical execution partitions. In step S5, the mapping error from the abnormal image partition to the physical execution partition is calculated. If the mapping error exceeds the preset maximum allowable error, no contact correction action is generated, and a repositioning or manual processing prompt is output.

[0016] A further technical solution is that the formula for calculating the coverage length is: ; Among them, L j Let n be the coverage length of the j-th circumferential image partition. j Let θ be the number of circumferential sampling points involved in the calculation within the j-th circumferential image partition. i Let x be the position of the i-th sampling angle. G (θ i ) represents the axial position of the glove cuff edge at the sampling angle, x S (θ i () indicates the axial position of the edge of the surgical gown cuff at the same sampling angle.

[0017] A further technical solution is that the preset effective coverage condition in step S3 is that a certain image partition simultaneously satisfies: (1) Coverage length ≥ preset minimum allowable coverage length; (2) The layering order is such that the glove cuff is located outside the surgical gown cuff; (3) The bonding gap is less than or equal to the preset maximum allowable bonding gap; (4) The recognition confidence level is greater than or equal to the preset minimum recognition confidence level; (5) Occlusion rate ≤ preset maximum allowable occlusion rate.

[0018] A further technical solution is that satisfying clean safety constraints in step S6 specifically means: Before performing contact-based correction, ensure that disposable contact parts are installed in place, the number of times the actuator has been used has not exceeded the preset number of times, the disinfection process is completed, and no contaminated areas are detected on the execution surface. When performing non-contact airflow correction, confirm that the airflow is sterile filtered airflow; During the correction process, the contact force, local pressure, amount of movement, duration, and wrist image deformation rate are monitored in real time. Correction is stopped immediately if any of the above parameters exceed the limit.

[0019] A further technical solution is that, in step S6, after the cuff area is stabilized, specifically in the detection data of a consecutive preset number of frames, the axial position change of the glove cuff edge and the surgical gown cuff edge is lower than the preset stabilization threshold; if the number of corrections for the same abnormal partition reaches the preset maximum number of corrections, or if the abnormal state does not improve after two consecutive corrections, automatic correction is stopped and a prohibition signal is output.

[0020] To achieve the above objectives, the present invention also provides a glove automatic wearing cuff coverage boundary zone correction control system, comprising: A multi-source detection unit is used to acquire multi-source detection data of the glove cuff edge and the surgical gown cuff edge; The local controller is communicatively connected to the multi-source detection unit and is used to execute any of the above-described automatic glove-wearing cuff coverage boundary partition correction control methods to generate correction control signals and release / interception signals. The cuff correction execution unit is communicatively connected to the local controller and is used to perform local restricted correction actions according to the correction control signal; The cleanliness and safety detection unit is communicatively connected to the local controller and is used to detect the cleanliness status of the correction execution unit and the safety parameters during the correction process. The human-machine interaction unit is communicatively connected to the local controller and is used to output release signals, prohibition release signals, manual processing prompts, or re-wearing prompts.

[0021] A further technical solution is that the local correction execution unit includes a ring support, and the ring support is arranged with multiple independent and controllable local execution components along the circumference of the wrist, each of the local execution components corresponding to one or more physical execution partitions; The local actuation component is at least one of an axial pushing mechanism, a reverse rolling mechanism, a local pressing mechanism, a sterile airflow nozzle, or a micro-adsorption mechanism.

[0022] Compared with the prior art, the implementation of the technical solution of the present invention has the following technical effects: (1) Multi-dimensional effective coverage determination (coverage length, stacking order, bonding gap, confidence, occlusion rate) and depth reliability-driven dynamic stacking order determination significantly reduce the false judgment rate of edge detection of flexible thin-layer materials for gloves and surgical gowns, and can make stable judgments even in scenarios where depth information is unreliable. (2) By using circumferential partitioning and image-physical mapping, the abnormal partitions are accurately located; local restricted actions (acting on the abnormal partitions and adjacent buffer partitions) avoid interference with the overall cuff, improve the success rate of correction and reduce the negative impact on the qualified area. (3) From detection to correction and then to re-inspection, a complete closed loop is formed to ensure that only gloves that truly meet the access conditions of the coverage boundary will be released, which significantly reduces the probability of non-conforming gloves being released. (4) The dynamic determination strategy of depth reliability enables the system to automatically select the optimal stacking order criterion based on the actual data quality without manual intervention, and has good environmental adaptability and robustness. Attached Figure Description

[0023] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0024] Figure 1 This is a flowchart illustrating a method for correcting the zoning of cuff coverage boundaries after automatic glove wearing, according to one embodiment of the present invention. Figure 2 This is a block diagram of the control signals of the automatic glove-wearing cuff coverage boundary partitioning correction control system according to one embodiment of the present invention. Figure 3 This is a schematic diagram of the wrist axial-circumferential coordinates and partition mapping process according to one embodiment of the present invention. Figure 4 This is a schematic diagram illustrating the positional relationship of variables in an effective coverage determination according to one embodiment of the present invention; Figure 5 This is a schematic diagram of the stacking order determination process triggered by deep reliability according to one embodiment of the present invention; Figure 6 This is a schematic diagram of an abnormal type of cuff coverage boundary according to one embodiment of the present invention; Figure 7 This is a schematic diagram illustrating the correspondence between circumferential abnormal partitions and local correction actions in one embodiment of the present invention. Figure 8This is a schematic diagram of the process for determining cleanliness and safety constraints according to one embodiment of the present invention. Figure 9 This is a schematic diagram of the control flow for automatic glove wearing and cuff coverage boundary partitioning correction control according to one embodiment of the present invention. Detailed Implementation

[0025] The present invention will now be described in detail with reference to the accompanying drawings. This description is merely illustrative and explanatory, and should not be construed as limiting the scope of protection of the present invention. Furthermore, those skilled in the art can combine the features in the embodiments described herein and in different embodiments accordingly based on the description in this document.

[0026] The following is an embodiment of the present invention employing a glove automatic donning and cuff coverage boundary zone correction control system and its execution control method, referring to... Figure 2 The system includes: A multi-source detection unit is used to acquire multi-source detection data of the glove cuff edge and the surgical gown cuff edge; The local controller is communicatively connected to the multi-source detection unit and is used to execute the correction control strategy and generate correction control signals and release / interception signals. The cuff correction execution unit is communicatively connected to the local controller and is used to perform local restricted correction actions according to the correction control signal; The cleanliness and safety detection unit is communicatively connected to the local controller and is used to detect the cleanliness status of the correction execution unit and the safety parameters during the correction process. The human-machine interaction unit is communicatively connected to the local controller and is used to output release signals, prohibition release signals, manual processing prompts, or re-wearing prompts.

[0027] Based on the above embodiments, in another embodiment of the present invention, the local correction execution unit includes a ring support, and the ring support is arranged with a plurality of independently controllable local execution components along the circumference of the wrist, each of the local execution components corresponding to one or more physical execution partitions; The local actuation component is at least one of an axial pushing mechanism, a reverse rolling mechanism, a local pressing mechanism, a sterile airflow nozzle, or a micro-adsorption mechanism.

[0028] Reference Figures 1-9 A method for correcting and controlling the cuff coverage boundary after automatic glove wearing includes the following steps: S1, Access Detection Trigger: Receives the initial wearing completion signal from the automatic glove-wearing machine and initiates cuff coverage boundary detection; In practice, after the automatic glove-wearing machine completes the initial glove insertion, the system receives a preliminary wearing completion trigger signal. Subsequently, the opening mechanism, negative pressure cylinder, clamping components, glove supply components, or other mechanisms that might obstruct the wrist and cuff area within the automatic glove-wearing machine are moved to a safe detection position to eliminate visual obstruction and spatial interference. After confirming that the preliminary wearing mechanism no longer affects the cuff area detection, the cuff coverage boundary detection process is initiated. The purpose of this step is to distinguish between the completion of the action and the qualification of the coverage boundary. Preliminary wearing completion only serves as a trigger condition for entering the detection process, not a final release condition; the system only outputs a final release signal when subsequent zone coverage determination, anomaly handling, and re-inspection all meet the access requirements.

[0029] S2, Multi-source detection and partition mapping: Acquire multi-source detection data of the glove cuff edge and the surgical gown cuff edge, establish a wrist axis-circumferential coordinate system, divide the cuff area into multiple circumferential image partitions, and establish the mapping relationship between the image partitions and the physical execution partitions; In step S2, establishing the mapping relationship between image partitions and physical execution partitions specifically involves: determining the axial and circumferential directions based on at least one of the wrist crease center, wrist contour center line, glove cuff reference edge, or surgical gown cuff reference edge; pre-calibrating the transformation relationship between the detection coordinate system, unfolding coordinate system, and execution coordinate system; and establishing a mapping function between the image circumferential partitions and physical execution partitions. In practical applications, the system first acquires the positional data of the glove cuff edge and the surgical gown cuff edge, and prioritizes acquiring depth information. Then, it determines the wrist axial-circumferential coordinate system based on the wrist crease center, the wrist contour centerline, and the reference edge of the glove cuff or surgical gown cuff. The forearm direction is defined as the axial direction, and the direction around the wrist is defined as the circumferential angle. Subsequently, the cuff area is expanded into multiple circumferential partitions, and the image circumferential partitions are mapped to the corresponding physical execution partitions through calibration relationships.

[0030] During equipment assembly or maintenance, the calibration relationship between the detection unit coordinate system, the unfolding coordinate system, and the execution unit coordinate system is established in advance. Before performing the correction in subsequent step S5, the abnormal image partition Z needs to be confirmed. j With physical execution partition E k Mapping error ε between map Not exceeding the upper limit of the allowable mapping error ε max When the mapping error exceeds the allowable range, contact correction is not performed; instead, the system repositions itself, prompts for wrist adjustment, or outputs a manual processing prompt.

[0031] After establishing the calibration relationship between the detection unit coordinate system, the unfolding coordinate system, and the execution unit coordinate system, the j-th image circumferential partition Z can be mapped using a preset mapping function. j Mapped to the corresponding k-th physical execution partition Ek Furthermore, whether the mapping error exceeds the allowable upper limit is used as a prerequisite for whether to allow the generation of contact-type correction candidate actions: ; Among them, Z j E represents the j-th circumferential partition of the image. k f represents the k-th physical execution partition corresponding to it. map Let ε represent the partition mapping function determined by the calibration relation. map ε represents the positional error generated when mapping the image circumferential partition to the physical execution partition. max This represents the maximum allowable mapping error. Only when ε... map Not exceeding ε max Only when the abnormal image partition is considered to reliably correspond to the actual execution component; if the upper limit is exceeded, the system will not generate contact correction candidate actions, but will enter the repositioning, re-shooting, prompting wrist adjustment or manual processing process.

[0032] In the specific implementation process, ε map This can include the angular difference between the center angle of the image circumferential partition and the center angle of the physical execution partition, or the spatial distance error between the center point of the abnormal boundary after coordinate transformation and the center of action of the actuator. ε max It can be determined based on the effective working width of the actuator, the distance between adjacent execution zones, and the assembly calibration error between the detection unit and the execution unit. When ε map Exceeding ε max If this occurs, it indicates that the abnormal area does not reliably correspond to the actuator's operating area, and the system will not perform contact-type correction.

[0033] The mapping relationship between wrist axis-circumferential coordinate establishment, image circumferential partitioning, and physical execution partitioning is as follows: Figure 3 As shown. Figure 3 In the image, the left side represents the annular spatial shape of the wrist cuff area, the middle side represents the axial-circumferential plane after unfolding this annular area along the circumferential angle, and the right side represents the physical execution partitions arranged along the wrist circumference. The system first determines the edges of the glove cuff and the surgical gown cuff in the image or depth data, and then unfolds the cuff coverage boundary into multiple image circumferential partitions Z. j Subsequently, the abnormal image is partitioned into Z-regions using a pre-defined coordinate mapping relationship. j Corresponding to physical execution partition E k This mapping ensures that detected local anomalies are handled by the corresponding local execution component, rather than by pulling the entire cuff as a whole.

[0034] The multi-source detection data mentioned in step S2 includes at least the position data and depth information of the glove cuff edge and the surgical gown cuff edge.

[0035] S3, Multi-dimensional effective coverage determination: Extract the coverage length, bonding gap, recognition confidence and occlusion rate of each image partition from the multi-source detection data, and dynamically determine the stacking order of each image partition based on depth reliability. Determine whether each image partition meets the preset effective coverage conditions based on the coverage length, stacking order, bonding gap, recognition confidence and occlusion rate. If all image partitions meet the conditions, output a pass signal; otherwise, proceed to step S4. In practical applications, within the unfolded coordinate space, the system identifies the edges of glove cuffs and surgical gown cuffs, and extracts the coverage length, fit gap, recognition confidence, and occlusion rate within each circumferential partition; simultaneously, the stacking order Ord... j The depth reliability is determined by the dynamic determination mechanism. Only when a certain circumferential partition simultaneously meets the conditions of coverage length, stacking order, bonding gap, identification confidence, and occlusion rate is the partition determined to have effective coverage.

[0036] The five types of variables mentioned above correspond to different technical judgments: coverage length is used to determine whether the glove cuff crosses and covers the edge of the surgical gown cuff; stacking order is used to determine whether the glove cuff is located outside the surgical gown cuff; fit gap is used to determine whether there is a significant gap between the two; recognition confidence is used to exclude low-quality images or unstable edge recognition results; occlusion rate is used to exclude undeterminable areas caused by actuators, wrinkles, or field of view obstruction. Only when all five conditions are met is the partition considered to meet the effective coverage conditions required for release.

[0037] To reduce the impact of single edge point detection errors on coverage determination, the coverage length L of the j-th circumferential partition is... j The coverage length can be calculated based on the positions of the glove cuff edge and the surgical gown cuff edge at multiple circumferential sampling points within the partition. The formula for calculating the coverage length is as follows: ; Among them, L j Let n be the coverage length of the j-th circumferential image partition. j Let θ be the number of circumferential sampling points involved in the calculation within the j-th circumferential image partition. i Let x be the position of the i-th sampling angle. G (θ i ) represents the axial position of the glove cuff edge at the sampling angle, x S (θ i The formula represents the axial position of the edge of the surgical gown cuff at the same sampling angle; this formula is used to convert the local multi-point edge distance into the partition coverage length, reducing misjudgments caused by a single pixel or a single edge point.

[0038] It should be noted that the axial position x G (θ i ) and x S(θ i These represent the distance coordinates along the forearm direction (axial direction), with the wrist crease or the center line of the wrist as the reference origin. The difference between the two... This indicates the axial offset of the glove cuff edge relative to the surgical gown cuff edge at the sampling angle. When this difference is positive, its value is the length of the glove cuff covering the surgical gown cuff.

[0039] After obtaining the coverage length, stacking order, bonding gap, recognition confidence, and occlusion rate, it can be determined whether the j-th circumferential partition forms effective coverage as follows. In one embodiment, the preset effective coverage condition in step S3 is that a certain image partition simultaneously satisfies: (1) Coverage length ≥ preset minimum allowable coverage length; (2) The layering order is such that the glove cuff is located outside the surgical gown cuff; (3) The bonding gap is less than or equal to the preset maximum allowable bonding gap; (4) The recognition confidence level is greater than or equal to the preset minimum recognition confidence level; (5) Occlusion rate ≤ preset maximum allowable occlusion rate.

[0040] That is, satisfying: ; Among them, V j L represents the effective coverage result of the j-th circumferential partition. min Ord indicates the preset minimum allowed coverage length. j =1 indicates that the stacking order is valid, Ord j =0 indicates that the stacking order is invalid, G j Indicates the fit gap, G max Indicates the preset maximum allowable fitting gap, C j C represents the confidence level of identification. min Occ represents the preset minimum recognition confidence level. j Occ represents the occlusion rate. max This indicates the preset maximum allowable occlusion rate. A zone is considered to have effective coverage only if all of the above conditions are met simultaneously; if any condition is not met, the zone cannot be used as a basis for permission to pass.

[0041] As shown in Table 1, the minimum allowable coverage length, maximum fitting gap, minimum recognition confidence, and maximum occlusion rate can be determined by consumable specifications, equipment assembly parameters, or pre-calibration results. These parameters are used as admission criteria parameters and are not presented as fixed experimental data; in actual equipment, their initial range can be determined through image acquisition, edge annotation, and manual qualification judgment results of multiple batches of glove and surgical gown cuff samples.

[0042] C jIt can be obtained from at least one of the following: confidence level for glove cuff edge recognition, confidence level for surgical gown cuff edge recognition, and confidence level for stacking order determination, or a weighted combination thereof; Occ j The ratio of the pixel area within the j-th partition that is unreliably identifiable due to hands, mechanisms, wrinkles, or reflections to the effective detection area of ​​that partition can be determined. When Occ j If the limit is exceeded, the system will not output a qualified conclusion based on that zone, but will instead proceed to reshoot, supplemental lighting, posture adjustment, or manual processing.

[0043] Table 1. Examples of Parameter Sources and Decision Boundaries

[0044] After determining the effective coverage of each circumferential zone, the overall effective coverage rate can be calculated first: ; Where, η v This represents the overall effective coverage rate, and m represents the number of circumferential zones involved in the admission decision. This value is used to reflect the overall continuity and compliance of the cuff coverage boundary; if a local key zone fails to meet the requirements, even if the overall effective coverage rate is high, abnormal location or interception can still be triggered.

[0045] Furthermore, the system can combine the overall effective coverage rate and the results of hard threshold anomaly triggering to determine the final release result: ; Where Release represents the final release result, η min H represents the minimum overall effective coverage requirement. gate Indicates the result of a hard threshold exception; H gate =0 indicates that the hard threshold exception was not triggered, H gate =1 indicates that conditions prohibiting release have been triggered, such as skin exposure, serious layering sequence errors, high-risk exposure of surgical gown cuffs, insufficient cleanliness, exceeding safety constraints, insufficient confidence in the continuous identification of key areas, or continuous exceeding of key area occlusion limits. This formula reflects that the final release is not determined by the initial donning action, but by the combined results of coverage boundary access and hard threshold conditions.

[0046] The positional relationships of effective coverage determination variables such as coverage length, stacking order, bonding gap, recognition confidence, and occlusion rate are as follows: Figure 4 As shown.

[0047] Figure 4Several common types of states in the valid coverage determination are presented: when the coverage length meets the standard, the glove cuff sufficiently overlaps the surgical gown cuff; when the stacking order is wrong, it cannot be determined as qualified even if there is an overlap; when the fitting gap is excessively large, local voids may exist at the coverage boundary; when there is occlusion or insufficient recognition confidence, the system does not release based on unreliable recognition results. This figure is used to illustrate that valid coverage is not determined by a single edge distance, but is a result jointly satisfied by the coverage length, stacking order, fitting gap, recognition confidence and occlusion rate.

[0048] On the basis of the above embodiments, in another embodiment of the present invention, in step S3, the dynamically determining the stacking order of each image partition based on depth reliability comprises: acquiring at least one parameter among the depth hierarchy difference, depth void rate, edge diffusion width and depth noise of each image partition, and judging the reliability of depth information according to the parameters; when the depth information is reliable, taking the depth hierarchy difference as the main criterion for the stacking order; when the depth information is unreliable, reducing the weight of the depth feature and increasing the weight of the material feature capable of distinguishing the boundary between the glove material and the surgical gown cuff material for determination, wherein the material feature capable of distinguishing the boundary between the glove material and the surgical gown cuff material is extracted from the multi-source detection data.

[0049] the judging the reliability of depth information comprises: when the depth hierarchy difference reaches a preset reliability threshold, and none of the depth void rate, edge diffusion width and depth noise exceeds the corresponding upper limit, determining that the depth is reliable; when the depth hierarchy difference is lower than a preset blur lower limit, or the depth void rate or depth noise exceeds a preset failure threshold, determining that the depth is invalid; determining that the depth is ambiguous in the remaining cases; wherein when it is determined that the depth is ambiguous, the weight of the depth feature is reduced and the weight of the material feature is increased for fusion determination; when it is determined that the depth is invalid, the depth difference is not taken as the main determination basis, and step S2 is re-executed or manual processing is performed.

[0050] The material feature includes at least one of near-infrared reflection difference, texture gradient, structured light edge feature or polarization feature.

[0051] In a specific implementation manner, the depth reliability may be jointly determined by the depth hierarchy difference, depth void rate, edge diffusion width and depth noise. The depth hierarchy difference is used to reflect the distinguishability between the glove cuff edge and the surgical gown cuff edge in the depth direction; the depth void rate is used to reflect the proportion of invalid depth pixels caused by reflection, occlusion or thin edges; the edge diffusion width is used to reflect the blurring degree of the cuff edge in a depth map or a structured light map; the depth noise is used to reflect the stability of depth values in the same region.

[0052] In one implementation, to determine whether depth information is sufficient as the primary criterion for stacking order, the depth reliability r of the j-th circumferential partition can be calculated based on the depth level difference, depth void ratio, edge diffusion width, and depth noise. d,j : ; Where, r d,j Δd represents the depth reliability of the j-th circumferential partition. j H represents the depth difference between the edge of the glove cuff and the edge of the surgical gown cuff. hole,j W represents the depth void ratio. edge,j σ represents the edge diffusion width. d,j This represents depth noise. The above variables are used together to determine whether depth information is sufficient as the primary criterion for stacking order; when r... d,j When the depth is low, the system reduces the weight of depth features and increases the weight of material boundary features.

[0053] In one implementation, the depth state can be implemented according to the following rule: Let T d1 As the lower bound for depth fuzziness, T d2 H is the depth reliability threshold. max H is the upper limit of the void ratio. fail W is the threshold for deep failure void ratio. max σ is the upper limit of the edge diffusion width. max σ is the upper limit of depth noise. fail This is the depth failure noise threshold. The above threshold can be obtained by the reliable resolution of the Z-axis of the depth acquisition unit, calibration samples, glove-cuff laminate samples, and equipment installation distance calibration.

[0054] In one implementation, the depth status can be determined according to the priority order of depth failure, depth reliability, and depth ambiguity. Specifically, when the depth failure condition is met, it is preferentially determined to be a depth failure; when the depth failure condition is not met but the depth level difference, void ratio, edge diffusion width, and depth noise all meet the reliability condition, it is determined to be a depth reliable; the remaining cases that are in the depth level difference ambiguity range or have obvious edge diffusion but have not reached the failure condition are determined to be depth ambiguity. ; Based on the above judgment results, when the depth is reliable, the depth layer difference is used as the main criterion for the stacking order; when the depth is blurry, the weight of the depth feature is reduced and the weight of the material boundary feature is increased; when the depth fails or a stable depth boundary cannot be obtained in multiple consecutive frames, the depth difference is not used as the main criterion, and the process of reshooting, adjusting the posture, prohibiting passage or manual processing is initiated.

[0055] Table 2 Implementation status of depth state boundary rules

[0056] When a partition is determined to be depth-blurred according to Table 2, the system reduces the weight of depth features in the stacking order determination and increases the weight of material boundary features such as near-infrared reflectance difference, texture gradient, or structured light edge features. When the partition is determined to be depth-failed according to Table 2, depth difference is no longer used as the primary criterion; instead, the system initiates a re-shoot, prompts wrist adjustment, prohibits passage, or requires manual processing. If the confidence level of the fused recognition is still lower than the admission requirement, no qualified conclusion is output.

[0057] The stacking order determination process triggered by deep reliability is as follows: Figure 5 As shown, the input is the target's circumferential partition and its multi-source detection data. If the depth information is reliable, the depth layer difference is used as the primary criterion for stacking order. If the depth information is blurry or invalid, the weight of the depth features is reduced, and near-infrared reflectance difference, texture gradient, or structured light edge features are used for fusion judgment. If the confidence level after fusion is still insufficient, the process enters the reshoot, adjustment, or interception stage, instead of outputting a qualified conclusion. This figure illustrates the conditional branch triggered by depth reliability.

[0058] In one implementation, the boundary of the depth state is based on the rule implementation in Table 2; Table 3 is only used to summarize the stacking order control strategies corresponding to different depth states, and the triggering conditions are not redefined.

[0059] Table 3. Determination Strategy for Stacking Order of Deep Reliability Triggering

[0060] Table 3 illustrates the processing strategies corresponding to the depth states determined in Table 2: when the depth is reliable, the depth layer difference is used as the primary criterion; when the depth is ambiguous, the depth weight is reduced and material boundary features are introduced; when the depth fails, the depth difference is not used as the primary criterion, but rather a reshoot, adjustment, or interception is initiated. This strategy is used to avoid directly using unreliable depth results as the basis for approval in scenarios involving flexible thin-layer material bonding, reflection, or occlusion.

[0061] S4, Anomaly Classification and Handling: For partitions that do not meet the effective coverage conditions, anomaly is attributed, and hard threshold anomalies and correctable anomalies are distinguished. If a hard threshold anomaly exists, a prohibition signal is directly output and manual handling or re-wearing is prompted. If only correctable anomalies exist, step S5 is executed. In practice, the system categorizes abnormal states that fail to meet effective coverage conditions into two main types: hard-threshold anomalies and correctable anomalies. All judgment thresholds are preset parameters that can be calibrated and adjusted based on the consumable specifications of gloves and surgical gown cuffs, equipment assembly precision, sterile barrier requirements, and clinical usage scenarios. Anomaly identification prioritizes situations that would affect access and release, including insufficient coverage, incorrect layering order, excessive gaps, slight curling, exposed surgical gown cuffs, exposed skin, and insufficient identification confidence. Among these, exposed skin, severe incorrect layering order, high-risk exposure of surgical gown cuffs, unmet cleanliness requirements, exceeding safety constraints, and persistently insufficient identification confidence or persistently excessive occlusion rate in key areas are considered hard-threshold anomalies. These do not undergo ordinary contact-based automatic correction but instead output a prohibition on release, manual handling, or re-wearing prompt.

[0062] Anomalies such as insufficient coverage, slight curling, and excessive gaps in the fit can be considered as locally restricted correction candidate anomalies and enter the subsequent action generation process, provided they do not trigger a hard threshold anomaly. Here, "local" means that the action applies to the target anomaly zone or a necessary adjacent buffer zone, without applying overall pulling to the entire cuff; "restricted" means that the amplitude of the candidate action, contact pressure, duration, number of repetitions, and wrist image deformation rate must all be limited by subsequent safety constraints.

[0063] Common anomaly types of cuff coverage boundary such as Figure 6 As shown.

[0064] Figure 6 The anomalies addressed in this invention are not arbitrary image defects, but rather abnormal states that affect the accessibility of the cuff coverage boundary. Insufficient coverage mainly affects the overlap length, surgical gown cuff exposure mainly affects the continuity of barrier coverage, incorrect layering order affects the inner and outer relationship between the glove and surgical gown cuff, excessive gap affects boundary adhesion, slight curling affects edge continuity, exposed skin and high-risk surgical gown cuff exposure are considered hard threshold interception situations, and insufficient recognition confidence triggers re-shooting, adjustment, or manual processing.

[0065] Insufficient coverage, exposed surgical gown cuffs, and exposed skin are classified as different criteria: Insufficient coverage is mainly judged based on the coverage length L. j Is it lower than L? min Judgment: Exposure of surgical gown cuffs is mainly determined by the exposed length L of the visible area of ​​the surgical gown cuff. exp,j and / or exposed area A exp,j Judgment; exposed skin is determined based on whether the skin area of ​​the wrist is detected, for example, the area of ​​exposed skin on the wrist, A. skin,j The anomaly is determined when the skin exposure exceeds a preset threshold or when skin on the wrist is detected in multiple consecutive frames. All three conditions can occur simultaneously; if all three occur, the anomaly with the higher risk will be treated as the one that causes the most problems.

[0066] The preset exposure limit includes the maximum exposure length L. exp,max and the upper limit of exposed area A exp,max If the equipment uses both exposed length and exposed area as criteria, then the requirement for slight exposure of the surgical gown cuffs is L. exp,j ≤L exp,max And A exp,j ≤A exp,max Furthermore, no skin exposure at the wrist was detected. If the device uses only one indicator, the indicator used should not exceed the corresponding preset upper limit. When the condition of slight exposure of the surgical gown cuff is met, it can be regarded as a candidate abnormality for local restricted correction and enter the subsequent abnormal area mapping and local restricted correction action generation process. Whether to generate a contact correction candidate action and whether to actually execute it are determined according to steps six and seven, respectively. If any of the indicators used exceeds the corresponding preset upper limit, the slight exposure correction process of the surgical gown cuff will not be entered. Among them, if the exposed area is located in a critical area or accompanied by skin exposure, it is judged as a high-risk exposure of the surgical gown cuff. Other cases exceeding the limit will not output a release conclusion and will enter the manual confirmation, re-inspection or re-wearing process.

[0067] In one implementation, the anomaly grading rules can be determined according to Table 4. This table is used to convert descriptions such as "minor" and "serious" into control boundaries that can be triggered by detection data, and the specific thresholds can still be obtained by device configuration and sample calibration.

[0068] When the curl width, curl angle, or curl area does not exceed the preset curl upper limit, it is judged as a slight curl and can enter the airflow deployment or reverse rolling candidate process; when the curl exceeds the preset curl upper limit, crosses multiple adjacent zones continuously, or is accompanied by incorrect stacking order, exposed skin, or high-risk exposure, it will not enter the normal curl correction process, but will output an interception or manual handling prompt.

[0069] When the fitting gap G j More than G max However, if the stacking order is correct and no skin exposure, high-risk exposure of surgical gown cuffs, or other hard threshold abnormalities are triggered, it can be considered a candidate for localized pressing abnormality and enter the subsequent action generation process. If it is accompanied by serious stacking order errors, skin exposure, high-risk exposure, or other hard threshold abnormalities, it will not enter the ordinary contact-based correction candidate process, but will instead enter the interception or manual processing process. If the confidence level of a single detection is insufficient, a re-shoot, supplementary lighting, or posture adjustment can be performed; if the number of consecutive preset attempts is still lower than the recognition confidence threshold C, further action can be taken. min Or, the occlusion rate of key partitions continues to exceed the upper limit of Occ. max If the result is not satisfactory, the system will not output a qualified conclusion and will instead proceed to manual processing or re-wearing.

[0070] Table 4 Examples of Anomaly Classification Rules

[0071] S5, Locally Restricted Correction Action Generation: Map the image partition corresponding to the correctable anomaly to the target physical execution partition, and generate a locally restricted correction action that only applies to the target physical execution partition and its adjacent buffer partitions. In step S5, the mapping error from the abnormal image partition to the physical execution partition is calculated. If the mapping error exceeds the preset maximum allowable error, no contact correction action is generated, and a repositioning or manual processing prompt is output.

[0072] The locally restricted correction action mentioned in step S5 includes at least one of axial pushing, reverse rolling, local pressing, tangential airflow deployment, or micro-adsorption traction. The amplitude, contact pressure, duration, number of repetitions, and wrist image deformation rate of the locally restricted correction action do not exceed the preset safety limit.

[0073] In a specific application, in one embodiment, the local correction execution unit includes a ring-shaped support arranged around the wrist detection area. Multiple local execution components are arranged circumferentially on the ring-shaped support. Each local execution component has a preset installation angle position or number and corresponds to one or more physical execution partitions E. k The controller determines the location based on the mounting angle and number of the local actuators, as well as the circumferential Z-axis of the image. j The mapping relationship determines the target execution component. Each local execution component can be driven independently or in groups by the controller, and performs local pressing, tangential unfolding, axial pushing, or micro-adsorption traction according to the anomaly type. The effective operating width of the execution component covers the corresponding physical execution partition E. k It can include a preset adjacent buffer; the buffer is used to reduce the missed correction caused by boundary error, but does not extend the action to the entire cuff circumferential area, so as to reduce the disturbance to the qualified partition.

[0074] After the image partition containing the anomaly is mapped to the corresponding physical execution partition through a calibration relationship, the system outputs a localized restricted correction action or interception signal based on the anomaly type. The action does not require all actuators to be present simultaneously; the specific action type is determined by the device configuration. During action generation, the anomaly type and severity are first determined, then the target execution partition, action direction, candidate action amplitude, and whether adjacent partitions need to be maintained are selected. If the mapping error exceeds the allowable range, or the anomaly type is a hard threshold anomaly, no contact-based correction candidate action is generated; instead, an interception, manual processing, re-detection, or re-wearing prompt is output. Whether a contact-based correction candidate action is actually executed is determined by the cleanliness status and safety constraint judgment results in step seven.

[0075] After determining the target anomalous partition and its corresponding physical execution partition, the system can represent the candidate correction action as the action vector U of the j-th partition.j : ; Among them, U j Let u be the correction action vector for the j-th partition. push,j u roll,j u press,j u air,j and u suction,j These represent the axial push amount, reverse rolling action amount, local pressing action amount, tangential airflow action amount, and micro-adsorption traction action amount, respectively. Each action component corresponds to a different physical dimension, therefore, cross-dimensional scalar summation is not performed on the unnormalized action components; the amplitude limiting processing of candidate action components is described below, and safety monitoring during actual execution is performed according to step seven.

[0076] The direction of motion is determined by the type of anomaly, while the amplitude of motion is determined by both the degree of anomaly and a preset upper limit. For example, when there is insufficient coverage, the direction of motion is along the axial direction of insufficient coverage, and the amplitude of motion is related to L. min -L j Correspondingly, for slight curling, the direction of movement is along the curling unfolding direction, and the amplitude of movement can correspond to the curling width, curling angle, curling area, or edge curvature change, and is constrained by airflow intensity, rolling pressure, or upper limit of rolling stroke; when the bonding gap is too large, the direction of movement is the local pressing direction, and the amplitude of movement is related to G. j -G max correspond.

[0077] To prevent candidate actions from exceeding the safety limit, each action component can be limited according to its corresponding safety limit: ; Where, k L and k G For preset or calibrated motion ratio coefficients, clip indicates that the amount of motion is limited to a safe upper limit. U j The motion components can be further converted into control signals such as valve opening, airflow pulse, roller rotation direction, pressing stroke, or adsorption duration for the corresponding actuator. Different equipment configurations can select one or more equivalent execution methods.

[0078] The correspondence between circumferential anomaly zones and local correction actions is as follows: Figure 7 As shown.

[0079] Figure 7In the diagram, the left side represents the abnormal zones divided circumferentially, while the right side shows optional local correction actions. Insufficient coverage can be addressed with axial pushing or micro-adsorption traction, slight edge curling can be addressed with reverse rolling or tangential airflow unfolding, and excessive bonding gaps can be addressed with local compression. Each action component in the diagram acts on the target abnormal zone or a necessary adjacent buffer zone and is constrained by the action amount, safety threshold, and re-inspection results.

[0080] In one implementation, exception types and local handling methods can be mapped as shown in Table 5. The actions in the table are exemplary engineering implementations; actual equipment can select one of these methods or an equivalent method based on its structural configuration.

[0081] Table 5 Mapping Table of Anomaly Types and Locally Restricted Corrective Actions

[0082] In Table 5, axial pushing or micro-adsorption traction corresponding to insufficient coverage is used to move the edge of the glove cuff towards the target coverage direction; axial pushing or local pressing corresponding to slight exposure of the surgical gown cuff is used to re-cover the exposed area with the glove cuff; high-risk exposure of the surgical gown cuff, serious layering sequence errors, and skin exposure do not enter the ordinary contact-based automatic correction process, but instead enter the prohibition release, manual handling, or re-wearing process; airflow unfolding or reverse rolling corresponding to slight curling is used to unfold the curled edge; local pressing corresponding to excessive bonding gap is used to reduce the gap between the two layers of flexible material; insufficient recognition confidence prioritizes re-shooting, supplementary lighting, posture adjustment, or manual handling. The above processing methods correspond one-to-one with the anomaly type classification, avoiding the use of all actuators as an indiscriminate toolbox.

[0083] S6, Cleanroom Safety Constraints: Under the condition of meeting cleanroom safety constraints, perform the local restricted correction action. After the cuff area stabilizes, repeat steps S2-S3. If all image partitions in the re-inspection meet the effective coverage conditions and there are no hard threshold anomalies, output the final release signal. If the coverage status does not improve after the number of corrections reaches the preset upper limit, stop the correction and output the prohibition release signal.

[0084] The requirement to meet clean safety constraints in step S6 specifically includes: Before performing contact-based correction, ensure that disposable contact parts are installed in place, the number of times the actuator has been used has not exceeded the preset number of times, the disinfection process is completed, and no contaminated areas are detected on the execution surface. When performing non-contact airflow correction, confirm that the airflow is sterile filtered airflow; During the correction process, the contact force, local pressure, amount of movement, duration, and wrist image deformation rate are monitored in real time. Correction is stopped immediately if any parameter exceeds the limit.

[0085] In step S6, after the cuff area is stabilized, specifically in the detection data of a consecutive preset number of frames, the axial position change of the glove cuff edge and the surgical gown cuff edge is lower than the preset stabilization threshold; if the number of corrections for the same abnormal partition reaches the preset maximum number of corrections, or if the abnormal state does not improve after two consecutive corrections, automatic correction is stopped and a prohibition signal is output.

[0086] In one specific implementation, after receiving the locally restricted correction candidate action generated in step six, the system first determines the cleanliness of the correction execution unit before executing contact correction. During execution, the contact force, local pressure, action amount, duration, number of corrections, and wrist image deformation rate are monitored simultaneously. If any safety constraint is not met, contact correction is stopped, and the system switches to non-contact processing, prohibits release, or provides a manual processing prompt. The aforementioned safety constraints are used to prevent local correction from damaging the glove material, disturbing the surgical gown cuff, or causing secondary impacts on already qualified zones.

[0087] Cleanliness status determination result S clean The determination of whether contact-based corrective actions are permissible can be based on a signal indicating the placement of a disposable sterile covering, the usage status of the actuator, the disinfection completion signal, the image detection results of the execution surface, or a combination of the above information. It should be noted that S... clean This mainly corresponds to the cleanliness permitting conditions for contact-type correction execution surfaces. If only non-contact airflow treatment is performed, the contact execution surfaces may not be required to meet contact cleanliness conditions depending on the equipment configuration. However, the final release still needs to meet the cuff coverage boundary access conditions and aseptic / clean process requirements. The aseptic / clean process requirements may include one or more of the following: disposable contact parts are installed in place, contact execution parts have not exceeded the preset number of uses, the disinfection process completion signal is valid, no contaminated areas are detected in the execution surface image, and the clean airflow or non-contact treatment process is completed.

[0088] In one implementation, the contact-type correction permission conditions are met when the disposable contact element is installed in place, the actuator has not exceeded the preset number of uses, the disinfection process completion signal is valid, and no contaminated area is detected in the execution surface image; if any condition is not met, contact-type automatic correction is not allowed. If a non-contact airflow deployment action is used, the non-contact processing branch can be entered, but it cannot be released directly just because the non-contact processing is completed. The final release signal can only be output after a re-inspection confirms that all cuff coverage boundary access conditions are met.

[0089] The process for determining cleanliness and safety constraints is as follows: Figure 8 As shown, the process first determines whether contact-based corrective action is required. If contact-based corrective action is not required, it can proceed to the non-contact or manual processing branch. If contact-based corrective action is required, the cleanliness status S is first determined. cleanTo determine if the requirements are met, further assessment is needed to determine if the contact force, pressure, amount of motion, duration, number of corrections, and wrist image deformation rate exceed the upper limits. Local correction actions are only permitted when both cleanliness and safety constraints are met simultaneously.

[0090] The stabilization waiting can be performed in two ways: one is to wait for a preset time; the other is to end the process after detecting a change in the cuff boundary position below a preset stabilization threshold in continuous images or depth data. After the stabilization waiting period ends, the cuff coverage boundary detection data is reacquired and a re-inspection is performed.

[0091] In one implementation, the stabilization wait is defined as the axial positional change of both the glove cuff edge and the surgical gown cuff edge being below a preset stabilization threshold δ across n consecutive frames of image or depth data. stable The process ends when the time is up; if the waiting time reaches the maximum waiting time and still does not stabilize, the process will proceed to re-inspection failure or manual processing. ; If the number of correction attempts for the same abnormal partition reaches the preset upper limit M max If the coverage length, bonding gap, or curling condition does not improve after two consecutive corrections, automatic correction will stop and a prompt to prohibit release or require manual handling will be output to avoid the closed-loop correction from entering an invalid cycle.

[0092] The release criteria are based on the coverage boundary access results, rather than the completion of the initial donning mechanism's actions as the sole basis. Rejection criteria include situations such as exposed skin, severe layering sequence errors, insufficient cleanliness, persistently low identification confidence, exceeding safety constraints, and abnormal deterioration after re-inspection.

[0093] The complete control process from detection, correction, stabilization, re-inspection to final release or interception is as follows: Figure 9 As shown. Figure 9 This demonstrates the complete process from initial donning to final release or interception under one of the above embodiments. The system sequentially performs mechanism retreat, multi-source data acquisition, coordinate establishment, boundary identification, effective coverage determination, anomaly location, correction permission determination, action execution, stabilization waiting, and re-inspection. If the re-inspection is successful, the system outputs final release; if the re-inspection fails, a hard threshold anomaly occurs, or safety constraints are exceeded, the system outputs a prohibition on release or a manual handling prompt.

[0094] To better illustrate the technical solution of this invention, the following terms are explained: (1) Automatic glove-wearing machine: A device that can assist users in the initial wearing of medical gloves. This invention does not limit its supply, opening, negative pressure, airflow assistance or mechanical wearing structure.

[0095] (2) Initial donning completed: The gloves have been put on the hands with the help of the equipment, but have not yet undergone the access detection and re-inspection of the cuff coverage boundary.

[0096] (3) Cuff coverage boundary: The overlapping, fitting, layering and exposure relationship between the glove cuff and the surgical gown cuff in the circumferential area of ​​the wrist.

[0097] (4) Effective coverage: The glove cuff is located outside the cuff of the surgical gown, and the coverage length, layering order, fitting gap, identification confidence and occlusion rate meet the admission criteria.

[0098] (5) Multi-source detection data: at least includes position data for determining the edge of the glove cuff and the edge of the surgical gown cuff, and preferably includes depth information; when the depth information is unreliable, near-infrared reflectance difference, texture gradient, structured light edge features or other material feature information that can distinguish the boundary between the glove material and the surgical gown cuff material can be further called.

[0099] (6) Critical zones: These are circumferential zones determined based on sterile barrier requirements, wrist flexion positions, main contact areas of the cuff, historically high-risk areas of abnormalities, or currently detected exposures, skin exposures, or serious layering sequence errors. Critical zones can be preset zones or include abnormal zones and their adjacent buffer zones. When skin exposure, high-risk exposure of surgical gown cuffs, or persistently insufficient identification confidence occur within a critical zone, it is not treated as a normal correctable abnormality.

[0100] (7) Hard threshold anomaly: refers to an abnormal state that does not allow direct passage and usually does not enter the normal contact automatic correction process, including skin exposure, serious layering sequence error, high risk of surgical gown cuff exposure, cleanliness not met, safety constraints exceeding limits, insufficient confidence in continuous identification of key areas or continuous exceeding of key area occlusion rate, etc.

[0101] (8) Local restricted correction: refers to the correction method that applies axial pushing, local pressing, reverse rolling, airflow unfolding or micro-adsorption traction to the target abnormal zone and necessary adjacent buffer zones, and keeps the amplitude of the action, contact pressure, duration, number of repetitions and wrist image deformation rate within the preset safety limit.

[0102] (9) Adjacent buffer zones: These refer to one or more circumferential zones adjacent to the target anomalous zone, used to compensate for mapping errors or the continuity of flexible material edges. Buffer zones are only used to assist in correcting the target anomalous zone and do not extend the correction action to the entire cuff circumferential area.

[0103] (10) Safety limit: refers to the control boundary set to avoid glove damage, surgical gown cuff disturbance, qualified area being pulled or excessive pressure on the wrist area, including one or more of the following: upper limit of motion displacement, upper limit of contact pressure, upper limit of duration, upper limit of airflow intensity, upper limit of rolling pressure, upper limit of rolling stroke, upper limit of correction times and upper limit of wrist image deformation rate.

[0104] (11) Aseptic / clean process requirements: refers to one or more of the following conditions that the equipment must meet before release: disposable contact parts are installed in place, contact actuators have not exceeded the preset number of uses, the disinfection process completion signal is valid, no contaminated areas are detected in the execution surface image, and the clean airflow or non-contact processing process is completed.

[0105] (12) Preset thresholds and parameters: The preset exposure limit, preset edge curling limit, maximum bonding gap, minimum recognition confidence, maximum occlusion rate, stability threshold and maximum number of corrections mentioned in the text can be determined by consumable specifications, equipment structure, sample labeling, manual qualification judgment results, sterile barrier requirements and equipment calibration results, and do not represent fixed experimental data.

[0106] (13) Manual handling and re-wearing: Manual handling refers to the operator visually confirming the cuff coverage status, manually adjusting, replacing consumables, or deciding to re-wear gloves; Re-wearing refers to the system terminating the current automatic correction process, allowing the user to re-enter the glove wearing process or completing the re-wearing process with manual assistance.

[0107] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for correcting and controlling the cuff coverage boundary after automatic glove wearing, characterized in that, Includes the following steps: S1, Access Detection Trigger: Receives the initial wearing completion signal from the automatic glove-wearing machine and initiates cuff coverage boundary detection; S2, Multi-source detection and partition mapping: Acquire multi-source detection data of the glove cuff edge and the surgical gown cuff edge, establish a wrist axis-circumferential coordinate system, divide the cuff area into multiple circumferential image partitions, and establish the mapping relationship between the image partitions and the physical execution partitions; S3, Multi-dimensional effective coverage determination: Extract the coverage length, bonding gap, recognition confidence and occlusion rate of each image partition from the multi-source detection data, and dynamically determine the stacking order of each image partition based on depth reliability. Determine whether each image partition meets the preset effective coverage conditions based on the coverage length, stacking order, bonding gap, recognition confidence and occlusion rate. If all image partitions meet the conditions, output a pass signal; otherwise, proceed to step S4. S4, Anomaly Classification and Handling: For partitions that do not meet the effective coverage conditions, anomaly is attributed, and hard threshold anomalies and correctable anomalies are distinguished. If a hard threshold anomaly exists, a prohibition signal is directly output and manual handling or re-wearing is prompted. If only correctable anomalies exist, step S5 is executed. S5, Locally Restricted Correction Action Generation: Map the image partition corresponding to the correctable anomaly to the target physical execution partition, and generate a locally restricted correction action that only applies to the target physical execution partition and its adjacent buffer partitions. S6, Correction and re-inspection release under clean safety constraints: Perform the local restricted correction action under the condition of meeting clean safety constraints, and repeat steps S2-S3 after the cuff area is stable. If all image partitions are re-inspected and meet the effective coverage conditions without any hard threshold anomalies, the final clearance signal is output; if the coverage status does not improve after the number of corrections reaches the preset upper limit, correction is stopped and a clearance prohibition signal is output.

2. The method for correcting and controlling the cuff coverage boundary after automatic glove wearing according to claim 1, characterized in that, In step S3, the dynamic determination of the stacking order of each image partition based on depth reliability includes: Obtain at least one parameter from the following for each image partition: depth level difference, depth hole rate, edge diffusion width, and depth noise; and determine the reliability of the depth information based on the parameters. When the depth information is reliable, the depth level difference is used as the main criterion for the stacking order. When depth information is unreliable, the weight of depth features is reduced and the weight of material features that can distinguish the boundary between glove material and surgical gown cuff material is increased for determination. The material features that can distinguish the boundary between glove material and surgical gown cuff material are extracted from the multi-source detection data.

3. The method for correcting and controlling the cuff coverage boundary after automatic glove wearing according to claim 2, characterized in that, The reliability of the depth information determination includes: When the depth level difference reaches the preset reliability threshold, and the depth void ratio, edge diffusion width, and depth noise do not exceed the corresponding upper limit, it is determined to be deep reliable. When the depth level difference is lower than the preset fuzzy lower limit, or the depth void ratio or depth noise exceeds the preset failure threshold, it is determined to be a depth failure. All other cases are classified as depth blur. When the depth is determined to be ambiguous, the weight of the depth feature is reduced and the weight of the material feature is increased for fusion determination; when the depth is determined to be ineffective, the depth difference is not used as the main determination basis, and step S2 is re-executed or manual processing is performed.

4. The method for correcting and controlling the cuff coverage boundary after automatic glove wearing according to claim 1, characterized in that, In step S2, establishing the mapping relationship between image partitions and physical execution partitions specifically involves: determining the axial and circumferential directions based on at least one of the wrist crease center, wrist contour center line, glove cuff reference edge, or surgical gown cuff reference edge; pre-calibrating the transformation relationship between the detection coordinate system, unfolding coordinate system, and execution coordinate system; and establishing a mapping function between the image circumferential partitions and physical execution partitions. In step S5, the mapping error from the abnormal image partition to the physical execution partition is calculated. If the mapping error exceeds the preset maximum allowable error, no contact correction action is generated, and a repositioning or manual processing prompt is output.

5. The method for correcting and controlling the cuff coverage boundary after automatic glove wearing according to claim 4, characterized in that, The formula for calculating the coverage length is: ； Among them, L j Let n be the coverage length of the j-th circumferential image partition. j Let θ be the number of circumferential sampling points involved in the calculation within the j-th circumferential image partition. i Let x be the position of the i-th sampling angle. G (θ i ) represents the axial position of the glove cuff edge at the sampling angle, x S (θ i () indicates the axial position of the edge of the surgical gown cuff at the same sampling angle.

6. The method for correcting and controlling the cuff coverage boundary after automatic glove wearing according to claim 1, characterized in that, The preset effective coverage condition in step S3 is that a certain image partition simultaneously satisfies the following: (1) Coverage length ≥ preset minimum allowable coverage length; (2) The layering order is such that the glove cuff is located outside the surgical gown cuff; (3) The bonding gap is less than or equal to the preset maximum allowable bonding gap; (4) The recognition confidence level is greater than or equal to the preset minimum recognition confidence level; (5) Occlusion rate ≤ preset maximum allowable occlusion rate.

7. The method for correcting and controlling the cuff coverage boundary after automatic glove wearing according to claim 1, characterized in that, The requirement to meet clean safety constraints in step S6 specifically includes: Before performing contact-based correction, ensure that disposable contact parts are installed in place, the number of times the actuator has been used has not exceeded the preset number of times, the disinfection process is completed, and no contaminated areas are detected on the execution surface. When performing non-contact airflow correction, confirm that the airflow is sterile filtered airflow; During the correction process, the contact force, local pressure, amount of movement, duration, and wrist image deformation rate are monitored in real time. Correction is stopped immediately if any of the above parameters exceed the limit.

8. The method for correcting and controlling the cuff coverage boundary after automatic glove wearing according to claim 1, characterized in that, In step S6, after the cuff area is stabilized, specifically in the detection data of a consecutive preset number of frames, the axial position change of the glove cuff edge and the surgical gown cuff edge is lower than the preset stabilization threshold; if the number of corrections for the same abnormal partition reaches the preset maximum number of corrections, or if the abnormal state does not improve after two consecutive corrections, automatic correction is stopped and a prohibition signal is output.

9. A glove automatic wearing and cuff coverage boundary zone correction control system, characterized in that, include: A multi-source detection unit is used to acquire multi-source detection data of the glove cuff edge and the surgical gown cuff edge; A local controller, which is communicatively connected to the multi-source detection unit, is used to execute the glove automatic wearing post-cuff coverage boundary partition correction control method as described in any one of claims 1-8, and to generate correction control signals and release / interception signals; The cuff correction execution unit is communicatively connected to the local controller and is used to perform local restricted correction actions according to the correction control signal; The cleanliness and safety detection unit is communicatively connected to the local controller and is used to detect the cleanliness status of the correction execution unit and the safety parameters during the correction process. The human-machine interaction unit is communicatively connected to the local controller and is used to output release signals, prohibition release signals, manual processing prompts, or re-wearing prompts.

10. The automatic glove-wearing cuff coverage boundary zone correction control system according to claim 9, characterized in that, The local correction execution unit includes a ring support, and multiple independently controllable local execution components are arranged along the circumference of the wrist on the ring support. Each local execution component corresponds to one or more physical execution partitions. The local actuation component is at least one of an axial pushing mechanism, a reverse rolling mechanism, a local pressing mechanism, a sterile airflow nozzle, or a micro-adsorption mechanism.

Images