Integrated material checkout stand and use method thereof

Abstract

The invention provides an integrated material checkout counter and a use method thereof, one side of a weighing platform is provided with a vertical support column, the vertical support column is provided with a control module, the control module is provided with a man-machine interaction display screen, a binocular depth visual identification camera is arranged above the man-machine interaction display screen, and the binocular depth visual identification camera is arranged above the man-machine interaction display screen. The field angle of the binocular depth visual identification camera covers the area where the weighing platform is located; an RFID read-write device is arranged on one side of the vertical supporting stand column. The residual error of the total weight and the RFID theoretical weight is utilized, visual secondary feature extraction and integer programming matching are combined, automatic additional recording of low-value and easy-to-consume products such as screws and gaskets is achieved, and the blank of management of the traditional RFID checkout counter for small materials without labels is filled.

Description

Technical Field

[0001] This invention relates to the field of intelligent warehousing, and in particular to an integrated material checkout counter and its usage method. Background Technology

[0002] With the development of modern materials management and intelligent warehousing technology, traditional materials entry and exit management is gradually shifting from manual recording to intelligent and automated processes. Among existing technical solutions, RFID (Radio Frequency Identification) technology has been widely applied in warehouse material management systems. CN111915234A discloses an RFID-based warehouse material management system that uses RFID readers to read goods tag information and perform quantity statistics, achieving paperless management.

[0003] To further improve identification accuracy, existing technologies attempt to combine RFID with weighing technology. CN217637627U discloses an intelligent scale with RFID tag identification function, which integrates a gravity sensor, RFID read / write antenna and identity recognition module, and verifies by comparing tag data with weight data. Similarly, in mobile operation scenarios, CN120117305A discloses an RFID intelligent operating device that can weigh and collect data, which uses an RFID data collection workbench on a trolley and a high-precision weighing instrument to achieve material information binding and verification on the mobile terminal.

[0004] In addition, to solve the problem of RFID signal collision, existing self-checkout systems, such as the self-checkout system disclosed in CN101625785A, use channel sorting and anti-collision algorithms to optimize multi-tag reading.

[0005] However, the aforementioned existing technologies still have significant limitations in practical applications. First, the simple combination of "RFID + weighing" usually only performs a simple linear comparison, that is, "theoretical weight" to "actual weight", lacking the perception of the spatial volume and visual characteristics of the materials. When encountering high-density stacking of materials, metal obstruction leading to RFID missed readings, or the presence of "untagged foreign objects", the system has difficulty distinguishing whether it is due to damaged tags or abnormal weight.

[0006] Secondly, most existing devices simply overlay sensors as independent data sources, lacking in-depth multimodal data fusion algorithms. This prevents them from dynamically adjusting the trust weights of each sensor based on environmental changes (such as visual occlusion or tag density), leading to a high false alarm rate in complex lighting or electromagnetic environments. Finally, for tiny consumables that cannot be labeled, current technologies often require manual data entry, lacking automated processing methods based on weight residuals and visual inverse reasoning. Furthermore, in case of anomalies, there is a lack of intuitive augmented reality (AR) guidance to help operators quickly correct errors. Summary of the Invention

[0007] The main objective of this invention is to provide an integrated material checkout counter and its usage method, which solves the technical problems of existing material checkout equipment in complex scenarios such as densely stacked materials, shielded RFID signals, or mixed unlabeled materials. These problems arise because the single or simple dual-modal recognition mechanism cannot dynamically compensate for sensor errors, resulting in low recognition accuracy, difficulty in achieving closed-loop verification of the consistency of "visual radio frequency gravity" three-dimensional features, and inability to automatically process small unlabeled materials.

[0008] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: an integrated material settlement counter, wherein a vertical support column is provided on one side of the weighing platform, a control module is provided on the vertical support column, a human-machine interaction display screen is provided on the control module, and a binocular depth vision recognition camera is provided above the human-machine interaction display screen, the field of view of the binocular depth vision recognition camera covers the area where the weighing platform is located; An RFID reader / writer is installed on one side of the vertical support column.

[0009] In the preferred embodiment, the control module integrates a core data processing unit, which is communicatively connected to the human-machine interaction display screen, the weighing platform, the RFID reader / writer, and the binocular depth vision recognition camera. The core data processing unit is configured to execute a material consistency verification algorithm based on visual radio frequency gravity three-dimensional feature fusion.

[0010] In the preferred scheme, the weighing platform collects simulated gravity signals; RFID reader / writer collects radio frequency tag data; Image stream data acquired by the binocular depth vision recognition camera are all aggregated into the control module for spatiotemporal alignment and fusion processing.

[0011] A method for using an integrated materials checkout counter, the method being executed by the core data processing unit in the control module, including: S1. Initialize the system and activate the binocular depth vision recognition camera to monitor in real time whether an object is placed on the weighing platform. S2. Once the placement of the materials is detected to be stable, multi-dimensional data acquisition is triggered simultaneously: the total physical weight is obtained through the weighing platform. Read the set of electronic tags through an RFID reader / writer. The three-dimensional point cloud and texture data of the materials are acquired through a binocular depth vision recognition camera. ; S3. Perform visual volumetric physical density coupling estimation, based on Calculate the visual estimated weight of materials ; S4. Perform RF visual weight ternary consistency verification and calculate the credibility index of this checkout process. ; S5, if Greater than the preset safety threshold If so, a settlement list will be generated on the human-computer interaction display screen; if Less than or equal to If this occurs, an abnormal alarm will be triggered.

[0012] In the preferred embodiment, step S3 involves calculating the visually estimated weight. The steps are as follows: Using point cloud segmentation algorithms from Extract the target area of ​​materials on the shelf and calculate the bounding box volume of the materials. ; Deep neural networks are used to identify material types and match corresponding density coefficients from a database. ; The calculation formula is: ; in, This represents the number of objects visually identified. This is a stacking correction factor used to compensate for volume estimation errors caused by visual occlusion.

[0013] In the preferred scheme, the confidence index is calculated in step S4. The mathematical model is as follows: First, analyze the electronic tag set. Obtain the theoretical total weight based on tag information ; Define RF deviation rate ; Define visual deviation rate ; Construct a dynamic weight verification function: ; in, The sensitivity coefficient of the Sigmoid function. This represents the system's basic error tolerance. and These are dynamic weighting coefficients, and they satisfy... , ,in The system automatically reduces the weight of radio frequency data as the number of RFID tags read increases. To prevent errors caused by interference from multiple labels.

[0014] In the preferred embodiment, considering the structural characteristics of the RFID reader / writer device being side-mounted, step S2 further includes a lateral radio frequency enhancement scanning strategy: When the binocular depth vision recognition camera detects that the height of the stacked materials exceeds a preset threshold At the same time, the control module controls the RFID reader to increase the transmission power and enables a multi-timeslot polling mechanism to ensure that it can penetrate the stacked materials and read the RFID tags on the bottom layer.

[0015] In the preferred embodiment, the method also includes a residual inference step for unlabeled micromaterials: When the system detects Furthermore, when the difference exceeds the set value, the control module initiates residual analysis: Calculate the remaining weight not interpreted by the RFID tag ; exist Secondary feature extraction is performed on regions in the image that do not match labels; Will Perform integer programming matching with the unit weight of low-value consumables in the database. If an integer solution exists: ; in It is an integer. If the unit weight is a low-value material of a certain type, then that type of material will be automatically added to the settlement list.

[0016] In the preferred embodiment, the method also includes an anomaly correction step based on human-computer interaction: When step S5 triggers an abnormal alarm, the control module projects the real-time image captured by the binocular depth vision recognition camera onto the human-computer interaction display screen, and uses augmented reality technology to highlight the suspected abnormal material areas in the image, that is, areas where the visual volume and weight do not match, or areas where the label is not read, guiding the user to adjust the placement of materials or remove abnormal items.

[0017] In the preferred embodiment, step S1 also incorporates biometric authorization verification: By using a binocular depth vision recognition camera to collect facial depth information of operators, an anti-spoofing facial feature vector is constructed. ; Will Compare with the pre-stored permission database, only if the similarity is... Only when the time is right will the control module's control over the RFID reader / writer be unlocked to prevent unauthorized personnel from operating the device and causing leakage of material data.

[0018] This invention provides an integrated material checkout counter and its usage method. This application adopts a material consistency verification algorithm based on the fusion of visual, radio frequency, and gravity three-dimensional features. By constructing a dynamic weight verification function containing a Sigmoid function, it can automatically adjust the trust weight of each sensor data according to the visual occlusion factor and RFID tag density. Thus, it relies on radio frequency and gravity when vision is obstructed, and on vision and gravity when radio frequency is disturbed, which significantly improves the robustness and confidence of the system in complex stacking scenarios.

[0019] This application introduces a residual inference mechanism for untagged small items. Utilizing the residual between the total weight and the theoretical weight of the RFID tag, combined with visual secondary feature extraction and integer programming matching, it achieves automated replenishment of low-value consumables such as screws and washers, filling the management gap of untagged small items in traditional RFID checkout counters. This application also employs a vision-guided lateral RF enhancement scanning strategy, using binocular vision to locate the core stacked area and guide the RFID reader / writer in beamforming and power adjustment, effectively solving the problem of insufficient penetration of side-mounted antennas when facing high-density stacked items.

[0020] In addition, this application also incorporates augmented reality (AR) anomaly correction function. When low consistency confidence is detected, the area where the visual volume and weight do not match is directly highlighted on the screen, realizing an interactive upgrade from "error reporting" to "intelligent error correction", which greatly reduces the difficulty of troubleshooting for operators. Attached Figure Description

[0021] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a structural diagram of the integrated material settlement counter of the present invention; Figure 2 This is a flowchart of the integrated material checkout algorithm of the present invention; Figure 3 This is a diagram of the human-computer interaction interface of the integrated material checkout counter of the present invention.

[0022] In the diagram: 1. Control module; 2. Human-machine interaction display screen; 3. Weighing platform; 4. RFID reader / writer; 5. Binocular depth vision recognition camera. Detailed Implementation

[0023] Example 1 like Figure 1-3 As shown, an integrated material checkout counter has a vertical support column on one side of the weighing platform 3, a control module 1 on the vertical support column, a human-machine interaction display screen 2 on the control module 1, and a binocular depth vision recognition camera 5 above the human-machine interaction display screen 2. The field of view of the binocular depth vision recognition camera 5 covers the area where the weighing platform 3 is located. An RFID reader / writer 4 is installed on one side of the vertical support column.

[0024] In the preferred embodiment, the control module 1 integrates a core data processing unit, which is communicatively connected to the human-machine interaction display screen 2, the weighing platform 3, the RFID reader / writer 4, and the binocular depth vision recognition camera 5. The core data processing unit is configured to execute a material consistency verification algorithm based on visual radio frequency gravity three-dimensional feature fusion.

[0025] In the preferred embodiment, the weighing platform 3 collects the simulated gravity signal; RFID reader / writer 4 collects radio frequency tag data; The image stream data collected by the binocular depth vision recognition camera 5 are all aggregated to the control module 1 for spatiotemporal alignment and fusion processing.

[0026] This embodiment provides an integrated material checkout counter, whose hardware structure mainly includes a weighing platform 3, a vertical support column, a human-machine interface display screen 2, a control module 1, an RFID reader / writer 4, and a binocular depth vision recognition camera 5. The weighing platform 3 serves as a base to support materials and acquire weight data. The vertical support column is located on one side of the weighing platform 3 for support. The control module 1 is mounted on the vertical support column. The human-machine interface display screen 2 is installed on the control module 1 to display information to the user. The binocular depth vision recognition camera 5 is located above the human-machine interface display screen 2, with its field of view covering the entire area of ​​the weighing platform 3 downwards to capture visual images of the materials. An RFID reader / writer 4 is also located on one side of the vertical support column for reading electronic tag information from the materials. The core data processing unit integrated within the control module 1 is connected to the human-machine interface display screen 2, the weighing platform 3, the RFID reader / writer 4, and the binocular depth vision recognition camera 5 via communication interfaces to achieve centralized data aggregation and processing.

[0027] The core data processing unit is configured to execute a material consistency verification algorithm based on the fusion of visual, radio frequency, and gravity three-dimensional features. This algorithm first performs spatiotemporal alignment and fusion processing on multi-source data. Specifically, the simulated gravity signal collected by the weighing platform 3, the radio frequency tag data collected by the RFID reader / writer 4, and the image stream data collected by the binocular depth vision recognition camera 5 are synchronized and aligned according to the data acquisition timestamps of each sensor after entering the control module 1, ensuring that data frames at the same time correspond. Subsequently, the system uses the binocular vision data to calculate the visually estimated weight of the materials. The calculation formula is as follows: ; In the formula, Represents the theoretical total weight estimated based on visual volume. This represents the number of objects visually identified. Representing the The volume of the outer bounding box of an object. Represents the first match found in the database Density coefficient of the type of material, This represents the stacking correction factor.

[0028] The meaning of this formula is that by visually identifying the volume and type density of materials, its theoretical weight can be calculated, and a stacking correction factor is introduced to compensate for the volume estimation error caused by visual obstruction, thereby obtaining a third-dimensional verification data independent of the weighing sensor and RFID data.

[0029] Based on this, the core data processing unit further calculates the reliability index of material consistency. This is used to determine the accuracy of the current checkout list, and its calculation model is constructed as a dynamic weighted verification function: ; In the formula, The confidence index represents the final consistency; the closer the value is to 1, the higher the confidence level. The sensitivity coefficient represents the sigmoid function and is used to control the steepness of the function curve; This represents the basic error tolerance allowed by the system; This represents the radio frequency deviation rate, which is the ratio of the actual weight to the theoretical weight corresponding to the RFID tag. This represents the visual bias rate, which is the ratio of the actual weight to the visually estimated weight. and These represent the dynamic weighting coefficients for radio frequency (RF) data and visual data, respectively. The physical meaning of this formula is that the system weights and fuses RF and visual biases, mapping them to a confidence probability between 0 and 1 using a Sigmoid nonlinear function. When the number of RFID tags is large or there is a risk of signal interference, the system will automatically reduce... The value and increase The value of makes the verification results more dependent on the matching degree of vision and gravity, and vice versa, thereby achieving adaptive adjustment to different environments.

[0030] By combining the aforementioned hardware layout and algorithms, the problem of low recognition accuracy in complex scenarios such as stacked materials, obscured labels, or unlabeled foreign objects mixed in at traditional checkout counters is effectively solved. On the hardware side, the high-angle, top-down layout of the binocular cameras and the side-mounted RFID layout create a complementary perception space, reducing blind spots. On the software algorithm side, by introducing visual volumetric weight estimation as a third-dimensional data point and using a dynamic weight verification function to non-linearly fuse physical weight, RFID data, and visual data, it can rely on visual and gravity verification when RFID signals are blocked, or on RFID and gravity verification when visual signals are obstructed, achieving complementary enhancement of multimodal features. This design not only ensures strict consistency in material entry and exit records but also provides an automated processing path based on visual and weight residuals for small materials that cannot be labeled, significantly improving the intelligence level and operational efficiency of material management, and fully meeting the requirements of patent law for sufficient disclosure of technical solutions and support of claims in the specification.

[0031] Example 2 Further explanation in conjunction with Example 1, such as Figure 1-3 As shown, a method for using an integrated materials checkout counter is implemented using the core data processing unit in control module 1, including: S1. Initialize the system and activate the binocular depth vision recognition camera 5 to monitor in real time whether an object is placed on the weighing platform 3. S2. Once the placement of the materials is detected to be stable, multi-dimensional data acquisition is triggered simultaneously: the total physical weight is obtained through the weighing platform 3. The electronic tag set is read by the RFID reader / writer 4. The three-dimensional point cloud and texture data of the materials were acquired through a binocular depth vision recognition camera 5. ; S3. Perform visual volumetric physical density coupling estimation, based on Calculate the visual estimated weight of materials ; S4. Perform RF visual weight ternary consistency verification and calculate the credibility index of this checkout process. ; S5, if Greater than the preset safety threshold If so, a settlement list will be generated on the human-computer interaction display screen 2; if Less than or equal to If this occurs, an abnormal alarm will be triggered.

[0032] In system initialization step S1, a binocular depth vision recognition camera is activated as the primary trigger sensing device. It can continuously monitor the weighing platform within its field of view in real time to determine whether any items awaiting payment have been placed on it. Compared to traditional infrared beam or gravity triggering, this vision-based triggering mechanism provides richer environmental context information, effectively preventing accidental system activation due to accidental touches or the placement of non-payment items.

[0033] Once the system detects in step S2 that the material has been stably placed on the weighing platform, the core data processing unit will issue a synchronization command to trigger concurrent acquisition of multi-dimensional data. At this time, the weighing platform collects the actual total physical weight of the material, denoted as... Simultaneously, the RFID reader / writer performs a full scan of all electronic tags within the radio frequency coverage area, obtaining a set containing the identity information of all identified tags, denoted as . The binocular depth vision recognition camera simultaneously captures high-precision 3D point cloud data and surface texture image data of the materials, denoted as... The key to this step lies in data synchronization, which ensures strict alignment of physical weight, radio frequency signals, and visual images in the time dimension, providing a reliable data foundation for subsequent fusion calculations.

[0034] After data acquisition, steps S3 and S4 mainly involve the processing of the core algorithm. The system first performs coupled estimation of visual volume and physical density, calculating the outer contour volume of the material using the acquired 3D point cloud data, and combining this with density parameters in the database based on texture features to derive the visually estimated weight of the material, denoted as . Subsequently, the system enters the crucial three-dimensional consistency verification stage, which involves RFID, visual, and weight measurements. In this stage, the system incorporates the actual physical weight, the theoretical weight derived from RFID tag parsing, and the visually estimated weight into a unified mathematical model for comparison and analysis. This model calculates a reliability index reflecting the consistency of the data during the checkout process, denoted as [index missing]. This index is a quantitative indicator used to characterize the probability of a match between current physical resources and the digital information read by the system.

[0035] Finally, in step S5, the system makes a logical judgment based on the calculated credibility index. The system has a preset security threshold, denoted as... This threshold represents the minimum acceptable level of credibility allowed by the system. If the calculated credibility index exceeds this threshold... This indicates a high degree of consistency in the multi-dimensional data, leading the system to determine that the checkout information is accurate. The system then generates a final settlement statement on the user-interactive display screen for confirmation. Conversely, if the credibility index is less than or equal to this threshold... If this occurs, it indicates a data conflict in the system. For example, there may be unread labels, labels that do not match the actual items, or abnormal weights. The system will then trigger an alarm, block the checkout process, and prompt manual intervention.

[0036] This method introduces visual weight estimation as a third dimension of verification data, breaking the limitations of traditional checkout systems that rely solely on two-dimensional data comparison between RFID and weighing. Through deep fusion calculations in steps S3 and S4, the system can sensitively detect inconsistencies in data when RFID signals are missed due to metal obstruction or liquid interference, or when unlabeled foreign objects cause weight deviations. This reduces the reliability index and triggers an alarm. This multimodal data verification logic not only significantly improves the accuracy and robustness of material identification but also effectively prevents material loss and accounting errors.

[0037] In the preferred embodiment, step S3 involves calculating the visually estimated weight. The steps are as follows: Using point cloud segmentation algorithms from Extract the target area of ​​materials on the shelf and calculate the bounding box volume of the materials. ; Deep neural networks are used to identify material types and match corresponding density coefficients from a database. The calculation formula is: ; in, This represents the number of objects visually identified. This is a stacking correction factor used to compensate for volume estimation errors caused by visual occlusion.

[0038] In this embodiment, the step of calculating the visual estimation weight is the core step in the system's 3D feature verification. First, the core data processing unit processes the 3D point cloud and texture data acquired by the binocular cameras. Preprocessing is performed using point cloud segmentation algorithms to distinguish the background environment from the material targets on the platform, accurately extracting the three-dimensional spatial coordinates of the materials. Based on this, the system further calculates the bounding box volume of each individual material target, thus forming a set of volume data. .

[0039] Simultaneously, the system utilizes a pre-trained deep neural network model to extract features and classify the texture images of materials, determining the specific type of materials, and indexing the corresponding standard density coefficients of that type of material from local or cloud databases based on the recognition results. This process maps geometry to physical properties, providing the necessary physical parameters for subsequent weight estimation.

[0040] After obtaining the volume and density, the system calculates the visually estimated weight of the material based on a preset mathematical model. Its calculation formula is expressed as In this formula, Represents the total number of objects currently recognized by the visual system, symbol This represents a summation operation over all identified objects. Each term in the formula represents the estimated weight of a single object, where... For the first The volume of an object For the first The density of an object. It is particularly important to explain the parameters. This refers to the stacking correction factor. During visual recognition, the stacking of materials often obscures part of their volume, causing the directly calculated volume to be smaller than the actual volume.

[0041] Therefore, the introduction of this stacking correction factor aims to nonlinearly compensate for the volume estimation error caused by visual occlusion, ensuring that the final calculated visually estimated weight is closer to the true physical weight of the material.

[0042] By introducing a visual weight estimation formula that includes a stacking correction factor, the system can independently generate third-dimensional weight reference data without relying on electronic tags and weighing sensors. This algorithm logic, based on volume multiplied by density and then corrected, not only clarifies the specific calculation path for converting visual data into weight data, avoiding the problem of insufficient disclosure caused by overly abstract or generalized technical solutions, but also effectively solves the problem of underestimation due to occlusion in densely stacked scenarios when using single visual recognition.

[0043] In the preferred scheme, the confidence index is calculated in step S4. The mathematical model is as follows: First, analyze the electronic tag set. Obtain the theoretical total weight based on tag information ; Define RF deviation rate ; Define visual deviation rate ; Construct a dynamic weight verification function: ; in, The sensitivity coefficient of the Sigmoid function. This represents the system's basic error tolerance. and These are dynamic weighting coefficients, and they satisfy... , ,in The system automatically reduces the weight of radio frequency data as the number of RFID tags read increases. To prevent errors caused by interference from multiple labels.

[0044] In this embodiment, the step of calculating the confidence index aims to transform the measurement differences of multimodal sensors into a quantified confidence index through a mathematical model. First, the system will process the collected electronic tags... The process involves parsing and retrieving the standard weight of the materials associated with the unique identifier of each electronic tag from the database. These standard weights are then summed to obtain the theoretical total weight based on the tag information. Based on this, the system defines two key deviation indicators, including the radio frequency deviation rate. Defined as the total physical weight obtained from actual weighing. Total weight of radio frequency theory The absolute value of the difference between them divided by the total physical weight This is used to characterize the degree of deviation between radio frequency data and physically measured data; similarly, visual deviation rate... Defined as total physical weight Weight estimation with visual estimation The absolute value of the difference between them divided by the total physical weight These two deviation rate indices characterize the degree of deviation between visually estimated data and physically measured data. They normalize the errors of data from different dimensions, allowing them to be compared and weighted within the same mathematical model.

[0045] To integrate the aforementioned biases and output the final judgment result, the system constructs a dynamic weight verification function based on the Sigmoid function, the mathematical expression of which is: In this formula, the symbol This represents the calculated credibility index, ranging from 0 to 1. A higher value indicates greater consistency in the settlement data. (Symbol) The sigmoid function has a sensitivity coefficient, which determines the steepness of the function curve near the critical point, i.e., the sensitivity of the control system to changes in error. (Symbol) This represents the system's allowable basic error tolerance. It is a preset constant used to set the range of background noise or inherent sensor error that the system can tolerate. When the weighted total deviation is less than this tolerance, the function will output a higher level of confidence.

[0046] The core innovation of this model lies in the introduction of dynamic weight coefficients. and These two coefficients are not fixed values, but are closely related to the number of tags in the current environment. The number of RFID tags read is set The function, i.e. ,and Then through the formula Maintain and The complementary relationship between them. The physical mechanism of this design lies in the fact that when the number of RFID tags read is small, the radio frequency signal is stable, and the system assigns a higher weight to the radio frequency data. However, as the number of tags increases, the reliability of radio frequency (RF) data decreases due to the increased probability of electromagnetic interference and signal collisions between tags. In this case, the system will automatically reduce the weight of the RF data. And correspondingly increase the weight of visual data. This dynamic adjustment mechanism can effectively suppress the impact of accumulated errors caused by multi-label interference on the final judgment result.

[0047] By constructing a nonlinear verification function with a dynamic weight adjustment mechanism, the system can adaptively adjust the confidence level of radio frequency (RF) and visual data according to the density of materials on site. This design overcomes the shortcomings of traditional algorithms that use fixed weights and cannot cope with complex electromagnetic environment changes, ensuring that when dense tag stacking leads to a decrease in RF identification accuracy, it can automatically rely on visual estimation results for compensation verification. Furthermore, the manual provides a detailed definition of the deviation rate, the specific mathematical expression of the verification function, and the dynamic adjustment logic of the parameters, clarifying the complete calculation path from raw sensor data to the final confidence level output.

[0048] In the preferred embodiment, considering the side-mounted structure of the RFID reader / writer 4, step S2 further includes a lateral radio frequency enhancement scanning strategy: When the binocular depth vision recognition camera 5 detects that the height of the stacked materials exceeds a preset threshold At the same time, the control module 1 controls the RFID reader / writer 4 to increase the transmission power and enable a multi-timeslot polling mechanism to ensure that it can penetrate the stacked materials and read the RFID tags on the bottom layer.

[0049] For the specific structural layout of the RFID reader / writer device with side mounting, a lateral radio frequency enhancement scanning strategy is introduced in the data acquisition step. This strategy is based on visual perception, utilizing a binocular depth vision camera located at the top to monitor the stacking status of materials on the shelf in real time. The system has a preset height threshold. This threshold represents the maximum height of a stack of materials that a side-mounted antenna can effectively penetrate under normal radio frequency power. When the actual stack height of the materials calculated by the vision system using depth data exceeds this preset threshold... When the situation arises, the control module automatically determines that there is a risk of radio frequency signal obstruction or attenuation in the current scenario, thereby triggering the enhancement mode. In enhancement mode, the control module first sends a command to the RFID reader to increase the radio frequency transmission power, enhancing the signal penetration capability and enabling the radio frequency beam to penetrate deeper into densely stacked material layers. Simultaneously, the system enables a multi-timeslot polling mechanism, which increases the number of polling time slots in the radio frequency communication protocol and performs multiple rounds of cyclic scanning to increase the probability that electronic tags at the bottom of the stack or on the side opposite to the antenna will successfully respond to the reader's polling signal.

[0050] This solution effectively overcomes the blind zone problem inherent in fixed side-mounted antennas when dealing with high-density or high-height stacked materials. Traditional side-mounted antennas often struggle to cover tags at the far end or bottom of the stacking platform, while this solution utilizes the depth perception capability of binocular vision as a priori guidance to achieve adaptive adjustment of RF parameters. By dynamically adjusting the transmission power and polling mechanism based on the stacking height, the system only increases power and scan density when necessary. This ensures that it can penetrate thick stacked material layers to read the bottom tags, significantly reducing the missed read rate, while avoiding energy waste and electromagnetic interference caused by maintaining high-power transmission when there are few materials. This vision-RF linkage control logic clearly defines the specific conditions and execution steps for triggering enhanced scanning, clarifying... The physical meaning of the trigger threshold and the technical function of the multiple time slot polling mechanism enable those skilled in the art to reproduce the control strategy based on the description in the specification.

[0051] In the preferred embodiment, the method also includes a residual inference step for unlabeled micromaterials: When the system detects When the difference exceeds the set value, control module 1 initiates residual analysis: Calculate the remaining weight not interpreted by the RFID tag ; exist Secondary feature extraction is performed on regions in the image that do not match labels; Will Perform integer programming matching with the unit weight of low-value consumables in the database. If an integer solution exists: ; in It is an integer. If the unit weight is a low-value material of a certain type, then that type of material will be automatically added to the settlement list.

[0052] The residual inference step for unlabeled small items aims to address the technical blind spot of traditional RFID checkout systems that cannot detect unlabeled items. When the control module detects the actual total physical weight collected by the weighing platform during data comparison... Significantly greater than the theoretical total weight calculated based on the read electronic tags. Furthermore, if the difference between the two exceeds a pre-set threshold, the system will automatically determine that there are unidentified physical materials and initiate a residual analysis program. This program first performs a precise calculation of the weight dimension using a formula. Calculate the remaining weight that is not interpreted by the RFID tag; this variable This represents the total physical weight of all unlabeled materials. Simultaneously, the system utilizes data collected by the binocular vision camera. Image data is used to perform secondary feature extraction on areas in the image that cannot be mapped to the spatial location of the read labels. Texture analysis or shape recognition is used to preliminarily determine the material category of objects in these unlabeled areas, such as distinguishing them as metal screws, plastic washers or other bulk consumables.

[0053] After obtaining the remaining weight and potential material categories, the core data processing unit will execute a matching algorithm based on integer programming. The system retrieves the standard unit weight data of low-value consumables stored in the database and attempts to find whether there exists an integer solution that satisfies the formula... It holds true. In this mathematical relation, the variables... The variable represents the integer value of the inferred quantity of unlabeled goods. This represents the standard unit weight of a specific low-value material in the database. If the calculation result indicates the remaining weight... Compared with the unit weight of a certain type of material If the integer multiples of the items are approximately equal within the allowable error range, and the category of the materials matches the results of visual feature extraction, the system determines that the match is successful and automatically calculates the quantity. These supplies will be added to the final settlement list without human intervention.

[0054] This innovative approach combines weight residual analysis with integer programming algorithms to provide an efficient and automated checkout process for low-value, small consumables that cannot be individually tagged with RFID tags. This design overcomes the inefficiency of existing technologies that require manual counting and data entry for small, bulk items such as screws and nuts, significantly improving the checkout counter's adaptability to mixed material scenarios. By clearly defining the criteria for triggering residual analysis, the method for calculating remaining weight, and the specific mathematical model for integer matching, the instruction manual comprehensively and clearly reveals the entire logical process from anomaly detection to automatic data entry.

[0055] In the preferred embodiment, the method also includes an anomaly correction step based on human-computer interaction: When step S5 triggers an abnormal alarm, the control module 1 projects the real-time image captured by the binocular depth vision recognition camera 5 onto the human-computer interaction display screen 2, and uses augmented reality technology to highlight the suspected abnormal material areas in the image, that is, areas where the visual volume and weight do not match, or areas where the label is not read, guiding the user to adjust the placement of materials or remove abnormal items.

[0056] The anomaly correction process based on human-computer interaction is a key step in improving user experience and checkout efficiency. When the system triggers an anomaly alarm in step S5 due to the credibility index falling below a preset threshold, control module 1 immediately initiates a visual guidance program. This program pushes the real-time video stream of the weighing platform captured by the binocular depth vision recognition camera 5 to the human-computer interaction display screen 2 and applies augmented reality technology to render the image in real time. The core data processing unit accurately locates the spatial coordinates of data anomalies based on previous consistency verification results. For example, areas where the visually estimated volume is large but does not match a sufficient weight share, or blind spots where objects are visually detected but the RFID reader fails to read the corresponding tag signal. The system marks these suspected abnormal material areas on the screen with highlighted borders, color overlays, or dynamic icons, thus intuitively instructing the user to adjust the posture of materials in specific locations to eliminate radio frequency obstruction or directly remove mixed foreign objects.

[0057] The beneficial effect of this technical solution lies in its ability to transform abstract data conflicts into intuitive visual feedback, significantly reducing the difficulty for operators in troubleshooting. Traditional checkout systems often only display "total discrepancy" or "weight error" messages when errors occur, requiring users to blindly search for the source of the problem among stacked materials, which is inefficient and frustrating. This application, however, utilizes augmented reality technology to achieve "what you see is what you get" anomaly localization, guiding users to precisely address problematic materials and achieving a qualitative leap from "error notification" to "intelligent error correction." This design fully leverages the spatial positioning value of binocular vision systems, clarifies the specific implementation logic of human-computer interaction, and not only solves the difficulty of anomaly handling mentioned in the background art but also provides sufficient implementation examples to support the technical features of human-computer interaction anomaly correction in the claims, ensuring the integrity and implementability of the technical solution.

[0058] In the preferred embodiment, step S1 also incorporates biometric authorization verification: Using a binocular depth vision camera 5 to collect facial depth information of operators, an anti-spoofing facial feature vector is constructed. ; Will Compare with the pre-stored permission database, only if the similarity is... Only when the time is right will the control module 1 unlock its control over the RFID reader / writer 4 to prevent unauthorized personnel from operating the device and causing leakage of material data.

[0059] Step S1, which incorporates biometric authentication, forms the first line of defense for ensuring device operational security. This step fully utilizes the hardware capabilities of the binocular depth vision camera 5, simultaneously acquiring 3D depth information of the face while capturing the operator's facial image. The core data processing unit then constructs a facial feature vector with anti-spoofing capabilities based on this depth data, denoted as... Unlike traditional two-dimensional face recognition, this feature vector contains spatial three-dimensional structural information of the face, effectively distinguishing real, live faces from flat photographs or video images. Subsequently, the system will construct... The similarity between the feature template of the user and the pre-existing feature template in the security chip or encrypted database is calculated and denoted as the feature similarity. The system has set extremely strict decision-making logic, that is, only if the calculated similarity is... When the value is greater than 0.98, the control module 1 will issue a command to unlock the control authority of the RFID reader / writer 4 and enable the radio frequency transmission circuit to enter the working state.

[0060] By introducing a depth-based liveness detection mechanism, the authentication level of the materials checkout counter has been significantly improved, effectively defending against spoofing attacks using photos and videos. The activation permissions of the RFID reader / writer are rigidly linked to high-precision facial recognition results, ensuring that only authorized personnel can trigger the materials scanning process. This design not only prevents unauthorized personnel from misoperating or maliciously tampering with inventory data at the source, but more importantly, it avoids the risk of leakage of sensitive material tag information due to unauthorized activation of RFID readers / writers, building a robust physical and digital dual security barrier for the materials management system. The instruction manual provides a detailed description of the feature vector construction method, threshold setting standards, and permission linkage logic.

[0061] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. An integrated materials checkout counter, characterized in that: A vertical support column is provided on one side of the weighing platform (3), and a control module (1) is provided on the vertical support column. A human-machine interaction display screen (2) is provided on the control module (1), and a binocular depth vision recognition camera (5) is provided above the human-machine interaction display screen (2). The field of view of the binocular depth vision recognition camera (5) covers the area where the weighing platform (3) is located. An RFID reader / writer device (4) is installed on one side of the vertical support column.

2. The integrated material checkout counter according to claim 1, characterized in that: The control module (1) integrates a core data processing unit, which is connected to the human-machine interaction display screen (2), the weighing platform (3), the RFID reader (4) and the binocular depth vision recognition camera (5). The core data processing unit is configured to execute a material consistency verification algorithm based on visual radio frequency gravity three-dimensional feature fusion.

3. The integrated material checkout counter according to claim 1, characterized in that: The simulated gravity signal collected by the weighing platform (3); RFID reader (4) collects radio frequency tag data; The image stream data collected by the binocular depth vision recognition camera (5) are all aggregated into the control module (1) for spatiotemporal alignment and fusion processing.

4. The method of using an integrated material checkout counter according to any one of claims 1-3, characterized in that: The method is executed by the core data processing unit in the control module (1), including: S1. Initialize the system, activate the binocular depth vision recognition camera (5), and monitor in real time whether there is an object placed on the weighing platform (3); S2. Once the placement of the materials is detected to be stable, multi-dimensional data acquisition is triggered simultaneously: the total physical weight is obtained through the weighing platform (3). The electronic tag set is read by the RFID reader (4). The three-dimensional point cloud and texture data of the materials are acquired by a binocular depth vision recognition camera (5). ; S3. Perform visual volumetric physical density coupling estimation, based on Calculate the visual estimated weight of materials ; S4. Perform RF visual weight ternary consistency verification and calculate the credibility index of this checkout process. ; S5, if Greater than the preset safety threshold If so, a settlement list will be generated on the human-computer interaction display screen (2); if Less than or equal to If this occurs, an abnormal alarm will be triggered.

5. The method of using the integrated material checkout counter according to claim 4, characterized in that: In step S3, the visually estimated weight is calculated. The steps are as follows: Using point cloud segmentation algorithms from Extract the target area of ​​materials on the shelf and calculate the bounding box volume of the materials. ; Deep neural networks are used to identify material types and match corresponding density coefficients from a database. ; The calculation formula is: ； in, This represents the number of objects visually identified. This is a stacking correction factor used to compensate for volume estimation errors caused by visual occlusion.

6. The method of using the integrated material checkout counter according to claim 5, characterized in that: In step S4, the credibility index is calculated. The mathematical model is as follows: First, analyze the electronic tag set. Obtain the theoretical total weight based on tag information ; Define RF deviation rate ; Define visual deviation rate ; Construct a dynamic weight verification function: ； in, The sensitivity coefficient of the Sigmoid function. This represents the system's basic error tolerance. and The dynamic weighting coefficients satisfy the following conditions: , ,in The system automatically reduces the weight of radio frequency data as the number of RFID tags read increases. To prevent errors caused by interference from multiple labels.

7. The method of using the integrated material checkout counter according to claim 4, characterized in that: In view of the structural characteristics of the RFID reader / writer device (4) being side-mounted, step S2 also includes a lateral radio frequency enhancement scanning strategy: When the binocular depth vision recognition camera (5) detects that the height of the stacked materials exceeds the preset threshold At that time, the control module (1) controls the RFID reader (4) to increase the transmission power and enable the multiple time slot polling mechanism to ensure that it can penetrate the stacked materials to read the RFID tags on the bottom layer.

8. The method of using the integrated material checkout counter according to claim 4, characterized in that: The method of use also includes residual inference steps for unlabeled micromaterials: When the system detects When the difference exceeds the set value, the control module (1) initiates residual analysis: Calculate the remaining weight not interpreted by the RFID tag ; exist Secondary feature extraction is performed on regions in the image that do not match labels; Will Perform integer programming matching with the unit weight of low-value consumables in the database. If an integer solution exists: ； in It is an integer. If the unit weight is a low-value material of a certain type, then that type of material will be automatically added to the settlement list.

9. The method of using the integrated material checkout counter according to claim 4, characterized in that: The usage method also includes anomaly correction steps based on human-computer interaction: When step S5 triggers an abnormal alarm, the control module (1) projects the real-time image captured by the binocular depth vision recognition camera (5) onto the human-computer interaction display screen (2), and uses augmented reality technology to highlight the suspected abnormal material area in the image, that is, the visual volume and weight do not match, or the area where the label is not read, to guide the user to adjust the placement of the material or remove the abnormal item.

10. The method of using the integrated material checkout counter according to claim 4, characterized in that: Step S1 also incorporates biometric authorization verification: Using a binocular depth vision recognition camera (5), facial depth information of the operator is collected to construct an anti-spoofing face feature vector. ; Will Compare with the pre-stored permission database, only if the similarity is... Only when the time is right will the control module (1) unlock its control authority over the RFID reader (4) to prevent unauthorized personnel from operating the device and causing leakage of material data.

Images