Passenger luggage management method and system based on channel vision and RFID fusion

Abstract

The application discloses a passenger luggage management method and system based on channel vision and RFID fusion, and the management method comprises the following steps: performing a line body collection mode and a vision collection mode in parallel in a luggage preparation area, and acquiring line body mode binding data and vision mode binding data associated with passengers and luggage baskets respectively; checking two groups of data based on a preset fusion rule to determine final person-basket binding data; binding security luggage images generated by an X-ray machine and recognized basket numbers to generate basket-luggage binding data; and associating luggage images to corresponding passenger identity information according to the person-basket binding data and the basket-luggage binding data to generate traceable security management records. The application solves the management defects in the prior art, such as the fact that passenger information and luggage information cannot be automatically associated, the fact that the association reliability is low, and the fact that tray reuse management is chaotic.

Description

Technical Field

[0001] This invention belongs to the field of security inspection management technology, and specifically relates to a passenger baggage management method and system based on the integration of channel vision and RFID. Background Technology

[0002] In the field of civil aviation security management, achieving full-process correlation between passengers and their carry-on baggage (hereinafter referred to as "person-baggage association") is the core requirement of security process management and a fundamental management means to prevent baggage from being taken by mistake, prohibited items from being missed, and to trace responsibility for security incidents. Relevant regulations of the Civil Aviation Administration of China clearly require that the security information system should be able to fully collect passenger identity information, baggage security images, and baggage inspection information, and have full-process traceability capabilities.

[0003] Currently, the management process for carry-on baggage security checkpoints at airports has the following main shortcomings: 1. Isolated and uncorrelated data management: In the current security check process, there is no direct link between the baggage inspection images generated by X-ray machines and passenger identity information, making baggage inspection images isolated data. When a security investigation is needed (e.g., suspicious baggage is discovered), security personnel can only perform vague tracing by reviewing multi-channel surveillance videos and manually comparing timestamps and passenger appearances. This method is extremely inefficient and inaccurate, failing to meet the Civil Aviation Administration's stringent traceability requirements for the security information management system.

[0004] 2. Management model relies on manual labor: The existing management methods do not incorporate automated information collection mechanisms, do not read luggage basket labels, and do not collect passenger identity information. The correspondence between luggage and passengers relies entirely on manual guidance from on-site security personnel and passengers' self-discipline. This extensive management model not only easily leads to management incidents such as misplaced luggage and fraudulent claims, but also results in a serious lack of management data, hindering the digital and intelligent upgrading of security management.

[0005] 3. Chaotic management of reused luggage baskets: Luggage baskets in the security checkpoint are reused by multiple passengers. The existing management system cannot effectively distinguish the ownership of the baskets at different times, which can easily lead to the incorrect association of historical passenger information with the current passenger, causing confusion in management data and affecting the accuracy of traceability.

[0006] 4. Lack of fault tolerance mechanisms in management processes: Some airports have begun to try introducing facial recognition or RFID technology for passenger baggage binding, but their technical architecture is mostly based on a single data source (pure vision or pure RFID), which is prone to binding failure in complex environments (such as passenger obstruction, changes in lighting, and tag reading failure). Existing management solutions lack effective compensation and integration mechanisms, leading to management process interruptions and making it impossible to achieve full-time, full-coverage passenger baggage association management.

[0007] Therefore, there is an urgent need for a management method that can accurately link passengers and their carry-on baggage throughout the entire process, in order to improve the informatization level and process reliability of security inspection management and meet the compliance management requirements of civil aviation security inspection. Summary of the Invention

[0008] The purpose of this invention is to provide a passenger and baggage management method and system based on the fusion of channel vision and RFID, which solves the management defects of existing security check management methods, such as the inability to automatically associate passenger information and baggage information, low reliability of association, and chaotic management of tray reuse.

[0009] To achieve the above objectives, this invention provides a passenger and baggage management method based on the fusion of channel vision and RFID, comprising the following steps: S1. Passenger basket binding management: In the luggage placement area, the line management mode and the visual management mode are executed in parallel to obtain the line mode binding data and the visual mode binding data used to represent the relationship between passengers and luggage baskets. The line management mode includes: reading the RFID tag of the luggage basket through the first RFID reader on the luggage preparation table, triggering the face auxiliary screen to obtain the current passenger's face information, and generating line mode binding data; The visual management mode includes: continuously acquiring images of the luggage area through a camera, identifying the passenger's facial information and the visual markings on the surface of the passenger's luggage basket from the images, and generating visual mode binding data; The code stored in the RFID tag of the luggage basket is the same as the code carried by the visual tag, both of which are the basket number of the luggage basket; Both the line mode binding data and the visual mode binding data contain corresponding confidence levels, which are calculated as follows: The following weighted recombinations were used to calculate the confidence level of line mode management: first weighted recombinations were assigned to the quality assessment scores of RFID tag reading, face capture by face-assisted screen, and face identity comparison and matching under the visual management mode; second weighted recombinations were assigned to the quality assessment scores of visual tag recognition, face capture by camera, and face identity comparison and matching under the visual management mode, and the results were calculated to obtain the confidence level of visual mode management. S2. Data Fusion Management: Before the luggage basket enters the X-ray machine, the binding data of the line mode and the binding data of the visual mode are retrieved using the basket number as the main index; then, the binding data of the two modes are checked for consistency based on preset fusion rules, and the final person-basket binding data is determined according to the check results; wherein, the fusion rules include: When only one of the line body mode binding data and the visual mode binding data is obtained, the binding data set is directly used as the final human basket binding data. When the passenger identities in the two sets of bound data are consistent, either set of data is selected as the final passenger-basket binding data. When the passenger identities in the two sets of bound data are inconsistent, the one with higher confidence level is selected as the final person-basket binding data. When only one set of data is valid among the two sets of binding data, the valid data is selected as the final person-basket binding data. S3, Bag Binding Management: When a luggage basket passes through an X-ray machine, the system acquires the luggage security inspection image generated by the X-ray machine, identifies the visual markings on the surface of the luggage basket from the image, obtains the basket number, binds the luggage security inspection image with the identified basket number, and generates bag binding data. S4. Person-bag binding management: Based on the person-basket binding data and basket-bag binding data, the baggage security inspection image is associated with the corresponding passenger identity information to generate person-bag binding data between the passenger and their carry-on baggage, and stored in the security inspection management database for security management traceability.

[0010] Furthermore, in step S2, the line body mode management confidence level is... The calculation formula is:

[0011] in, The score is used to evaluate the quality of RFID tag reading. The value range is [0,1]. A successful reading without interference is scored as 1, a successful reading after multiple retries is scored as 0.8, a reading due to weak signal or crosstalk causing unstable reading is scored as 0.5, and a reading failure is scored as 0. The value is the quality assessment score of the face capture by the face-assisted screen, with a range of [0,1]. A value of 1 is given for a complete and unobstructed frontal face capture, a value of 0.7 is given for a side face capture or a face with partial occlusion but which does not affect the extraction of facial features, a value of 0.3 is given for a face capture with occlusion or blur that affects the extraction of facial features, and a value of 0 is given for no valid face capture. The value is the evaluation score for face identity comparison and matching in the line body management mode. The value range is [0,1]. The value is 1 for a complete match of face identity features, 0.8 for a matching similarity of more than 50% and less than 100%, 0.5 for a matching similarity of less than 50%, and 0 for a failed match. , , The weight coefficients of the first weight group satisfy the following conditions: ,and =0.5, =0.3, =0.2; Visual pattern management confidence The calculation formula is:

[0012] in, The evaluation score for the confidence level of visual identification is in the range of [0,1]. The score is 1 for a complete and clear identification of the basket number, 0.8 for a blurry basket number but accurate identification, 0.5 for a basket number that is partially obscured, worn or blurred, which makes the identification uncertain, and 0 for no identification. The score is the quality assessment value for facial capture by the camera, ranging from [0,1]. The value selection rules are the same as those for facial recognition. same; The evaluation score for facial identity matching in visual management mode, and the scoring rules are the same as those for facial recognition. same; , , The weight coefficients of the second weight group satisfy the following conditions: ,and =0.4, =0.3, =0.3.

[0013] Furthermore, in step S1, in the online body management mode, after the first RFID reader reads the RFID tag of the luggage basket, it identifies the RFID tag to obtain the basket number B, and triggers the face-capturing auxiliary screen at the corresponding preparation station to start face capture. Based on the captured face information, it calls the OneID identity service to obtain the passenger's identity information. Record the current binding time And channel number C, generate line mode binding data :

[0014] Where R is the reset identifier, and at this time it is the initial state N, which indicates that the luggage basket has completed the current passenger's basket binding but has not yet completed the security check reset; Facial image feature data captured by a face-assisted screen; Confidence level for line management mode; In visual acquisition mode, images of the luggage area are continuously captured by a camera. OCR technology is used to identify visual markers from the images, obtaining the baggage number B. Passenger facial information is then extracted from the images, and the OneID identity service is invoked to retrieve passenger identity information. Record the current binding time And channel number C, generate visual pattern binding data :

[0015] in, Facial image feature data captured by the camera; Confidence management for visual patterns; The update rule for the reset flag R is: using Key=(B,C, The unique index key is used to construct the data storage record for the person-basket binding, where, The system binds the person-basket to a specific time. When the baggage basket carrying the luggage enters the exit of the X-ray machine, the second RFID reader deployed on the exit side of the X-ray machine reads the RFID tag of the baggage basket. Based on the same index key, the reset tag in the corresponding person-basket binding data record is updated to R=Y, indicating that the baggage basket has completed the security check process and entered the reset state. The baggage basket can then be reused.

[0016] Furthermore, in step S3, if the visual markings on the surface of the luggage basket cannot be effectively identified from the luggage security inspection image, the image anomaly fallback compensation management mechanism is triggered. The management method of this fallback compensation management mechanism is as follows: obtain the acquisition time and security channel number of the current luggage security inspection image, and find the luggage basket number under the same security channel whose RFID acquisition time is less than and closest to the luggage security inspection image acquisition time from the historical management data read by the second RFID reader deployed at the front end of the X-ray machine entrance or the back end of the exit. Use this as the basket number corresponding to the luggage security inspection image for compensation binding management, and generate the final basket binding data.

[0017] Furthermore, in step S4, both the person-basket binding data and the basket-bag binding data include the basket number, the security checkpoint number, and the corresponding binding time. Associating the baggage security check image with the corresponding passenger identity information includes: extracting the basket number, the security checkpoint number, and the basket-bag binding time from the basket-bag binding data; using the basket number and the security checkpoint number in the basket-bag binding data as the main index; and filtering out valid person-basket binding data from the person-basket binding data that has the same basket number, the same security checkpoint number, and the person-basket binding time that meets the following time constraints.

[0018] in, The time is bound to the basket; W is the preset time window threshold, which is the maximum time for the luggage basket to flow from the luggage preparation table to the X-ray machine; Binding time to the basket; The retrieved valid person-basket binding data is sorted in reverse order by the binding time. The sorted dataset is traversed, and the first record of the person-basket binding data is taken. The passenger identity information in the person-basket binding data is associated with the corresponding baggage security inspection image in the bag binding data.

[0019] Furthermore, in step S4, the time when the passenger identity information of the first record is associated with the baggage security image is set as the baseline binding time. The remaining baggage basket binding data is then traversed. If the difference between the binding time of the remaining baggage basket binding data and the baseline binding time does not exceed the baggage basket reuse time, it is determined to be valid binding data associated with the same baggage basket. The passenger identity information in it is also associated with the corresponding baggage security image. If multiple different passenger identity information is bound, it is manually verified through captured images to achieve unique passenger identity information binding. If it exceeds the baggage basket reuse time, it is determined to be historical invalid data generated by baggage basket reuse and is filtered out to avoid passenger mismatch caused by baggage basket cyclic reuse.

[0020] Furthermore, in step S4, the reuse time of the luggage basket is dynamically predicted using a sliding time window + mean drift clustering algorithm.

[0021] Furthermore, in step S4, when the corresponding passenger identity information cannot be retrieved in the passenger basket binding data, the passenger basket management data missing fallback compensation management mechanism is triggered. The management method of this fallback compensation management mechanism is as follows: based on the current binding time of the baggage basket, a set of passenger management information in the same security check channel and whose security check time is within the security check time window is selected from the passenger security check records. The passenger identity information in the passenger management information set is used as the corresponding passenger for the current baggage for fallback binding management, and a fallback management record is generated. The security check time window is the time period between the passenger's identity security check at the counter and the baggage security check in the baggage placement area.

[0022] Furthermore, in steps S1 and S3, OCR recognition technology is used to identify visual identifiers. Two visual identifiers are set on the side of the luggage basket, and an arrow is set between the two visual identifiers. The direction of the arrow indicates the running direction of the luggage on the track. The code of the visual identifier is read according to the direction indicated by the arrow. The visual identifiers on both sides of the arrow are identified and parsed. If the codes identified on both sides are consistent, a unique basket number binding result is output. If the codes identified on both sides are inconsistent, two independent recognition result sets are output, which are then manually reviewed or verified by the system.

[0023] This application also provides a passenger and baggage management system based on the fusion of channel vision and RFID, for implementing the above management method, including: The luggage baskets are equipped with RFID tags and visual tags. The RFID tags are located on the bottom of the luggage baskets, and the visual tags are located on the sides of the luggage baskets. The line management data acquisition module includes a first RFID reader and a face recognition auxiliary screen deployed at the baggage preparation station, used to execute the line management mode; The visual management acquisition module includes a camera deployed on the side of the baggage conveyor track for executing the visual management mode; The auxiliary RFID management and acquisition module includes a second RFID reader deployed on the entrance and exit sides of the X-ray machine to assist in the execution of basket binding management steps and image anomaly compensation management mechanism. The security inspection management data processing unit is connected to the line management acquisition module, the vision management acquisition module, the auxiliary RFID management acquisition module and the X-ray machine respectively. It is used to perform the steps of personnel binding management, data fusion management, basket and bag binding management, personnel and bag corresponding management and all fallback compensation management mechanisms, and generate traceable security inspection management records. The security inspection management database is used to store security inspection management records.

[0024] After adopting the above solution, the beneficial effects of the present invention are as follows: 1. Improved Efficiency and Accuracy of Security Check Management and Traceability: This invention automatically links passenger identity information, baggage basket labels, and X-ray images of baggage security checks throughout the entire process, generating a complete and reliable security check management data chain. When conducting security audits or incident tracing, managers can directly retrieve corresponding baggage security images and records from the security check management database within seconds based on passenger name, flight number, or ID information. This completely eliminates the inefficient method of manually reviewing surveillance videos frame by frame, significantly improving traceability management efficiency and accuracy while reducing labor costs.

[0025] 2. Optimize security checkpoint management processes and enhance passenger experience: The dual-mode parallel data collection architecture of this invention enables seamless person-bag binding without requiring passengers to cooperate or change their security check habits. Simultaneously, through intelligent data fusion and anomaly compensation mechanisms, it reduces manual intervention and duplicate security checks caused by recognition failures, making security checkpoint management smoother and more efficient, thus improving the passenger travel experience.

[0026] 3. Provide data support for security inspection management decisions: The precise correlation management data of "passenger-baggage basket-baggage security inspection image" generated by this invention can serve as the basic resource for security inspection big data analysis, providing sustainable data support for subsequent data analysis, security inspection process optimization and service upgrades, and promoting the intelligent transformation of the transportation sector.

[0027] 4. Enhance the integrity and compliance of security check management: This invention designs an anomaly compensation management mechanism. When anomalies occur, such as failure of baggage security check image recognition visualization or missing data of the person basket binding, the system can automatically activate backup management processes such as the image anomaly fallback compensation management mechanism and the person basket management data missing fallback compensation management mechanism to ensure that the management information of every piece of baggage and every passenger is not lost or interrupted.

[0028] 5. Low implementation cost and high management and promotion value: The hardware deployment solution of this invention only requires the addition of modular data acquisition equipment at key nodes of existing security checkpoints, without the need for large-scale modifications to the main structure of the security checkpoints, X-ray machines, or return basket lines. This results in low implementation cost and a short cycle. Furthermore, the core management methods can be standardized and replicated, making it applicable to various airport security checkpoint scenarios nationwide. It has significant demonstrative value and promotional significance for promoting the standardization of smart security checkpoint management and improving the overall intelligent level of security management in the industry. Attached Figure Description

[0029] Figure 1 This is a structural diagram of the baggage security inspection equipment of the present invention; Figure 2 This is a schematic diagram of the structure of the luggage basket of the present invention; Figure 3 This is a data flow diagram of the management method of the present invention.

[0030] Label Explanation: 1. Luggage placement area; 2. Luggage conveyor track; 3. Camera; 4. X-ray machine; 5. Luggage preparation table; 6. First RFID reader; 7. Second RFID reader; 8. Face recognition auxiliary screen; 9. Luggage basket; 91. Visual signage; 92. Signage arrows. Detailed Implementation

[0031] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] Key references Figures 1-3 This invention provides a passenger baggage management method based on the fusion of channel vision and RFID. The management method is executed by a management system that does not require large-scale modifications to the existing security checkpoint structure, X-ray machine, or baggage return line. It only requires the deployment of several first RFID readers 6 and facial recognition auxiliary screens 8 in the baggage preparation area 5, second RFID readers 7 on the entrance and exit sides of the X-ray machine 4, and a high-definition camera 3 installed on the side of the baggage conveyor track 2 of the baggage preparation area 5. The management system specifically includes baggage baskets 9, a line management acquisition module, a vision management acquisition module, an auxiliary RFID management acquisition module, a security check management data processing unit, and a security check management database.

[0033] The luggage basket 9 is equipped with an RFID tag (not shown in the figure) and a visual tag 91. The RFID tag is located at the bottom of the luggage basket 9, and the visual tag 91 is located on the side of the luggage basket 9. Specifically, two visual tags 91 can be set on the side of the luggage basket 9, with an arrow 92 between the two visual tags 91. The direction of the arrow 92 indicates the running direction of the luggage on the luggage conveyor track 2. The code of the visual tag 91 is read according to the direction indicated by the arrow 92. The visual tag 91 is preferably a printed tag with a string of codes printed on it. This code is the basket number of the luggage basket 9. The code stored in the RFID tag is consistent with the code carried by the visual tag 91, which is also the basket number of the luggage basket 9. Figure 2 As shown, the printed code on the surface of the luggage basket is "31331", so the basket number of the luggage basket is "31331", and the code obtained by the RFID reader from reading the RFID tag is also "31331".

[0034] like Figure 1 As shown, the line management acquisition module includes a first RFID reader 6 and a face recognition auxiliary screen 8 deployed at the luggage preparation station 5, which are used to execute the line management mode in the management method.

[0035] The visual management acquisition module includes a camera 3 deployed on the side of the baggage conveyor track 2, used to execute the visual management mode in the management method.

[0036] The auxiliary RFID management and acquisition module includes a second RFID reader 7 deployed on the entrance and exit sides of the X-ray machine 4, which is used to assist in the execution of the basket binding management steps and the image anomaly compensation management mechanism in the management method.

[0037] The security inspection management data processing unit is connected to the line management acquisition module, the vision management acquisition module, the auxiliary RFID management acquisition module and the X-ray machine 4, respectively. It is used to perform the steps of personnel-basket binding management, data fusion management, basket-bag binding management, personnel-bag correspondence management and various fallback compensation management mechanisms, and generate traceable security inspection management records and store them in the security inspection management database.

[0038] Key references Figure 3 The management method includes the following steps: S1. Baggage Binding Management: After a passenger arrives at the baggage placement area 1 and selects a baggage basket 9, and places the baggage basket 9 on the baggage preparation table 5, the line management mode and the visual management mode are executed in parallel to obtain line mode binding data and visual mode binding data used to represent the relationship between the passenger and the baggage basket.

[0039] The line management mode includes: reading the RFID tag on the bottom of the luggage basket through the first RFID reader 6 on the luggage preparation station 5, identifying the RFID tag to obtain the basket number B, and triggering the face capture screen 8 at the corresponding workstation to capture the current passenger's face information (i.e., facial biometrics). Based on the captured face information, the OneID identity service (a backend system that binds and verifies the passenger's face information with its unique identity identifier, which can be an ID card number or a unique identity number set by the system) is invoked to obtain the passenger's identity information. Record the current binding time And security checkpoint number C, generate line mode binding data :

[0040] Where R is the reset identifier, and at this time it is the initial state N, which indicates that the luggage basket has completed the current passenger's basket binding but has not yet completed the security check reset; Facial feature data captured by the face-assisting screen 8; The confidence score for the line management mode, with a value range of [0,1], is calculated by weighting the quality assessment scores of RFID tag reading, face capture by the face-assisted screen, and face identity matching under the visual management mode, respectively, using the first weighted reassembly. The calculation formula is as follows:

[0041] in, The quality assessment score for RFID tag reading ranges from [0,1]. A score of 1 indicates a successful reading without interference, a score of 0.8 indicates a successful reading after multiple retries, a score of 0.5 indicates an unstable reading (caused by weak signal, crosstalk, or intermittent reading), and a score of 0 indicates a failed reading. The quality assessment score for face capture by the face-assisting screen is based on the clarity of the captured image and whether it is occluded. The value range is [0,1]. A value of 1 is given for a complete and unobstructed frontal face with normal lighting. A value of 0.7 is given for a slightly obstructed side face or a face with partial occlusion but which does not affect the extraction of facial features. A value of 0.3 is given for a face with large-area occlusion, backlighting, or blurriness that affects the extraction of facial features. A value of 0 is given for no valid face captured. This is the evaluation score for facial identity comparison and matching in the line management mode. It compares the facial information extracted by the facial assistance screen with the passenger facial information with the highest matching degree in the OneID identity service. The value range is [0,1]. A value of 1 is set for a complete match of facial identity features (i.e., a matching similarity of 100%), a value of 0.8 is set for a high similarity (e.g., a matching similarity of more than 50% and less than 100%), a value of 0.5 is set for a low similarity (e.g., a matching similarity of less than 50%), and a value of 0 is set for a failed match. , , The weight coefficients of the first weight group satisfy the following conditions: And preferred =0.5, =0.3, =0.2.

[0042] The visual management mode includes: during the entire period when passengers are retrieving their baskets and placing their carry-on luggage, the camera continuously captures images of the luggage placement area 1. The camera's shooting area covers both the passenger's face and the luggage basket. Then, OCR recognition technology is used to identify a visual identifier from the image to obtain the basket number B, and the passenger's facial information is extracted from the image. Finally, the OneID identity service is invoked to obtain the passenger's identity information. Record the current binding time And channel number C, generate visual pattern binding data :

[0043] Here, R is the initial state. Facial image feature data captured by the camera; The confidence score for visual pattern management, with a value range of [0,1], is obtained by weighting the evaluation scores of the visual identifier recognition confidence score, the quality evaluation score of the face captured on-site by the camera, and the evaluation score of the face identity comparison and matching under the visual management mode, respectively, using a second weighting method. The calculation formula is as follows:

[0044] in, The evaluation score for the confidence level of visual identification marks represents the accuracy and reliability of the visual algorithm in recognizing luggage basket numbers. The score is graded according to the clarity of the basket number image and the accuracy of the recognition matching. The value range is [0,1]. The value is 1 for the basket number that is completely clear, unobstructed and undamaged, and has a very high matching degree. The value is 0.8 for the basket number that is slightly blurry or has a small amount of dirt interference but is recognized accurately. The value is 0.5 for the basket number that is partially obscured, severely worn or blurry, which makes the algorithm barely recognize or the recognition is uncertain. The value is 0 for the basket number that cannot be recognized (which may be caused by severe obstruction, inability to identify, or recognition errors). This is a quality assessment score for face capture by the camera, representing the completeness, clarity, and lighting adaptability of the high-definition camera's captured passenger face images. The score is graded according to the imaging state, with a value range of [0,1]. The scoring rules are the same as... same; The evaluation score for facial identity matching in the visual management mode compares the matching degree between the facial information extracted by the camera and the passenger facial information with the highest matching degree in the OneID identity service. The value selection rules are the same as those for... same; , , The weight coefficients of the second weight group satisfy the following conditions: And preferred =0.4, =0.3, =0.3.

[0045] The update rule for the reset flag R is: using Key=(B,C, The unique index key is used to construct the data storage record for the person-basket binding, where, The time for binding the person basket is set. When the baggage basket carrying the luggage enters the exit of the X-ray machine, the second RFID reader 7 deployed on the exit side of the X-ray machine reads the RFID tag of the baggage basket. According to the same index key, the reset tag in the corresponding person basket binding data record is updated to R=Y, indicating that the baggage basket has completed the security check process and entered the reset state. The baggage basket can be reused in the cycle.

[0046] Key references Figure 2 After the camera captures an image, when identifying and obtaining the basket number B from the image, the visual markings on both sides of the arrow are identified and analyzed. If the codes on both sides are consistent, a unique basket number binding result is output; if the codes on both sides are inconsistent, two independent sets of recognition results are output, which are then manually verified by the image to ensure accurate acquisition of the basket number and avoid recognition errors caused by image deformation, dirt occlusion, or angle deviation, thereby improving the fault tolerance and recognition robustness of basket binding.

[0047] S2. Data Fusion Management: Before the luggage basket enters the X-ray machine, the binding data of the line mode and the binding data of the visual mode are retrieved using the basket number as the main index. Then, the binding data of the two modes are checked for consistency based on the preset fusion rules, and the final binding data of the person and basket is determined according to the check results.

[0048] Specifically, the index fusion matching index condition is Key=(B,C,T,R=N), where T is... and Using basket number B as the primary index, passenger identity information can be retrieved. , and confidence level Then, the data is fused according to the fusion rules to output the final person-basket binding information. The fusion rules include: When only one of the line body mode binding data and the visual mode binding data is obtained, the binding data set is directly used as the final human basket binding data. When the passenger identities in the two sets of bound data are consistent, either set of data is selected as the final passenger-basket binding data. When the passenger identities in the two sets of bound data are inconsistent, the one with higher confidence level is selected as the final person-basket binding data. If only one set of data is valid among the two sets of binding data, the valid data will be selected as the final person-basket binding data.

[0049] The dual-mode data fusion mechanism described above compensates for the binding failure caused by environmental occlusion, shooting angle, and device omissions in the single recognition mode, thereby improving the integrity, accuracy, and stability of the human-basket binding.

[0050] S3. Baggage Binding Management: The baggage binding stage establishes a unique association between the RFID tag on the baggage basket and the baggage security inspection image. When the baggage basket carrying the passenger's carry-on luggage enters the X-ray machine along the conveyor track, the X-ray machine performs a through-scan scan of the luggage inside the basket, generating a baggage security inspection image, which is then pushed to the security inspection management data processing unit in real time. The security inspection management data processing unit uses high-precision OCR image recognition technology to identify the visual markings on the surface of the baggage basket from the baggage security inspection image, obtaining the baggage basket number. The identification method is the same as in step S1. The baggage security inspection image is then bound to the identified basket number to generate baggage binding data. ,in, Number the baggage security inspection images. The time is then used to bind the baggage to the security checkpoint. Therefore, the baggage binding data includes four core binding data points: security checkpoint number, baggage number, image number, and binding time. This enables a full attribute association between the baggage baggage and the security check image, providing a complete data foundation for subsequent person-baggage association based on the person-baggage binding data.

[0051] When identifying basket numbers from baggage security images, if the basket number cannot be identified due to image blurring or obstruction of visual markings, an image anomaly fallback compensation management mechanism is triggered. The management method of this fallback compensation management mechanism is as follows: obtain the current baggage security image acquisition time and security channel number. Using these data as indexes, search for the baggage basket number under the same security channel with an RFID acquisition time less than and closest to the baggage security image acquisition time from the historical management data read by the second RFID reader 7 deployed on the X-ray machine entrance or exit side. Use this as the basket number corresponding to the baggage security image for compensation binding management, and generate the final basket binding data to ensure that the image and baggage tray are not lost or lost.

[0052] S4. Person-bag binding management: Based on the person-basket binding data and basket-bag binding data, the baggage security inspection image is associated with the corresponding passenger identity information to generate person-bag binding data between the passenger and their carry-on baggage, and stored in the security inspection management database for security management traceability.

[0053] Specifically, associating baggage security inspection images with corresponding passenger identity information includes: extracting the basket number, security check channel number, and basket binding time from the basket binding data; constructing multi-dimensional search conditions; using the basket number and security check channel number from the basket binding data as the main index; and filtering out valid human-basket binding data from the human-basket binding data that has the same basket number, the same security check channel number, and the human-basket binding time meets the following time constraints.

[0054] in, Bind the time to the basket; W is the preset time window threshold, which is the maximum time for the luggage basket to flow from the luggage preparation table to the X-ray machine. Since this maximum time generally does not exceed 10 minutes, W is set to 10 minutes. Binding time to the basket; The retrieved valid person-basket binding data is sorted in reverse order by the binding time. The sorted dataset is traversed, and the first record of the person-basket binding data is taken. The passenger identity information of the person-basket binding data is associated with the corresponding baggage security inspection image in the basket binding data.

[0055] To ensure no data is missed, the remaining data also needs to be filtered. The time when the passenger identity information of the first record is associated with the baggage security image is set as the baseline binding time. The remaining baggage basket binding data is then traversed. If the difference between the binding time of the remaining baggage basket binding data and the baseline binding time does not exceed the baggage basket reuse time, it is determined to be valid binding data associated with the same baggage basket. The passenger identity information in this data is also associated with the corresponding baggage security image. If multiple different passenger identity information is bound, it is manually verified through captured images to achieve unique passenger identity binding. If it exceeds the baggage basket reuse time, it is determined to be invalid historical data generated by baggage basket reuse and is filtered out to avoid passenger mismatch caused by baggage basket recycling.

[0056] Preferably, the reuse time of the luggage basket is affected by passenger flow. If it is set to a fixed value, an excessively large value may cause the loss of passenger data binding, while a small value may cause multiple bindings of passenger data binding. Therefore, a sliding time window + mean drift clustering algorithm is used to dynamically predict and obtain the value. It is a dynamic value that changes according to the peak changes of passenger flow. For example, during peak flow, the reuse time of the luggage basket is predicted to be 1-2 minutes, and during off-peak flow, it is predicted to be 3 minutes or more.

[0057] When the corresponding passenger identity information cannot be retrieved in the passenger basket binding data, the passenger basket management data missing fallback compensation management mechanism is triggered. The management method of this fallback compensation management mechanism is as follows: based on the current binding time of the baggage basket, the passenger management information set of the same security check channel and the security check time is within the security check time window is selected from the passenger security check records. The passenger identity information in the passenger management information set is used as the corresponding passenger of the current baggage for fallback binding management, and fallback management records are generated to achieve no omissions and full coverage of passenger baggage traceability and control.

[0058] The security check time window refers to the time period between when a passenger undergoes identity security check at the counter and when their luggage is checked in the baggage area. The security check time window can be represented as... , , These are the upper and lower offsets of the security check time window, used to define the earliest start and nearest end of the security check time window, and are optimized accordingly. =75s, =20s.

[0059] The following examples, through specific embodiments, further illustrate this point: The embodiment sets the luggage basket reuse time to 2 minutes and the time window threshold W to 10 minutes. The person-basket binding records for each embodiment are shown in Table 1: Table 1 - Person-Basket Binding Record Table

[0060] As shown in Table 1, both the line management mode and the visual management mode in Example 1 identified the passenger ID (i.e., the passenger's identity identifier in the OneID identity service), and the identified passenger IDs were consistent, so they were directly bound to that passenger ID. However, neither the line management mode in Example 2 nor the visual management mode in Example 3 retrieved the passenger ID, so both were directly bound to the passenger ID identified by the other mode. In Example 4, the passenger IDs identified by the two modes were inconsistent, so the mode with higher confidence was bound. In Example 5, the basket numbers were the same as in Example 1, but the collection times were different, indicating that the binding records were after the luggage baskets were reused. The above examples fully demonstrate the fusion rules for dual-mode binding data set in this application.

[0061] Table 2 shows the binding records of the frame package and the person package for each embodiment: Table 2 - Frame and Person-Package Binding Record Table

[0062] According to Table 2, all the above embodiments identify the basket number from the X-ray machine image and retrieve the corresponding passenger ID through the basket number, thus completing the person-bag binding. These embodiments demonstrate the automatic association between passenger information and baggage information in security check management achieved through this method, which can improve the accuracy of passenger ID acquisition, avoid data confusion, and improve traceability accuracy.

[0063] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0064] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A passenger and baggage management method based on the fusion of channel vision and RFID, characterized in that, Includes the following steps: S1. Passenger basket binding management: In the luggage placement area, the line management mode and the visual management mode are executed in parallel to obtain the line mode binding data and the visual mode binding data used to represent the relationship between passengers and luggage baskets. The line management mode includes: reading the RFID tag of the luggage basket through the first RFID reader on the luggage preparation table, triggering the face auxiliary screen to obtain the current passenger's face information, and generating line mode binding data; The visual management mode includes: continuously acquiring images of the luggage area through a camera, identifying the passenger's facial information and the visual markings on the surface of the passenger's luggage basket from the images, and generating visual mode binding data; The code stored in the RFID tag of the luggage basket is the same as the code carried by the visual tag, both of which are the basket number of the luggage basket; Both the line mode binding data and the visual mode binding data contain corresponding confidence levels, which are calculated as follows: The following weighted recombinations were used to calculate the confidence level of line mode management: first weighted recombinations were assigned to the quality assessment scores of RFID tag reading, face capture by face-assisted screen, and face identity comparison and matching under the visual management mode; second weighted recombinations were assigned to the quality assessment scores of visual tag recognition, face capture by camera, and face identity comparison and matching under the visual management mode, and the results were calculated to obtain the confidence level of visual mode management. S2. Data Fusion Management: Before the luggage basket enters the X-ray machine, the binding data of the line mode and the binding data of the visual mode are retrieved using the basket number as the main index; then, the binding data of the two modes are checked for consistency based on preset fusion rules, and the final person-basket binding data is determined according to the check results; wherein, the fusion rules include: When only one of the line body mode binding data and the visual mode binding data is obtained, the binding data set is directly used as the final human basket binding data. When the passenger identities in the two sets of bound data are consistent, either set of data is selected as the final passenger-basket binding data. When the passenger identities in the two sets of bound data are inconsistent, the one with higher confidence level is selected as the final person-basket binding data. When only one set of data is valid among the two sets of binding data, the valid data is selected as the final person-basket binding data. S3, Bag Binding Management: When a luggage basket passes through an X-ray machine, the system acquires the luggage security inspection image generated by the X-ray machine, identifies the visual markings on the surface of the luggage basket from the image, obtains the basket number, binds the luggage security inspection image with the identified basket number, and generates bag binding data. S4. Person-bag binding management: Based on the person-basket binding data and basket-bag binding data, the baggage security inspection image is associated with the corresponding passenger identity information to generate person-bag binding data between the passenger and their carry-on baggage, and stored in the security inspection management database for security management traceability.

2. The passenger and baggage management method based on the fusion of channel vision and RFID as described in claim 1, characterized in that: In step S2, the line body mode manages the confidence level. The calculation formula is: in, The score is used to evaluate the quality of RFID tag reading. The value range is [0,1]. A successful reading without interference is scored as 1, a successful reading after multiple retries is scored as 0.8, a reading due to weak signal or crosstalk causing unstable reading is scored as 0.5, and a reading failure is scored as 0. The value is the quality assessment score of the face capture by the face-assisted screen, with a range of [0,1]. A value of 1 is given for a complete and unobstructed frontal face capture, a value of 0.7 is given for a side face capture or a face with partial occlusion but which does not affect the extraction of facial features, a value of 0.3 is given for a face capture with occlusion or blur that affects the extraction of facial features, and a value of 0 is given for no valid face capture. The value is the evaluation score for face identity comparison and matching in the line body management mode. The value range is [0,1]. The value is 1 for a complete match of face identity features, 0.8 for a matching similarity of more than 50% and less than 100%, 0.5 for a matching similarity of less than 50%, and 0 for a failed match. , , The weight coefficients of the first weight group satisfy the following conditions: ,and =0.5, =0.3, =0.2; Visual pattern management confidence The calculation formula is: in, The evaluation score for the confidence level of visual identification is in the range of [0,1]. The score is 1 for a complete and clear identification of the basket number, 0.8 for a blurry basket number but accurate identification, 0.5 for a basket number that is partially obscured, worn or blurred, which makes the identification uncertain, and 0 for no identification. The score is the quality assessment value for facial capture by the camera, ranging from [0,1]. The value selection rules are the same as those for facial recognition. same; The evaluation score for facial identity matching in visual management mode, and the scoring rules are the same as those for facial recognition. same; , , The weight coefficients of the second weight group satisfy the following conditions: ,and =0.4, =0.3, =0.

3.

3. The passenger and baggage management method based on the fusion of channel vision and RFID as described in claim 2, characterized in that: In step S1, in the online body management mode, after the first RFID reader reads the RFID tag of the luggage basket, it identifies the RFID tag to obtain the basket number B, and triggers the face-capturing auxiliary screen at the corresponding preparation station to start face capture. Based on the captured face information, it calls the OneID identity service to obtain the passenger's identity information. Record the current binding time And channel number C, generate line mode binding data : Where R is the reset identifier, and at this time it is the initial state N, which indicates that the luggage basket has completed the current passenger's basket binding but has not yet completed the security check reset; Facial image feature data captured by a face-assisted screen; Confidence level for line management mode; In visual acquisition mode, images of the luggage area are continuously captured by a camera. OCR technology is used to identify visual markers from the images, obtaining the baggage number B. Passenger facial information is then extracted from the images, and the OneID identity service is invoked to retrieve passenger identity information. Record the current binding time And channel number C, generate visual pattern binding data : in, Facial image feature data captured by the camera; Confidence management for visual patterns; The update rule for the reset flag R is: using Key=(B,C, The unique index key is used to construct the data storage record for the person-basket binding, where, The system binds the person-basket to a specific time. When the baggage basket carrying the luggage enters the exit of the X-ray machine, the second RFID reader deployed on the exit side of the X-ray machine reads the RFID tag of the baggage basket. Based on the same index key, the reset tag in the corresponding person-basket binding data record is updated to R=Y, indicating that the baggage basket has completed the security check process and entered the reset state. The baggage basket can then be reused.

4. The passenger and baggage management method based on the fusion of channel vision and RFID as described in claim 1, characterized in that: In step S3, if the visual markings on the surface of the luggage basket cannot be effectively identified from the luggage security inspection image, the image anomaly fallback compensation management mechanism is triggered. The management method of this fallback compensation management mechanism is as follows: obtain the acquisition time and security channel number of the current luggage security inspection image, and find the luggage basket number under the same security channel whose RFID acquisition time is less than and closest to the luggage security inspection image acquisition time from the historical management data read by the second RFID reader deployed at the front end of the X-ray machine entrance or the back end of the exit. Use this number as the basket number corresponding to the luggage security inspection image for compensation binding management, and generate the final basket binding data.

5. The passenger and baggage management method based on the fusion of channel vision and RFID as described in claim 4, characterized in that: In step S4, both the person-basket binding data and the basket-bag binding data include the basket number, security checkpoint number, and corresponding binding time. Associating the baggage security check image with the corresponding passenger identity information includes: extracting the basket number, security checkpoint number, and basket-bag binding time from the basket-bag binding data; using the basket number and security checkpoint number in the basket-bag binding data as the main index; and filtering out valid person-basket binding data from the person-basket binding data that has the same basket number, the same security checkpoint number, and the person-basket binding time that meets the following time constraints. in, The time is bound to the basket; W is the preset time window threshold, which is the maximum time for the luggage basket to flow from the luggage preparation table to the X-ray machine; Binding time to the basket; The retrieved valid person-basket binding data is sorted in reverse order by the binding time. The sorted dataset is traversed, and the first record of the person-basket binding data is taken. The passenger identity information in the person-basket binding data is associated with the corresponding baggage security inspection image in the bag binding data.

6. The passenger and baggage management method based on the fusion of channel vision and RFID as described in claim 5, characterized in that: In step S4, the time when the passenger identity information of the first record is associated with the baggage security inspection image is set as the baseline binding time. The remaining baggage basket binding data is further traversed. If the difference between the binding time of the remaining baggage basket binding data and the baseline binding time does not exceed the baggage basket reuse time, it is determined to be valid binding data associated with the same baggage basket. The passenger identity information in it is also associated with the corresponding baggage security inspection image. If multiple different passenger identity information are bound, they are manually verified by capturing images to achieve the binding of unique passenger identity information. If the reuse time of the luggage basket is exceeded, it will be judged as invalid historical data generated by the reuse of luggage baskets and will be filtered out to avoid passenger mismatch caused by the repeated reuse of luggage baskets.

7. The passenger and baggage management method based on the fusion of channel vision and RFID as described in claim 6, characterized in that: In step S4, the reuse time of the luggage basket is dynamically predicted using a sliding time window + mean drift clustering algorithm.

8. The passenger and baggage management method based on the fusion of channel vision and RFID as described in claim 5, characterized in that: In step S4, when the corresponding passenger identity information cannot be retrieved in the passenger basket binding data, the passenger basket management data missing fallback compensation management mechanism is triggered. The management method of this fallback compensation management mechanism is as follows: based on the current binding time of the baggage basket, a set of passenger management information in the same security check channel and whose security check time is within the security check time window is selected from the passenger security check records. The passenger identity information in the passenger management information set is used as the corresponding passenger for the current baggage for fallback binding management, and a fallback management record is generated. The security check time window is the time period between the passenger's identity security check at the counter and the baggage security check in the baggage placement area.

9. A passenger and baggage management method based on the fusion of channel vision and RFID as described in claim 4, characterized in that: In steps S1 and S3, OCR recognition technology is used to identify visual identifiers. Two visual identifiers are set on the side of the luggage basket, and an arrow is set between the two visual identifiers. The direction of the arrow indicates the running direction of the luggage on the track. The code of the visual identifier is read according to the direction indicated by the arrow. The visual identifiers on both sides of the arrow are identified and parsed. If the codes identified on both sides are consistent, a unique basket number binding result is output. If the codes identified on both sides are inconsistent, two independent recognition result sets are output, which are then manually reviewed or verified by the system.

10. A passenger and baggage management system based on the fusion of channel vision and RFID, used to execute the management method according to any one of claims 4-9, characterized in that, include: The luggage baskets are equipped with RFID tags and visual tags. The RFID tags are located on the bottom of the luggage baskets, and the visual tags are located on the sides of the luggage baskets. The line management data acquisition module includes a first RFID reader and a face recognition auxiliary screen deployed at the baggage preparation station, used to execute the line management mode; The visual management acquisition module includes a camera deployed on the side of the baggage conveyor track for executing the visual management mode; The auxiliary RFID management and acquisition module includes a second RFID reader deployed on the entrance and exit sides of the X-ray machine to assist in the execution of basket binding management steps and image anomaly compensation management mechanism. The security inspection management data processing unit is connected to the line management acquisition module, the vision management acquisition module, the auxiliary RFID management acquisition module and the X-ray machine respectively. It is used to perform the steps of personnel binding management, data fusion management, basket and bag binding management, personnel and bag corresponding management and all fallback compensation management mechanisms, and generate traceable security inspection management records. The security inspection management database is used to store security inspection management records.

Images