Parking lot temporary parking arrearage management and entrance control method and system

Abstract

The invention relates to the technical field of parking lot management, and particularly discloses a parking lot temporary parking arrearage management and entrance control method and system. According to the method, a tire pattern image and a Bluetooth address of a vehicle are collected, a feature fingerprint spectrum is generated, and a binding structure is constructed; combining the image sequence to extract path information, and generating a track confidence score; executing identity verification based on the scoring result, and constructing a digital credit voucher; and then arrearage verification and freezing control are executed through a block chain mechanism, entrance judgment and barrier gate operation are completed, and identification platform parameters are updated based on an execution record and image sequence training model. The method effectively improves the recognition accuracy and model adaptability in a complex environment, solves the problems of cross-field avoidance, recognition failure, model island and the like of arrearage vehicles, and has high stability, safety and popularization value.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent transportation and parking management, and particularly to a method and system for managing overdue temporary parking fees and controlling vehicle entry in a parking lot. Background Art

[0002] With the continuous construction of the urban intelligent transportation system, parking lot management systems have been widely used in reducing labor costs and improving vehicle passing efficiency. However, existing temporary parking management systems generally rely on a single license plate recognition technology. When a vehicle is in complex environments such as strong light reflection, rain and fog occlusion, and sunshade film coverage, problems such as misrecognition and tracking interruption are likely to occur, seriously affecting the accurate tracking of overdue vehicles and subsequent entry interception control.

[0003] On the other hand, to avoid payment management, many overdue vehicles bypass the original records by means such as changing license plates or switching parking areas across different lots, resulting in ineffective supervision. At the same time, the traditional manual verification process for supplementary payment vouchers is inefficient and easily forged, posing a significant potential credit risk. Although some systems introduce Bluetooth signal-assisted recognition or increase the dimension of image matching, there are still obvious deficiencies in data fusion, matching determination, and model generalization ability. Summary of the Invention

[0004] The present invention provides a method and system for managing overdue temporary parking fees and controlling vehicle entry in a parking lot, so as to solve the problem of how to construct a path matching graph structure based on the Bluetooth address, tire tread image, and image sequence during the vehicle entry process, and through a trajectory confidence scoring model and a cross-modal federated training mechanism, achieve accurate vehicle identity recognition and dynamic update of the recognition model in a strong interference environment, thereby improving the tracking ability of overdue temporary parking vehicles in the parking lot and the adaptive performance of entry control.

[0005] To solve the above technical problems, the present invention provides a method for managing overdue temporary parking fees and controlling vehicle entry in a parking lot, including:

[0006] Obtaining the tire tread image and Bluetooth address of the entering vehicle, generating a feature fingerprint map, and constructing a feature binding structure;

[0007] Obtaining the feature binding structure and image sequence, extracting path information, and constructing a path matching graph structure to generate a trajectory confidence scoring result;

[0008] Obtaining the trajectory confidence scoring result and performing identity verification to generate a digital identity structure and constructing a digital credit voucher;

[0009] Obtaining the digital credit voucher and binding structure information, performing overdue verification, and generating an entry status instruction to complete the entry control judgment;

[0010] Obtain the entrance status instruction and perform the barrier operation, generate an execution record and complete the control response operation;

[0011] Obtain the execution record and the image sequence, complete the local model training and update the recognition model platform parameters.

[0012] Furthermore, obtain the tire tread pattern image and the Bluetooth address of the entering vehicle, generate a feature fingerprint map and construct a feature binding structure, including:

[0013] Obtain the tire tread pattern image and the in-vehicle Bluetooth address of the entering vehicle, perform image filtering and data cleaning to obtain the tire tread pattern image and the Bluetooth address data;

[0014] Extract the contour edge features and the address label information from the tire tread pattern image and the Bluetooth address data to generate a feature fingerprint map;

[0015] Furthermore, obtain the feature binding structure and the image sequence, extract the path information and construct a path matching graph structure, including:

[0016] Obtain the feature binding structure and the image sequence of the entering vehicle, perform object detection and temporal tracking processing to obtain the image tracking path;

[0017] Extract the path point sequence and the matching status information from the image tracking path to construct a path matching graph structure.

[0018] Furthermore, perform a comprehensive scoring process on the path matching graph structure to generate a trajectory confidence scoring result.

[0019] Furthermore, obtain the trajectory confidence scoring result and perform identity verification, generate a digital identity structure and construct a digital credit voucher, including:

[0020] Obtain the trajectory confidence scoring result, combine the historical binding records, and perform the entrance identity verification;

[0021] Digitally mark the identity verification result to generate a digital identity structure;

[0022] Construct a credit voucher based on the digital identity structure and complete the digital signature to construct a digital credit voucher.

[0023] Furthermore, obtain the digital credit voucher and the binding structure information, perform the arrears verification and generate the entrance status instruction to complete the entrance control judgment, including:

[0024] Obtain the digital credit voucher and the binding structure information, match the historical arrears records and complete the verification;

[0025] Perform a freeze control judgment on the verification result to generate the entrance status instruction;

[0026] Update the barrier control logic based on the entry status instruction to complete the entry control judgment.

[0027] Further, obtain the entry status instruction and perform the barrier operation, generate an execution record and complete the control response operation, including:

[0028] Obtain the entry status instruction and call the control interface to complete the physical barrier execution operation;

[0029] Combine and process the physical barrier execution operation result with the identity verification record to generate an execution record;

[0030] Submit the execution record to the entry behavior storage module to complete the control response operation.

[0031] Further, obtain the execution record and the image sequence, and perform multi-modal data screening and structure registration processing.

[0032] Further, perform local model training based on the multi-modal data screening result, generate incremental model parameters, and synchronize the incremental model parameters to the recognition model platform to complete parameter update.

[0033] A temporary parking arrears management and entry control system for a parking lot, which is applied to the temporary parking arrears management and entry control method described in any one of the above, and is characterized by including:

[0034] A feature acquisition module, which is used to obtain the tire tread pattern image and Bluetooth address of the entry vehicle, generate a feature fingerprint map and construct a feature binding structure;

[0035] A trajectory recognition module, which is used to obtain the feature binding structure and the image sequence, construct a path matching graph structure and generate a trajectory confidence score result;

[0036] An identity verification module, which is used to obtain the trajectory confidence score result and generate a digital identity structure and a digital credit voucher;

[0037] An entry control module, which is used to obtain the digital credit voucher and the binding structure information, perform arrears verification and generate an entry status instruction;

[0038] An execution control module, which is used to obtain the entry status instruction and complete the barrier operation and control response;

[0039] A model training module, which is used to obtain the execution record and the image sequence and update the parameters of the recognition model platform.

[0040] The key innovation points of the present invention include:

[0041] (1) Innovatively propose to bind the 3D tire pattern image with the in-vehicle Bluetooth MAC address to form a biometric digital feature pair, constituting the unique identifier source of the vehicle, replacing the traditional single license plate recognition method, and providing a stable identity anchor point for trajectory construction.

[0042] (2) Introduce the edge weight calculation formula into the design diagram structure, and integrate visual similarity, Bluetooth matching degree, and geomagnetic deviation degree in the trajectory scoring. At the same time, combine Shannon entropy as the confidence weighting term to construct a generalizable and interpretable trajectory scoring system.

[0043] (3) Innovatively adopt zero-knowledge proof to generate the digital credit voucher structure, realize the arrears verification and contract execution under the condition of no exposure, and support the entry credit guarantee system of "mortgage first, then pass".

[0044] (4) Introduce triple optimization terms of registration error minimization, KL divergence constraint, and trajectory derivative regularization in the local model training stage, and adopt the weight synchronization method to construct a federated recognition model update path suitable for multiple scenarios.

[0045] The following are its main beneficial effects:

[0046] (1) In the present invention, the tire pattern image and the Bluetooth MAC address are synchronously collected when the vehicle enters the venue for the first time, and a unique feature binding structure is constructed. As the dual mapping of the physical and digital identities of the vehicle, this structure can still compensate for the recognition error through the Bluetooth address in the scenarios of strong light interference, rain and fog occlusion, or sunshade film interference, effectively reducing the matching failure rate caused by license plate recognition errors.

[0047] (2) The trajectory matching graph structure constructed by the system, and a joint scoring model of visual features, Bluetooth signals, and geomagnetic data is introduced to realize the dynamic scoring between path points and the elimination of abnormal points. This scoring mechanism solves the local inconsistency problem of traditional frame-level recognition, and improves the accuracy of trajectory continuity determination.

[0048] (3) The present invention introduces the digital credit voucher structure, completes the pre-authorization mortgage for entry through the blockchain contract mechanism, and automatically triggers the freezing and withholding agreement after arrears verification. Compared with the static matching mode relying on the local database, this mechanism supports cross-parking lot identity sharing and credit status synchronization, effectively preventing "escaping after changing the parking lot" of arrears vehicles.

[0049] (4) By designing the cross-modal structure registration function and the trajectory smoothing regularization term, the present invention constructs a multi-modal collaborative local model training framework, and uses federated learning to dynamically update the recognition model. This mechanism can quickly adapt to the recognition accuracy differences in different parking lot environments, and improve the generalization ability and stability of the overall platform model. Description of the Drawings

[0050] Figure 1Schematic flowchart of a method for managing overdue parking fees and controlling access in a parking lot provided by an embodiment of the present application;

[0051] Figure 2 Block diagram of a system for managing overdue parking fees and controlling access in a parking lot provided by an embodiment of the present application. Detailed implementation manners

[0052] Example 1: Refer to Figure 1 , which is a schematic flowchart of a method for managing overdue parking fees and controlling access in a parking lot provided by an embodiment of the present invention. The process may at least include steps S100 - S600:

[0053] S100. Obtain the tire tread pattern image and Bluetooth address of the incoming vehicle, generate a characteristic fingerprint map, and construct a characteristic binding structure;

[0054] S200. Obtain the characteristic binding structure and image sequence, extract path information, construct a path matching graph structure, and generate a trajectory confidence score result;

[0055] S300. Obtain the trajectory confidence score result and perform identity verification, generate a digital identity structure, and construct a digital credit voucher;

[0056] S400. Obtain the digital credit voucher and binding structure information, perform overdue fee verification, and generate an access status instruction to complete the access control judgment;

[0057] S500. Obtain the access status instruction and perform the barrier gate operation, generate an execution record, and complete the control response operation;

[0058] S600. Obtain the execution record and image sequence, complete local model training, and update the recognition model platform parameters.

[0059] Step S100 at least includes steps S110 - S130:

[0060] S110. Obtain the tire tread pattern image and in - vehicle Bluetooth address of the incoming vehicle, perform image filtering and data cleaning to obtain the tire tread pattern image and Bluetooth address data.

[0061] Specifically, when the incoming vehicle travels to the image acquisition area, the laser three - dimensional structure scanning device deployed under the ground of the vehicle passage starts the tire tread pattern image acquisition process. The laser scanning device uses a vertical multi - line array laser beam irradiation method to scan and model the tire surface to obtain the tire tread pattern image. During the scanning process, the image acquisition unit constructs a high - resolution structure image based on the point cloud depth data and the calculation result of the reflection gray - scale intensity, and performs noise removal processing through an image filtering algorithm after the image acquisition is completed.

[0062] Meanwhile, during the vehicle's driving process, it enters the detection area composed of lateral Bluetooth signal detectors. The system activates the low-power Bluetooth host mode and broadcasts a specific handshake signal packet. If the in-vehicle Bluetooth device is in the "discoverable" state, the system will receive the response broadcast frame and extract the Bluetooth address field as the Bluetooth address data. If the in-vehicle Bluetooth device is turned off or not discoverable, and the system does not receive a valid response within the specified time window, it automatically marks the Bluetooth channel recognition status as "no response" and enters the tire tread image enhancement path.

[0063] To further improve the robustness of image recognition, the system enhances the tire image acquisition frequency in this path, expands the scanning area to the front and rear wheels, and simultaneously transmits the image data to the backend contour analysis module to enhance the edge discrimination ability.

[0064] Finally, the system obtains the tire tread image and the corresponding Bluetooth address data, where the Bluetooth address data can be the measured value or a null value. The system performs format conversion and feature area cropping processing on the image data, and at the same time performs deduplication, channel strength screening, and protocol format verification operations on the Bluetooth address data to obtain the tire tread image and Bluetooth address data for the next step of processing.

[0065] S120. Extract the contour edge features and address label information from the tire tread image and Bluetooth address data to generate a feature fingerprint map.

[0066] Based on the tire tread image and Bluetooth address data obtained in step S110, the system performs multi-source information extraction and fusion processing. Specifically, for the tire tread image part, the edge detection module is called to perform gradient analysis on the image contour area, extract feature factors such as the pattern groove structure, block surface tangent distribution, and groove density change, and construct a description vector set.

[0067] After the above vector set is processed by the multi-scale direction normalization strategy, it forms contour edge features with geometric uniqueness; for the Bluetooth address data part, if there is valid broadcast information, the system extracts the main Bluetooth address field and the identification time field to form an address label information set. If the Bluetooth address is missing, the system automatically marks the current binding structure as "single-source binding" and supplements the position mark of the image strong feature area.

[0068] Subsequently, the system splices the contour edge features and address label information in structure and rearranges the format to generate a feature fingerprint map. The feature fingerprint map includes, but is not limited to, fields such as the contour direction vector matrix, edge patch weight index, Bluetooth address label, and device signal identifier. This map will be used as the only data source for constructing the subsequent feature binding structure and supports the multi-frame tire image compensation mode in the scenario of missing Bluetooth addresses.

[0069] S130. Encode and store the characteristic fingerprint map to construct a characteristic binding structure.

[0070] After the characteristic fingerprint map is generated in step S120, the system enters the characteristic encoding and structured binding stage. Specifically, the system performs field mapping, value range compression, block reorganization, and identity field specification processing on the fingerprint map based on the data structure specification to generate a structured encoding data packet. The multi-directional gradient principal component features of the tire image, the Bluetooth tag address (if any), and the vehicle entry timestamp are retained in the data packet.

[0071] During the process of constructing the binding structure, the system distinguishes the generation mode according to whether Bluetooth address data is included. If the Bluetooth address field exists, the two-way binding mode is executed to form a one-to-one mapping structure between the image feature encoding and the Bluetooth address; if the Bluetooth address is missing, the image enhancement binding mode is executed to incorporate the frame numbers of multiple tire tread images into the binding structure to enhance the robustness of target recognition.

[0072] After the binding structure is written into the local cache database, a characteristic binding structure is generated, and the structure identifier is synchronously sent to the downstream path recognition module. This characteristic binding structure will serve as the primary reference information in the image sequence matching and path composition process of step S200 and participate in the matching verification process of image target tracking.

[0073] By implementing steps S110 to S130 of the present invention, when the in-vehicle Bluetooth is turned off or in an undiscoverable state, the system can still achieve the complete generation of the characteristic binding structure by constructing an image reinforcement mechanism and a structure discrimination strategy, ensuring the consistency of target recognition and the accuracy of path tracking in complex scenarios. The characteristic binding structure forms the core basic data in subsequent steps such as path composition, identity verification, and arrears blocking, enhancing the recognition reliability of the overall system.

[0074] Step S200 at least includes steps S210 - S230:

[0075] S210. Obtain the characteristic binding structure and the image sequence of the entering vehicle, perform target detection and temporal tracking processing, and obtain the image tracking path.

[0076] Specifically, the system first obtains the characteristic binding structure generated in S130, which includes the edge feature vector of the tire tread contour of the vehicle and the Bluetooth address tag information. Subsequently, the system obtains the image sequence generated by the image acquisition device deployed on the entry channel, performs target detection on each frame of the image using the YOLOv7 model, identifies the vehicle bounding box and its center position, and extracts the vehicle appearance feature vector. To achieve the temporal continuity recognition of the target between image frames, the system constructs the following prediction update formula:

[0077]

[0078] Wherein:

[0079] P t : The predicted position state of the vehicle at time t;

[0080] F t : The appearance feature vector extracted by YOLOv7 in image frame t;

[0081] The gradient direction vector of F t in the image space;

[0082] M t : The matching feature vector of the vehicle Bluetooth device;

[0083] The Bluetooth matching change rate;

[0084] S t : The continuous guiding displacement vector of the geomagnetic path;

[0085] α1, α2, α3: Empirical weight coefficients, satisfying the normalization constraint.

[0086] This formula performs weighted derivation on three types of feature sources (visual features, Bluetooth status, geomagnetic path) to construct a continuous prediction model for the image tracking path, which is used to maintain path stability in low-light or occlusion scenarios.

[0087] S220. Extract the path point sequence and matching status information from the image tracking path to construct a path matching graph structure.

[0088] After obtaining the tracking path state sequence {P t}, the system sequentially establishes a path node set, and each node corresponds to the vehicle position state of one frame of the image. The system constructs the edge weights of the graph structure according to the spatial change and similarity between the image object detection frames. The edge weights are in the following form:

[0089] w i,i+1 = ‖P i - P i+1 ‖2 + λ·(1 - IoU(B i , B i+1 )) (Formula ②)

[0090] Wherein:

[0091] w i,i+1 : The edge weight between the i-th and (i + 1)-th frames in the graph structure;

[0092] ‖P i - P i+1 ‖2: The Euclidean distance between position states;

[0093] Bi : Detection box position of the i-th frame;

[0094] IoU(B i ,B i+1 ): Intersection over Union of two detection boxes;

[0095] λ: IoU difference penalty coefficient.

[0096] By constructing a graph structure, the system generates the matching paths and their weight distributions between consecutive frames, preparing the input structure for path consistency discrimination and scoring.

[0097] S230. Perform comprehensive scoring processing on the path matching graph structure to generate a trajectory confidence scoring result.

[0098] After the graph structure is established, the system traverses and evaluates the paths. By introducing a multi-modal matching degree scoring function, an information entropy adjustment term, and a confidence weighting mechanism, the following trajectory confidence scoring function is formed:

[0099]

[0100] Where:

[0101] Ψ track : Trajectory confidence scoring result;

[0102] φ img (P i ): Similarity function between image features and fingerprint maps;

[0103] φ bt (M i ): Bluetooth address matching function, measuring the matching degree between trajectory nodes and bound Bluetooth;

[0104] φ geo (S i ): Geomagnetic sensing path trajectory matching function;

[0105] p i : Matching uncertainty probability of path node i;

[0106] Shannon entropy term, guiding the system to attach importance to stable trajectory points;

[0107] Z: Normalization factor, ensuring that the final score is in a comparable range;

[0108] β1, β2, β3: Weighting factors, reflecting the importance of three feature sources, satisfying β1 + β2 + β3 = 1.

[0109] The final Ψ output by the system trackIt will be used as an input variable for the entry identity determination logic in S300 and is used to determine whether the current vehicle is a trusted bound object.

[0110] Step S300 at least includes steps S310 - S330:

[0111] S310. Obtain the trajectory confidence score result, and combine it with the historical binding record to perform entry identity verification.

[0112] Specifically, the system first obtains the trajectory confidence score result from step S230. This score result is generated by comprehensively comparing the image tracking path with the tire tread pattern feature, Bluetooth address feature, and geomagnetic path feature in the feature binding structure. Based on the trajectory confidence score result, the system calls the binding structure indexing module to retrieve the set of vehicle entry records containing matching tags from the historical binding database.

[0113] The system compares the current score result with the historical binding record according to the trajectory point mapping relationship and the matching degree of the image feature descriptor. The comparison parameters include feature indicators such as the difference in contour wheelbase, the residual of the tread direction tensor, the Bluetooth address hash consistency field, and the trajectory of channel strength change.

[0114] During the comparison process, the system sets a dynamic matching threshold and determines whether the current trajectory path belongs to the identity of the historically bound vehicle through the identity cross - verification module. If the trajectory confidence score is higher than the set threshold, the system marks the current vehicle identity status as "trusted binding"; if it is lower than the threshold, it is marked as "unregistered or identity ambiguous".

[0115] This identity determination mark serves as an important input parameter for generating the digital identity structure in subsequent step S320, and at the same time will be associated with the feature binding structure through field reference to maintain data structure consistency.

[0116] S320. Digitally mark the identity verification result to generate a digital identity structure.

[0117] Furthermore, after the system obtains the identity verification result in step S310, it starts the digital identity structure generation module. This module structurally integrates the identity verification status, confidence score value, feature binding structure number, and the current entry timestamp to generate a verifiable digital identity structure.

[0118] The digital identity structure includes the following fields: identity status identification field, confidence score field, binding structure index field, image feature summary field, and Bluetooth address summary field, and is attached with a location awareness tag and a time synchronization tag to ensure the uniqueness of this structure in the multi - field recognition platform.

[0119] The system calls the hash digest algorithm to perform summary compression on the image descriptor and the Bluetooth MAC address field, avoiding data leakage of the original image and MAC value during cross-platform transmission. Write the above data fields into the identity structure template, and perform encoding and compression processing on the field content to generate a standardized digital identity structure data packet.

[0120] This digital identity structure will serve as the input carrier for constructing digital credit vouchers, and link its status tag with the key binding mechanism as the basic condition for subsequent credit freezing and contract signing.

[0121] S330. Construct a credit voucher based on the digital identity structure and complete a digital signature to construct a digital credit voucher.

[0122] After the digital identity structure is generated, the system calls the credit voucher generation module to seal and package this structure and construct a digital credit voucher. First, the system calls the identity status field and the confidence score field in the digital identity structure to determine the credit status classification of the current vehicle. Combining the credit score mapping rules set by the platform, convert the confidence score into a digital credit value and generate a corresponding credit voucher level label.

[0123] Subsequently, the system calls the elliptic curve digital signature algorithm according to the key system of the platform end node to perform an immutable digital signature process on the digital identity structure. This process includes identity field digest generation, signature private key encryption, and trusted node public key verification field embedding.

[0124] The system combines the signature result with the digital identity structure data structure to generate a complete digital credit voucher. This credit voucher includes the following core fields: voucher encoding, identity structure hash, signature information block, credit value label, timestamp label, node verification public key information block, etc.

[0125] Finally, the digital credit voucher is synchronously written into the local contract pre-authorization queue and sent to the arrears verification module in step S400 as the primary key input for the frozen contract trigger condition.

[0126] Through the S300 module of the present invention, the system can accurately integrate the trajectory recognition result with the historical binding information, construct a structured, traceable, and encrypted-signed digital identity structure, and then form a digital credit voucher with on-chain application capabilities, realizing the tight coupling of the vehicle credit status and the access control logic, providing legal voucher support for subsequent cross-site anti-fare-evasion recognition and automatic deduction mechanisms.

[0127] Step S400 at least includes steps S410 - S430:

[0128] S410. Obtain the digital credit voucher and the binding structure information, match the historical arrears records and complete the verification.

[0129] Specifically, the system first obtains a digital credit voucher from step S330. The voucher includes an identity structure hash field, a voucher level label, a signature information block, and a node public key identification field. The system synchronously calls the feature binding database and extracts the vehicle unique identification item from the binding structure constructed in step S130. The identification item is composed of a tire tread pattern code, a Bluetooth address label, and an image matching index, and has the ability to identify across time and space.

[0130] The system uses the vehicle unique identification item as the matching primary key to access the historical overdue record database and perform the following comparison operations: First, match the historical order binding structure number through the structured index field; then compare the digital signature traceability information in the historical record according to the credit voucher hash field; finally, call the on-chain bill contract to verify whether there are any outstanding overdue contract entries for the vehicle.

[0131] After completing the above matching, the system generates a verification result structure, which includes the following fields: a matching status identification field, an overdue contract identification number, a bill status field, a freeze suggestion level field, and an authentication time label field. This structure will be used as an input variable for judging the freezing condition in S420.

[0132] S420. Perform a freezing control judgment on the verification result to generate an admission status instruction.

[0133] After obtaining the verification result structure generated in step S410, the system calls the freezing judgment engine to jointly compare according to the freeze suggestion level field and the current vehicle digital credit voucher level field to determine whether the freeze contract trigger condition is met.

[0134] The system has a built-in freezing control rule set, and the rule basis includes but is not limited to the following dimensions: 1) The cumulative number of overdue times of the vehicle in the past n days; 2) The amount threshold of a single unpaid bill; 3) The time ratio of the overdue duration to the interval since the last payment; 4) Whether there are multiple skip behaviors in multiple parking lots.

[0135] When the freezing condition is met, the system sets the freeze instruction flag field to "freeze execution", sets the barrier status instruction to "deny entry" at the same time, and attaches a freeze reason code and a credit punishment suggestion level. When the freezing condition is not met, the system sets the instruction field to "allow entry" and attaches a verification passed flag.

[0136] The admission status instruction is finally output in the form of a structure, including a status control code, an instruction generation timestamp, a processing node number, and a voucher primary key reference field. This structure will be used as an input object for updating the control logic in S430 to achieve the linkage trigger between the status and the behavior.

[0137] S430. Update the barrier control logic based on the entry status instruction to complete the entry control judgment.

[0138] Further, the system calls the barrier control module according to the entry status instruction generated in step S420, and updates the parameter logic of the barrier execution state machine based on the status control code. Specifically, the system injects the status control code into the barrier logic framework and replaces the response trigger parameters of the current status node.

[0139] If the entry status instruction is marked as "allow entry", the system automatically enables the passage path, releases the barrier execution permission, sets the vehicle as a "trusted identification object" at the same time, and pushes its credit certificate and identity structure to the next-hop blockchain node for chained deposit processing.

[0140] If the entry status instruction is marked as "deny entry", the system calls the rejection logic processing flow, outputs the freezing reason code to the vehicle prompt terminal, attaches the freezing record to the blockchain arrears contract record, and triggers the background alarm module to push relevant notifications at the same time.

[0141] Finally, after the system completes the logic update, it outputs an entry control identifier to the physical execution module described in step S500 for issuing the physical action execution instruction of the barrier. The whole step realizes the closed-loop linkage process from credit certificate → behavior control → execution response.

[0142] The present invention significantly improves the arrears determination efficiency before vehicle entry by constructing an automatic verification and freezing control mechanism with digital credit certificates as the core, and avoids the timeliness and forgery risks existing in traditional manual audits. At the same time, based on the dual-factor verification strategy of the binding structure and the bill chain, it ensures the accuracy and auditability of the freezing logic, effectively prevents the situation of arrears vehicles avoiding responsibilities through jumping fields, and provides refined management capabilities for urban-level parking control.

[0143] Step 500 at least includes steps S510 - S530:

[0144] S510. Obtain the entry status instruction and call the control interface to complete the physical barrier execution operation.

[0145] Specifically, obtain the entry status instruction, where the entry status instruction is the structured entry control parameter output after the arrears verification is completed based on the digital credit certificate and the binding structure information in the S400 module. The entry status instruction includes field contents such as instruction number, execution permission, trust level mark, and barrier control signal.

[0146] Further, call the control interface to complete the physical barrier execution operation, where the control interface is a predefined standard embedded communication protocol interface, supporting physical signal docking with multi-brand barrier controllers and compatible with both level trigger and command signal encoding mechanisms.

[0147] Understandably, during the execution of the physical barrier operation, it is also necessary to synchronously link and judge the current vehicle passage state, on-site geomagnetic detector state, and entrance video surveillance frame to ensure that the sending of control signals will not cause safety conflicts, especially automatically delaying execution in scenarios where there are still residual vehicles in front of the barrier or there are obstacles on-site.

[0148] S520. Combine and process the physical barrier execution operation result with the identity verification record to generate an execution record.

[0149] Specifically, obtain the physical barrier execution operation result, where the operation result includes information such as whether the barrier has successfully risen or fallen, command response time, feedback signal strength, execution exception flag, and its corresponding error code.

[0150] Meanwhile, obtain the identity verification record generated in the S300 module, where the identity verification record includes key fields such as trajectory confidence scoring result, bound structure identification number, historical matching record summary, and digital identity structure.

[0151] Combine and process the above two types of data through a field mapping structure to generate an execution record, where the execution record is a structure containing multi-source information fusion, unified with the timestamp and vehicle unique binding structure number as the main index key, and nested with sub-fields such as control behavior, response feedback, identity identification, and confidence information.

[0152] Further, the execution record also needs to mark control identifiers such as whether the current operation is an abnormal interruption, whether it is manual review intervention, and whether it is a test verification mode, etc., for distinguishing the actual business scenario from the training sample mode.

[0153] S530. Submit the execution record to the entrance behavior storage module to complete the control response operation.

[0154] Specifically, submit the execution record to the entrance behavior storage module in the form of a structured data packet, where the storage module is a dedicated sub-node in a distributed behavior log system, constructed based on an event sourcing architecture, supporting high-concurrency writing, traceability, and chained data compression.

[0155] Furthermore, during the submission process, a hash signature mechanism is invoked to generate a digest of the execution record and perform integrity verification, avoiding the risk of tampering during transmission. Additionally, by invoking the historical record interface, the execution record is bound to the binding structure information of the current vehicle, achieving vertical association of multiple rounds of entry behaviors.

[0156] In addition, the submission process also triggers a response completion signal within the system and, through a unified asynchronous event notification mechanism, pushes a data update signal to the model training scheduler of the S600 module, serving as a trigger for model parameter updates.

[0157] Through the execution of the S500 module, a closed-loop operation of the physical control response link based on multiple verification results can be achieved, ensuring that the barrier control has the capabilities of complete identity binding and behavior recording, providing reliable data support for subsequent model iteration and behavior tracking, and significantly enhancing the controllability and security prevention capabilities of the system in dynamic and complex scenarios.

[0158] Step S600 includes at least steps S610 - S630:

[0159] S610. Obtain the execution record and image sequence, and perform multi-modal data screening and structure registration processing.

[0160] Specifically, in step S610, the system first extracts the image sequence index segment corresponding to a specific time period based on the execution record obtained from step S530, and filters out the key frame set containing the key nodes of vehicle entry. On this basis, a multi-modal feature structure alignment function is invoked to perform the structure registration of the millimeter-wave radar image and the lane geomagnetic identifier.

[0161] The structure registration error is defined in the form of the least squares distance loss as follows:

[0162] ① Multi-modal structure registration loss function:

[0163]

[0164] Where:

[0165] F align (θ): Structure registration loss function;

[0166] θ: Current set of structure registration rotation and translation parameters;

[0167] R(i,θ): The i-th point of the radar modality after attitude adjustment;

[0168] M(i): Point cloud reference marker in the i-th visual modality;

[0169] n2: Total number of point clouds participating in the registration calculation.

[0170] The minimization objective of the registration loss function is used to construct the cross-modal structure fusion output, providing input feature support for the next local model training.

[0171] S620. Perform local model training based on the multi-modal data screening results to generate incremental model parameters.

[0172] In step S620, the system extracts the vehicle contour change sequence in the image modality, the reflection intensity trajectory sequence in the millimeter-wave modality, and the driving track sequence in the geomagnetic modality respectively based on the fusion feature sequence after registration in S610. For the above data, a local model training objective function with spatial continuity constraints is constructed.

[0173] ② Incremental training loss function:

[0174]

[0175] Where:

[0176] t3: Sequence frame time index;

[0177] T2: Total number of sample training frames;

[0178] P vis (t3): Vehicle contour probability distribution in the image modality;

[0179] P rad (t3): Vehicle reflection feature distribution in the radar modality;

[0180] P geo (t3): Trajectory position distribution in the geomagnetic modality;

[0181] λ: Weight factor of the trajectory smoothing term;

[0182] KL(·,·): Kullback–Leibler divergence, used to measure the difference degree between two distributions.

[0183] The loss function reflects the coupling relationship between multi-modal distribution consistency and trajectory continuity. Further, based on the above loss function, a parameter update gradient expression is constructed:

[0184] ③ Incremental model parameter update formula:

[0185]

[0186] Where:

[0187] ΔW local : Local model incremental weight;

[0188] η: Learning rate;

[0189] W(t): The set of weight parameters of the model at the t-th moment.

[0190] The local model weight gradient is used to generate a new set of parameters adapted to the local scenario for reference by the next federated aggregation module.

[0191] S630: Synchronize the incremental model parameters to the recognition model platform to complete the parameter update.

[0192] Specifically, the local model incremental parameter ΔW generated in step S620 local Will be used as input, synchronized to the system recognition model platform, and perform the following federated aggregation operations:

[0193] ④ Federated weight update formula:

[0194]

[0195] Where:

[0196] The updated global model parameters;

[0197] The previous round of federated model parameters;

[0198] ΔW local,j : The local parameter increment uploaded by the j-th node;

[0199] N3: The number of edge nodes participating in synchronization.

[0200] This aggregation strategy combines Federated Averaging (FedAvg) with the multi-modal recognition model structure optimized for the scenario of the present invention to improve the global recognition robustness and anti-interference performance in the scenario of cross-parking lot model collaborative optimization.

[0201] By introducing cross-modal structure registration function, multi-modal distribution consistency function, trajectory derivative smoothing regularization term and federated synchronization weighting mechanism, the present invention constructs a model update path with spatio-temporal consistency and data fidelity in module S600. Compared with the traditional single-field recognition error-based parameter adjustment strategy, this method significantly improves the generalization ability of model parameters and cross-field adaptability, and is particularly suitable for the problem of model collaborative optimization in multi-source access control environments.

[0202] Embodiment 2: Figure 2 Shows a structural block diagram of a parking lot temporary parking overdue fee management and access control system according to an embodiment of the present invention. As Figure 2 Shown, this structure may include:

[0203] The feature acquisition module 10 is used to perform the combined acquisition operation of biometric features and digital features when the vehicle first enters the parking lot. This module includes a laser scanning sub-module and a Bluetooth listening sub-module. Specifically, the laser scanning sub-module is used to obtain the three-dimensional pattern structure image of the vehicle tire based on the laser point cloud technology, and extract the contour edge and geometric topology information; the Bluetooth listening sub-module captures the MAC address broadcast by the in-vehicle Bluetooth module through passive listening and performs uniqueness verification. The above image data and Bluetooth address data will be synchronously transmitted to the processing end to complete feature filtering, contour calibration and feature fingerprint map generation, and finally form a feature binding structure for subsequent identity recognition.

[0204] The trajectory recognition module 20 is used to perform spatio-temporal path recognition processing based on the fusion of vision and geomagnetic information after the vehicle enters the venue. This module receives the image sequence and the feature binding structure information, performs vehicle target detection based on the YOLOv7 algorithm, and uses an improved DeepSORT architecture combined with geomagnetic perception data for time-series trajectory tracking processing, extracts the path point sequence and timestamp label information, and generates a path matching map structure. Subsequently, combined with the Doppler vector data provided by the millimeter-wave radar, path confidence modeling is performed, and a trajectory confidence score result is output for subsequent identity verification processing.

[0205] The identity verification module 30 is used to complete dynamic identity recognition and the generation of digital credit vouchers according to the trajectory confidence score result. This module receives the score result and the historical binding records in the local database, and performs consistency verification on the matching of tire features and Bluetooth addresses; if the verification passes, the identity information is marked with a hash code, a digital identity structure is constructed, and the generation of credit vouchers and digital signatures are completed based on the preset public-private key pair. The credit voucher will be used as a prerequisite for contract invocation to complete admission authorization and freeze judgment during the admission control stage.

[0206] The admission control module 40 is used to perform overdue status verification and admission judgment based on the digital credit voucher and the binding structure information before the vehicle is ready to enter the venue. The module is built-in with a blockchain query engine and a freeze rule engine. The former is used to match overdue behavior records across the parking lot environment, and the latter is used to call the freeze function according to the verification result to lock the voucher status. If the overdue record is established, the system automatically triggers the withholding contract process and generates a "reject admission" instruction; if no abnormality is found, an "admit admission" instruction is output and passed to the subsequent control module.

[0207] The execution control module 50 is used to complete the interaction operation with the physical barrier gate according to the entry status instruction and form an execution record. This module calls the control interface API to send the entry instruction to the barrier gate controller, and at the same time monitors the feedback of the execution status. If the barrier gate operation is successfully executed, the result will be merged with the identity verification record to form a complete execution record; this execution record will be passed as a feedback result to the entry behavior storage module for subsequent use by the model training module.

[0208] The model training module 60 is used to build a local incremental learning model and complete the update of the recognition model platform parameters based on the execution record and the image sequence data. This module performs multi-modal data screening and structure registration, fuses image frame data, geomagnetic data and execution labels, and extracts effective samples. Subsequently, the local training process is completed using the federated learning structure to generate incremental model parameters. Finally, the incremental parameters will be synchronized to the recognition model platform to complete the global model update, so as to continuously optimize the feature recognition performance in various scenarios.

[0209] Through the above modular structure design, the present invention achieves the following beneficial effects:

[0210] (1) The recognition accuracy is significantly improved. By introducing the two-way binding mechanism of the tire laser image and the Bluetooth MAC address, compared with the traditional license plate recognition-based scheme, the recognition accuracy is improved in environments such as low light, rain and fog occlusion, effectively solving the problem of OCR image failure.

[0211] (2) The behavior of avoiding arrears is curbed. Through the credit anchoring mechanism and the blockchain certificate synchronization design, the historical arrears information is traceable among multiple parking lots, eliminating the risk of avoiding arrears by changing the place or license plate.

[0212] (3) The ability to prevent identity forgery is enhanced. This system relies on the non-clonable physical fingerprint information (tire pattern and Bluetooth address), combined with the digital signature and hash anti-tampering mechanism, to achieve the unique binding of vehicle identity, effectively resisting MAC spoofing and license plate replacement attacks.

[0213] (4) The model can adaptively evolve. By building a multi-source data federated learning platform, the model collaborative training is realized without transmitting the original data, continuously improving the robustness and generalization ability of the model deployed across regions, and supporting the dynamic deployment of a large number of scenarios.

[0214] (5) The system closed-loop feedback chain is complete. From feature collection to entry execution and then to model update, it constitutes a complete closed loop of data collection - recognition modeling - control execution - self-optimization, effectively ensuring the real-time performance and security of the system response.

[0215] Obviously, the embodiments described above are only a part of the embodiments of this application, rather than all of them. The preferred embodiments of this application are shown in the accompanying drawings, but they do not limit the patent scope of this application. This application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosed content of this application more thorough and comprehensive. Although this application has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions described in the foregoing specific embodiments, or perform equivalent replacements on some of the technical features. Any equivalent structure made by using the content of the specification and drawings of this application, directly or indirectly applied in other related technical fields, is equally within the scope of patent protection of this application.

Claims

1. A method for managing overdue temporary parking fees and controlling access to a parking lot, characterized in that Including: Obtain the tire tread pattern image and Bluetooth address of the entering vehicle, generate a characteristic fingerprint map, and construct a characteristic binding structure; Obtain the characteristic binding structure and image sequence, extract path information, construct a path matching graph structure, and generate a trajectory confidence scoring result; Obtain the trajectory confidence scoring result and perform identity verification, generate a digital identity structure, and construct a digital credit voucher; Obtain the digital credit voucher and binding structure information, perform overdue payment verification, and generate an entrance status instruction to complete the entrance control judgment; Obtain the entrance status instruction and perform the barrier gate operation, generate an execution record, and complete the control response operation; Obtain the execution record and image sequence, complete local model training, and update the recognition model platform parameters.

2. The method according to claim 1, characterized in that Obtain the tire tread pattern image and Bluetooth address of the entering vehicle, generate a characteristic fingerprint map, and construct a characteristic binding structure, including: Obtain the tire tread pattern image of the entering vehicle and the in-vehicle Bluetooth address, perform image filtering and data cleaning to obtain the tire tread pattern image and Bluetooth address data; Extract the contour edge features and address label information from the tire tread pattern image and Bluetooth address data to generate a characteristic fingerprint map.

3. The method according to claim 1, characterized in that Obtain the characteristic binding structure and image sequence, extract path information, and construct a path matching graph structure, including: Obtain the characteristic binding structure and the image sequence of the entering vehicle, perform object detection and temporal tracking processing to obtain the image tracking path; Extract the path point sequence and matching status information from the image tracking path to construct a path matching graph structure.

4. The method according to claim 3, wherein Perform a comprehensive scoring process on the path matching graph structure to generate a trajectory confidence scoring result.

5. The method according to claim 1, wherein Obtain the trajectory confidence scoring result and perform identity verification, generate a digital identity structure, and construct a digital credit voucher, including: Obtain the trajectory confidence scoring result, combine it with the historical binding record, and perform the entrance identity verification; Digitally mark the identity verification result to generate a digital identity structure; Construct a credit voucher based on the digital identity structure and complete the digital signature to construct a digital credit voucher.

6. The method according to claim 1, wherein Obtain the digital credit voucher and binding structure information, perform overdue payment verification, and generate an entrance status instruction to complete the entrance control judgment, including: Obtain the digital credit voucher and binding structure information, match the historical overdue payment record, and complete the verification; Perform a freeze control judgment on the verification result to generate an entrance status instruction; Update the barrier gate control logic based on the entrance status instruction to complete the entrance control judgment.

7. The method according to claim 1, characterized in that, Obtain the entrance status instruction and perform the barrier gate operation, generate an execution record, and complete the control response operation, including: Obtain the entrance status instruction and call the control interface to complete the physical barrier gate execution operation; Combine the physical barrier gate execution operation result with the identity verification record to generate an execution record; Submit the execution record to the entrance behavior storage module to complete the control response operation.

8. The method according to claim 1, wherein Obtain the execution record and image sequence, and perform multi-modal data screening and structure registration processing.

9. The method according to claim 8, characterized in that, Perform local model training based on the multi-modal data screening result to generate incremental model parameters, and synchronize the incremental model parameters to the recognition model platform to complete parameter update.

10. A temporary parking fee arrears management and entry control system for a parking lot, applied to the temporary parking fee arrears management and entry control method described in any one of claims 1-9, characterized in that, Including: A feature acquisition module for obtaining the tire tread pattern image and Bluetooth address of the entering vehicle, generating a characteristic fingerprint map, and constructing a characteristic binding structure; A trajectory recognition module, which is used to obtain a feature binding structure and an image sequence, construct a path matching graph structure and generate a trajectory confidence scoring result; An identity verification module, which is used to obtain the trajectory confidence scoring result and generate a digital identity structure and a digital credit voucher; An admission control module, which is used to obtain the digital credit voucher and the binding structure information, perform overdue payment verification and generate an admission status instruction; An execution control module, which is used to obtain the admission status instruction and complete the barrier gate operation and control response; A model training module, which is used to obtain the execution record and the image sequence and update the parameters of the recognition model platform.

Images