A parking lot intelligent navigation and car searching system and method based on offline caching and multi-source fusion positioning

Abstract

The application relates to the technical field of parking lot management, in particular to a parking lot intelligent navigation and car searching system and method based on offline caching and multi-source fusion positioning, which comprises a user terminal applet module, a multi-source fusion positioning module, a map and path planning module, a backend service and management background module and a data synchronization and verification module. The application adopts an offline caching priority strategy, pre-stores map data, channel node information, Bluetooth beacon coordinates and camera calibration parameters in the local, and realizes near-field updating of parking space states through Bluetooth broadcasting, so that the car searching function can be independently operated in a network-free environment; differentiated vertical movement cost models are established for the elevator, escalator and fire escape to optimize cross-floor path planning, effectively solve the problems of positioning drift and car searching difficulty in a weak network environment of an underground parking lot, and improve car searching efficiency, accuracy and user experience.

Description

Technical Field

[0001] This invention relates to the field of parking management technology, specifically to an intelligent navigation and vehicle-finding system and method for parking lots based on offline caching and multi-source fusion positioning. Background Technology

[0002] With the continuous growth of car ownership in my country, large indoor parking lots are constantly extending underground and developing into multi-level complexes to alleviate parking pressure, resulting in increasingly larger scales and more complex structures. However, car owners often get lost when retrieving their vehicles due to the vast parking spaces, similar environmental signs, difficulty in identifying indoor directions, and complex multi-level structures, wasting a lot of time and energy searching for their cars. Statistics show that over 87% of car owners have encountered difficulties finding their cars in indoor parking lots, and over 76% of car owners reported that the weak network environment in underground parking garages caused parking-related applications to malfunction. At the same time, parking lot operators face challenges such as insufficient parking space utilization (generally below 60%), high manual management costs, and a lack of data-driven operational support. Therefore, intelligent parking navigation and car-finding technology has become an important direction for smart city construction, and several related patent technologies have been published in China. Patent document 1 (CN116013104B) discloses a parking lot car-finding method, device, and system. Its technical solution includes: acquiring the user's car-finding request; determining a cross-floor or same-floor car-finding strategy based on the current location and parking location; calculating horizontal and vertical movement costs based on the information of each floor's passageway nodes, and selecting the optimal path through path cost; and implementing the same-floor strategy based on the A-star algorithm. The main shortcomings of this solution are: First, the entire method heavily relies on real-time network acquisition of map data and user location information. When the network signal in the underground parking lot is weak or interrupted, the user cannot download the map or obtain location information, and the system completely fails. Second, the positioning method is singular, relying solely on external positioning signal sources, without considering positioning drift and signal loss caused by GPS signal obstruction in underground spaces, resulting in unreliable indoor positioning accuracy and continuity. Third, if the user fails to scan the code successfully or forgets to scan it when parking, the system cannot obtain the parking location, lacking backup recording methods and insufficient fault tolerance. Fourth, it only focuses on the car-finding process, failing to form a complete closed loop from entering the parking lot to parking payment, resulting in fragmented functionality. Patent document 2 (CN104269072B) discloses a method, device, and system for finding a car in an underground parking lot. Its technical solution includes: receiving a user's route query command through a query service terminal or mobile terminal, obtaining user location information and parking space information, calculating route navigation information, and generating map data. The main shortcomings of this solution are: First, it also relies on a real-time network connection to obtain map data and perform route calculations, making it unusable in environments without a network. Second, user location acquisition depends on Bluetooth signals or the fixed location of the query terminal, resulting in limited positioning coverage and insufficient accuracy; it lacks the ability to integrate multiple sensor sources for continuous positioning. Third, it only provides a single route calculation from the query terminal to the parking space, lacking real-time location tracking and navigation correction functions; if the user deviates from the route during walking, replanning is not possible. Fourth, the parking space information needs to be obtained and stored by the user in advance through a mobile terminal, which is cumbersome and lacks manual error correction design. Patent document 3 (CN103489328B) discloses a reverse car-finding method for intelligent parking lots. The technical solution includes: storing XML configuration files for each parking lot on a server; when parking, the driver scans a parking space QR code to download the XML file and record the parking location; when retrieving the car, the driver scans any QR code again, and the system displays a route from the scanned location to the parking space. The main shortcomings of this solution are: First, the QR code only provides a download address and location identifier; map data needs to be downloaded in real-time each time, with no offline caching mechanism, making it impossible to obtain maps and routes after network outages; Second, positioning relies entirely on the fixed location of the QR code scan, failing to achieve continuous positioning and real-time navigation during the user's movement, forcing the user to rely on memory after viewing a static route map; Third, it lacks the ability to plan complex paths across multiple floors, and the navigation logic for cross-floor scenarios is not optimized; Fourth, there is no manual data entry fault tolerance mechanism, and the function is completely disabled if the QR code is damaged or scanning fails; Fifth, the XML configuration file method has low data interaction efficiency, and map rendering and path calculation response speeds are slow. Comprehensive comparative analysis reveals that existing parking lot vehicle location technologies generally suffer from the following common shortcomings: (1) Strong network dependence, unusable in weak network environment. All three patent documents mentioned above regard real-time network communication as a prerequisite for system operation. Map data acquisition, location information transmission, and route planning calculation all rely on cloud servers. However, due to the shielding of the building structure, the mobile network signal coverage is poor and the strength is weak in underground parking lots. Users often face the problems of request timeout and data loading failure in actual use scenarios, which makes the car finding function completely unusable at critical moments, forming the paradox of "unusable when most needed".

[0003] (2) Low positioning accuracy and lack of continuous tracking capability. Existing solutions either rely on a single GPS positioning source (which is basically ineffective in underground spaces) or on fixed beacons such as QR codes or Bluetooth (which can only provide discrete location points and cannot achieve continuous position tracking during walking), lacking a multi-source sensor fusion positioning mechanism. During actual walking, the system cannot perceive changes in the user's position in real time, cannot provide route deviation reminders and dynamic path correction, resulting in a poor navigation experience.

[0004] (3) Fragmented functions, failing to form a closed-loop service process. Most existing solutions focus on the single "car finding" step, ignoring the complete user experience chain of "entering the venue to find a parking space → parking record → leaving the venue to find the car → paying and leaving." Users need to use different tools or systems at different stages, resulting in a fragmented experience and low efficiency.

[0005] (4) Lack of fault tolerance mechanism and incomplete scenario coverage. The existing solution assumes that users can successfully scan the QR code when parking, without considering various real-world scenarios such as damaged or missing QR codes, user's mobile phone camera malfunction, or user forgetting to scan the code in a hurry. It also lacks backup solutions such as manual data entry. Once a problem occurs in the scanning process, the entire car-finding process is interrupted.

[0006] (5) Cross-floor path planning is coarse and does not distinguish between different types of passages. Although patent document 1 considers cross-floor scenarios, the calculation of vertical movement costs is relatively rough. It does not perform fine-grained modeling of the differences in user passage efficiency and physical consumption based on different types of passages (elevators, escalators, fire stairs), which may result in the recommended path not being the actual optimal one.

[0007] To address the aforementioned shortcomings, this invention proposes a parking lot intelligent navigation and vehicle locating system and method based on offline caching and multi-source fusion positioning. Summary of the Invention

[0008] The purpose of this invention is to provide a parking lot intelligent navigation and vehicle finding system and method based on offline caching and multi-source fusion positioning, so as to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution: A parking lot intelligent navigation and car-finding system based on offline caching and multi-source fusion positioning includes: The user-side mini-program module provides an interactive interface for scanning and binding parking spaces, manually entering parking spaces, offline car finding, indoor navigation, finding a parking space upon entry, and online payment. It has a built-in offline data cache and local storage unit. The offline data cache unit downloads and stores parking lot map data, access node information, Bluetooth beacon location coordinates, camera installation location mapping table, and parking space mapping relationships to a local database before the user first enters the parking lot and when the network is available. The user-side mini-program module also supports receiving map update data and real-time parking space status information sent by Bluetooth broadcasting devices within the parking lot via Bluetooth communication. This map update data and real-time parking space status information are periodically distributed by the parking management system through a Bluetooth gateway or Bluetooth broadcasting device. It records and reads parking location information during parking and car finding processes, and performs two-way synchronization of local and cloud data after the network is restored. A multi-source fusion positioning module is used to fuse GPS signals, inertial measurement unit (IMU) sensor data, parking lot camera image recognition and parking space occlusion detection data, and Bluetooth beacon positioning data for fusion positioning. It achieves seamless indoor and outdoor positioning and error correction through a Kalman filter algorithm. The IMU sensors include an accelerometer, a gyroscope, and a magnetometer. When the GPS signal is valid, the multi-source fusion positioning module uses GPS positioning as the observation value and IMU estimation as the state prediction for filtering and updating. When the GPS signal is lost, it switches to an indoor positioning mode, which includes the following three sub-modes: (I) Image Recognition and Parking Space Occlusion Detection and Localization Mode: Acquire video images of parking space areas captured by surveillance cameras in the parking lot, identify the vehicle occupancy status of each parking space in the image through target detection algorithm, and when a specific parking space is detected to be occupied by a vehicle, combine the pre-calibrated position of the parking space in the parking lot physical coordinate system and use the pre-calibrated coordinates of the parking space as the vehicle positioning anchor point; at the same time, during the user's car-finding navigation process, the image coordinates are mapped to the parking lot physical coordinate system by combining the pixel position of the screen spot or human body outline of the user's handheld mobile terminal in the image with the camera's intrinsic and extrinsic parameter matrix, and the real-time position estimate of the user is calculated; through the user tracking results of multiple consecutive frames of images, a smooth position estimate is output; (ii) Bluetooth beacon positioning mode: By scanning the Bluetooth broadcasting devices (including Bluetooth gate, Bluetooth geomagnetic detector, and Bluetooth beacon) in the parking lot, the signal strength and media access control address of each device are obtained. Combined with the pre-stored Bluetooth device location coordinate map, the trilateration calculation is performed using the signal arrival time difference or signal strength indication value to obtain the estimated user location value. (III) Image recognition and Bluetooth fusion positioning mode: When the surveillance camera recognizes the light spot on the user's mobile terminal screen or the outline of a human body, it combines the Bluetooth beacon positioning results and outputs the fused position estimate through a weighted fusion algorithm. When none of the above-mentioned auxiliary positioning sources are available, the multi-source fusion positioning module enters a pure inertial calculation mode and performs position tracking based on the second integral of acceleration and the gyroscope heading angle. After GPS signal loss, the Kalman filter algorithm uses any one or more combinations of image recognition and parking space occlusion detection positioning results, Bluetooth beacon positioning results, and image recognition and Bluetooth fusion positioning results as observations to periodically correct the cumulative drift of inertial estimation and output a continuous and stable indoor position estimate.

[0010] Map and Path Planning Module: The map and path planning module is used to store and load offline map data, and performs path planning on the same floor and across floors based on the improved A-star algorithm. The cross-floor path planning comprehensively considers the horizontal and vertical movement costs of the passage nodes on each floor. The vertical movement cost is calculated differently according to the type of passage node.

[0011] Backend Service and Management Backend Module: The backend service and management backend module provides cloud-based management of parking lot map data, parking space information, QR code resources, parking orders and charging standards, as well as functions for adding, deleting, modifying and querying parking space information, QR code resources and parking data, and statistical analysis of parking duration, parking space utilization rate and user behavior.

[0012] Data Synchronization and Verification Module: The data synchronization and verification module is used to automatically synchronize local offline cached data with cloud data bidirectionally when the network recovers, and to verify data consistency through conflict resolution strategies based on timestamps and version numbers.

[0013] A parking lot intelligent navigation and vehicle locating method based on offline caching and multi-source fusion positioning, applied to the aforementioned parking lot intelligent navigation and vehicle locating system based on offline caching and multi-source fusion positioning, includes the following steps: S1: The user-side mini-program module obtains the user's car-finding request, detects the current network status, and determines the user's current location and parking location. If the network is available, the system retrieves parking lot map data from the cloud server, reads the user's latest parking location information from the cloud database, and obtains the user's current location information via GPS or base station positioning. If the network is unavailable, the user-side mini-program module first attempts to receive map update data and parking space status information sent by the Bluetooth broadcasting device in the parking lot via Bluetooth communication, incrementally updates the local offline cache, then reads the parking lot map data and parking location information from the local offline cache, and uses the multi-source fusion positioning module to calculate the user's current indoor location using image recognition and parking space occlusion detection positioning mode, Bluetooth beacon positioning mode, or image recognition and Bluetooth fusion positioning mode. The network status detection includes network quality classification judgment: When the heartbeat request response time is less than the first preset threshold, it is determined to be a high-quality network, and full data synchronization in online mode is performed; When the response time is between the first preset threshold and the second preset threshold, the network is determined to be available, online mode is executed, and the loading of non-critical data is postponed until navigation is completed. If there is no response after a timeout or the response time exceeds the second preset threshold, the network is determined to be unavailable, and the system immediately switches to a fully offline mode without blocking the user's operation process.

[0014] S2: The map and route planning module determines whether cross-floor car searching is involved based on the user's current location and parking location. If the vehicle is located across multiple floors, the cross-floor vehicle search strategy will be used; otherwise, the same-floor vehicle search strategy will be used. The cross-floor vehicle search strategy calculates the optimal cross-floor path based on the information of each floor's passageway nodes, the vertical movement cost corresponding to the passageway node type, and offline map data. The same-floor vehicle search strategy calculates the optimal same-floor path based on the improved A-star algorithm. S3: The map and path planning module, based on the selected car-finding strategy and combined with parking lot map data, current location and parking location, uses a path planning algorithm to generate at least one candidate car-finding path from the current location to the parking location. The final car-finding path is determined based on the total movement cost of each candidate path, where the total movement cost is obtained by summing the horizontal movement cost and the vertical movement cost. S4: During navigation, the multi-source fusion positioning module collects acceleration, gyroscope and geomagnetic data from the built-in inertial measurement unit of the mobile phone in real time, and fuses GPS signal, inertial measurement unit inferred data, image recognition and parking space occlusion detection positioning data and Bluetooth beacon positioning data through Kalman filtering algorithm. Dynamic error correction is performed on indoor positioning data, and the map and route planning module matches the corrected location information with the final car-finding route on the map. The user terminal applet module updates the user's location in real time and guides the user to the parking location along the final car-finding route. S5: Upon arrival at the parking location, the user-side mini-program module sends a notification signal, automatically updates the parking status, and synchronizes it to the cloud server and local offline cache.

[0015] As a preferred embodiment of the present invention, the indoor positioning mode of the multi-source fusion positioning module after GPS signal loss specifically includes: (a) Image recognition and parking space occlusion detection and positioning mode: The system acquires video streams of parking space areas captured by surveillance cameras in the parking lot, extracts frames from the video streams, and uses a target detection algorithm to identify the vehicle occupancy status of each parking space in the image. The parking space occupancy status includes two states: "idle" and "occupied". When a vehicle is detected to be blocking a specific parking space, the coordinates of the blocking parking space are used as the vehicle positioning anchor point, based on the pre-calibrated position of the parking space in the parking lot's physical coordinate system. At the same time, by using the pixel position of the screen light spot or human outline of the user's handheld mobile device in the image, combined with the camera's intrinsic and extrinsic parameter matrix, the image coordinates are mapped to the parking lot's physical coordinate system to calculate the user's rough position estimate. The system outputs a smooth position estimate based on the vehicle tracking and user tracking results from multiple consecutive frames of images. (ii) Bluetooth beacon positioning mode: The user-side mini-program module scans the Bluetooth broadcasting devices in the parking lot to obtain the signal strength and media access control address of each device. The Bluetooth broadcasting devices include Bluetooth barrier gates, Bluetooth geomagnetic detectors, and Bluetooth beacons. Combining the pre-stored Bluetooth device location coordinate map, trilateration is calculated using signal strength indicators. The specific calculation formula is as follows: For the i-th Bluetooth device, its position coordinates are The signal strength is According to the signal propagation model Calculate the distance from the user to the device ,in, For the signal strength at the reference distance, The path loss index; The user's location is obtained by solving the following system of nonlinear equations. The estimated value: Where m≥3, the least squares method or the weighted centroid method is used to solve the problem; (III) Image recognition and Bluetooth fusion positioning mode: When the surveillance camera detects a light spot on the user's mobile terminal screen or a human silhouette, it obtains a first position estimate through the camera calibration parameters. ; When Bluetooth beacon positioning can obtain a second location estimate At that time, the weighted fusion algorithm outputs the fused location estimate: Among them, weight and Dynamically adjusted based on image recognition confidence level and Bluetooth signal strength to meet [the requirements]. + =1; When only one location source is available, the location estimate of that location source is used directly as the output. (iv) Pure inertial calculation model: When none of the above three positioning modes are available, the displacement is obtained by double integration based on the acceleration data of the inertial measurement unit, the change in heading angle is obtained based on the gyroscope data, and the heading angle is corrected based on the geomagnetic sensor data. The current indoor position estimate is then obtained by recursion through the dead reckoning algorithm. The Kalman filter algorithm takes any one or more combinations of the above image recognition and parking space occlusion detection positioning results, Bluetooth beacon positioning results, and image recognition and Bluetooth fusion positioning results as the observation input, performs periodic correction on the cumulative drift of inertial estimation, outputs a continuous and stable indoor position estimate, and uses GPS observations to correct the cumulative error of dead reckoning after the GPS signal is recovered. The state equation and observation equation of the Kalman filter algorithm are as follows: Equations of state: Observation equation: in: This is the state vector at time k, containing position coordinates, velocity, and attitude angle information; This is the state transition matrix; The acceleration and angular velocity measured by the inertial measurement unit; For control matrix; It is process noise, and follows a zero-mean Gaussian distribution; These are GPS or geomagnetic observations; The observation matrix; To observe the noise, it follows a zero-mean Gaussian distribution; The time update equation and measurement update equation for the Kalman filter are as follows: Updated in time: Measurement Update: in: The process noise covariance matrix is... To observe the noise covariance matrix, Here is the Kalman gain matrix. To estimate the error covariance matrix, It is an identity matrix.

[0016] As a preferred embodiment of the present invention, the Bluetooth communication and offline data transmission mechanism includes: When a user enters the parking lot for the first time and the network is available, the system automatically downloads and stores the complete map data of the current parking lot, information on the passage nodes of each floor, information on the accessible floors corresponding to each passage node, the mapping relationship between QR code identifiers and parking spaces, the location coordinate map of Bluetooth broadcasting devices (including Bluetooth gate barriers, Bluetooth geomagnetic detectors, and Bluetooth beacons), the media access control address and location mapping table of each Bluetooth device, the mapping table of camera numbers and installation locations, and the internal and external parameter matrices of cameras to the local database. The map data includes vectorized floor plans of each floor, the topology of the passageway, parking space numbers and coordinates, and the coordinates and types of passage nodes.

[0017] In addition, the parking management system periodically transmits updated parking map data and real-time parking space status information via Bluetooth gateways or Bluetooth broadcast devices in the form of Bluetooth broadcast packets. The user-side mini-program module periodically scans for Bluetooth broadcast signals when offline. Upon receiving a broadcast packet, it verifies data integrity using a checksum. If the data version number is higher than the local cache version number, it decodes the data content and incrementally updates the local offline cache, ensuring the timeliness of local data even when offline.

[0018] When a user parks, the user-side mini-program module records the parking location information by scanning the parking space QR code or manually entering the parking space number. The parking location information includes the parking space number, floor, coordinates, parking timestamp, and vehicle identifier. It is stored in the local offline cache and asynchronously synchronized to the cloud server by the data synchronization and verification module after the network is restored. When the network is unavailable, all vehicle search requests, map readings, route calculations, and navigation operations are completed independently based on locally cached data. Once the network is restored, the data synchronization and verification module automatically performs bidirectional synchronization and consistency verification between the local offline cached data and the cloud data. The consistency verification adopts a conflict resolution strategy based on timestamps and version numbers, with the most recently updated data taking precedence.

[0019] As a preferred embodiment of the present invention, the cross-floor car-finding strategy in step S2 specifically includes: The starting floor and the coordinates of the starting point within the starting floor are determined based on the user's current location, and the ending floor and the coordinates of the ending point within the ending floor are determined based on the parking location.

[0020] Read all passage nodes in the starting floor map data, and obtain the node type, node location coordinates and reachable floor information list for each passage node. The node types include elevators, escalators and fire escape stairs.

[0021] Select candidate channel nodes from all channel nodes that meet one of the following conditions: For the two types of candidate channel nodes selected through screening, different methods are used to calculate the total path cost: For the first type of channel node (directly accessible to the destination floor), the horizontal movement cost from the starting point to the starting floor of the channel node, the vertical movement cost between floors corresponding to the channel node, and the horizontal movement cost from the exit of the destination floor to the destination floor are directly calculated, and the sum of these three costs is taken as the total path cost of the path corresponding to the channel node; For the second type of channel node (requiring a transfer), the node is first marked as a transfer node, and the complete total path cost, including the horizontal movement cost of the transfer floor, is calculated after the transfer floor is processed. The total path cost of both types of nodes is quantified using the same metric (total travel distance converted value) to ensure the fairness and accuracy of the total path cost comparison.

[0022] For each candidate channel node, calculate the horizontal movement cost from the starting point to the starting floor of the candidate channel node, the vertical movement cost between floors corresponding to the candidate channel node, and the horizontal movement cost from the end floor channel exit to the end point. The sum of these three costs is taken as the total path cost of the path corresponding to the candidate channel node.

[0023] The path corresponding to the candidate channel node with the minimum total path cost is selected as the cross-floor vehicle search path.

[0024] As a preferred embodiment of the present invention, the horizontal movement cost is the actual travel path length between the two points, which is obtained by performing the A-star algorithm on the floor plan.

[0025] The vertical movement cost is determined based on the type of passageway node: when the passageway node type is a vertical elevator, the vertical movement cost is a preset distance value calculated from the elevator waiting time; when the passageway node type is an escalator, the vertical movement cost is zero; when the passageway node type is a fire escape staircase, the vertical movement cost is calculated based on the floor height, the staircase inclination angle, and the number of floors that need to be traversed, using the following formula: in: For the cost of vertical movement, The height of a single floor. The angle of inclination of the fire escape staircase. This represents the number of floors that need to be walked through. As a weighting factor, during upward movement =2, downlink =1.

[0026] When there is no direct passage node to the destination floor, the cross-floor vehicle search strategy further includes: determining intermediate floors, finding transfer passage nodes in the intermediate floors that can reach the destination floor, calculating the horizontal movement cost from the exit of the candidate passage node at the starting floor to the transfer passage node at the intermediate floor, and including it in the total path cost calculation.

[0027] As a preferred embodiment of the present invention, it further includes a dual-fault-tolerant parking recording mechanism, specifically comprising: The first recording method: After parking, the user scans the QR code on the parking space with a mobile terminal. The system parses the parking space number and parking lot sign encoded in the QR code, binds the obtained parking space number to the current user account, and records the parking location.

[0028] The second recording method: When the QR code scan fails or the QR code is missing, the user can manually enter the parking space number through the mobile terminal interface. The system will then match the entered parking space number with the local cache or cloud map data to find the corresponding parking space coordinates and complete the parking location record.

[0029] The third recording method: When a user neither scans the code nor manually enters information when parking, the user initiates a car-finding request and enters the license plate number, parking time range, or parking space area information on the mobile terminal interface. The system performs fuzzy matching through local cache or cloud database and returns a list of matching parking location candidates for the user to select and confirm. The fuzzy matching supports partial matching of license plate numbers and filtering of time ranges.

[0030] As a preferred embodiment of the present invention, it also includes an intelligent positioning function for entry: Users can scan the QR code at the parking lot entrance or elevator lobby, or trigger an entry and parking space request through a mobile terminal interface.

[0031] The system obtains real-time status information of all available parking spaces in the current parking lot. The real-time status information is obtained from the cloud server or read from the most recently updated data cached locally.

[0032] Based on the user's current location, the coordinates of each available parking space, and the topology of the traffic paths within the parking lot, a path planning algorithm is used to calculate the candidate entry paths from the user's current location to each available parking space and their path lengths.

[0033] Available parking spaces are sorted in ascending order of path length. The available parking space with the shortest path length is selected as the recommended parking space. An entry navigation path from the user's current location to the recommended parking space is generated and presented to the user.

[0034] Once a user arrives at a recommended parking space, the system automatically binds the parking space to the user's account, records the parking location, and stores it in the local cache and cloud server. If a user parks in another parking space instead of the recommended one, the system will detect the inconsistency between the user's location and the recommended parking space through continuous location tracking. When the system detects that the user's stay time exceeds a preset threshold, it will automatically trigger location re-identification and update the parking space closest to the user's actual location to the bound parking space.

[0035] Furthermore, to address situations where a user selects a recommended parking space but parks in an adjacent space due to a parking operation error, the system is configured with adjacent parking space determination logic: When the multi-source fusion positioning module detects that the distance between the user's stopping position and the recommended parking space is less than a first preset threshold (e.g., 3 meters) but greater than a second preset threshold (e.g., 1 meter), a precise positioning confirmation process is initiated. This process uses fine-grained displacement detection (acceleration rate of change analysis) by the inertial measurement unit to determine whether the user has actually entered the recommended parking space area. If it is determined that the user has entered an adjacent parking space instead of a recommended parking space, the system matches the nearest parking space based on the user's final stopping position and binds it, then sends a confirmation prompt to the user, avoiding incorrect parking space binding due to positioning accuracy errors.

[0036] As a preferred embodiment of the present invention, the specific method for map matching in step S4 includes: The user's location coordinates, corrected by Kalman filtering, are projected onto the traffic path network of the parking lot map, and the vertical distance between the projected point and each path node on the final vehicle-finding path is calculated.

[0037] If the vertical distance is less than a preset matching threshold, the user's position will be corrected to the nearest node on the final vehicle search path to achieve position correction.

[0038] If the vertical distance exceeds a preset matching threshold, the user is determined to have deviated from the final car-finding path. Upon determining this deviation, to avoid unnecessary route replanning due to brief positional fluctuations (such as temporary avoidance of pedestrians or vehicles, or temporary stops), the system first activates a deviation confirmation timer. If the user returns to the vicinity of the path within a preset time (e.g., 10 seconds) (vertical distance less than the matching threshold), the deviation warning is canceled, and navigation continues. If the user remains deviated within the preset time, the system proactively prompts the user via voice and text that they have deviated from the planned path and asks if they wish to replan an alternative route. If the user confirms the replanning via the interface, or if the user continues to walk beyond a preset distance (e.g., 20 meters) in the direction of deviation without confirmation, the system automatically replans the optimal alternative route from the current user location to the parking location and updates the navigation guidance. This mechanism ensures navigation accuracy while respecting the user's autonomous route selection, avoiding interference from mechanical automatic replanning.

[0039] During navigation, the final vehicle-finding route is represented by a sequence of path segments. The user's location is updated in real time and moves along the sequence of path segments. When the user's location crosses the current target path node, it automatically switches to the next path segment and guides the user to turn at path turning points through voice and text prompts.

[0040] As a preferred embodiment of the present invention, the method for constructing the parking lot map data includes: Based on the actual layout of the parking lot, the passage areas on each floor are mapped as a passage path network. This network consists of path segments and nodes. Each path segment is a straight line connecting two nodes, located in the center of the passageway. The endpoint coordinates of the path segments are generated using Mercator coordinates. For two intersecting path segments, a new node is generated at the intersection, dividing the original path segment into four path segments. Passage nodes between floors are recorded as cross-floor connection nodes. Each passage node records its node type, node coordinates, current floor, and a list of accessible floors.

[0041] Complete the parameter calibration of the surveillance cameras in the parking lot, and record the device number, installation location coordinates, shooting coverage area (corresponding parking space number range), and intrinsic and extrinsic parameter matrices (including focal length, principal point coordinates, distortion coefficients, rotation matrix, and translation vector) for each camera. The camera intrinsic and extrinsic parameter matrices are used to map the pixel coordinates in the image coordinate system to the physical coordinate system of the parking lot.

[0042] Complete the parameter calibration of the Bluetooth broadcasting equipment in the parking lot, and record the device number, device type (Bluetooth barrier gate / Bluetooth geomagnetic detector / Bluetooth beacon), installation location coordinates, media access control address, broadcast signal coverage (effective radius), and signal propagation model parameters (signal strength RSSI0 at reference distance, path loss index n) for each Bluetooth device.

[0043] Parking map data is stored in a structured file format (GeoJSON or a custom binary format) and deployed on a cloud server. It can also be downloaded and stored in a local offline cache with the same data structure.

[0044] Compared with the prior art, the beneficial effects of the present invention are: This invention fundamentally solves the rigid dependence of existing technologies on real-time networks by using an offline cache-first data access strategy combined with a Bluetooth near-field data blind spot compensation mechanism. All core data, such as parking lot map data, channel node information, Bluetooth beacon coordinates, and camera calibration parameters, are pre-stored locally on the user terminal. All functions, such as map loading, route planning, location calculation, and navigation guidance, can run independently in a completely network-free environment. At the same time, the parking management system periodically broadcasts parking space status and incremental map update data through a Bluetooth gateway. Even when offline, users can still obtain the latest parking space occupancy information and map change data by scanning the Bluetooth broadcast signal, realizing near-field real-time data updates in offline environments. This completely frees the navigation function from the constraints of network conditions, allowing users to smoothly complete all car-finding operations in underground parking lots with no signal coverage. This invention constructs an integrated positioning architecture that deeply fuses outdoor GPS and indoor multi-source heterogeneous positioning. Utilizing the surveillance cameras already widely deployed in parking lots, image recognition and parking space occlusion detection technologies are used to establish pre-calibrated coordinates of occluded parking spaces as absolute position reference anchors. Simultaneously, real-time position tracking is achieved by recognizing light spots on the user's mobile phone screen or human silhouettes. Existing Bluetooth barriers, Bluetooth geomagnetic detectors, and Bluetooth beacons within the parking lot are used to achieve continuous position estimation through trilateration based on signal strength. In the absence of any external positioning sources, short-term, high-precision relative displacement calculations are performed using the mobile phone's built-in inertial measurement unit. The three positioning sources are deeply fused using a Kalman filter, with the weights of each source dynamically adjusted based on signal quality and confidence level. Abnormal observations are automatically downweighted or removed. This multi-source heterogeneous fusion positioning architecture significantly improves positioning accuracy in underground parking lots where GPS is completely blocked. It enables real-time smooth position tracking, automatic route deviation detection, and dynamic path correction during user movement, effectively solving problems such as positioning drift and inaccurate positioning caused by signal occlusion in underground parking lots. This invention establishes a differentiated vertical movement cost assessment system for different types of cross-floor passage nodes: elevators consider waiting time as an equivalent passage cost, escalators have a cost of zero, and fire escape staircases quantify physical exertion costs based on floor height, inclination angle, and direction of travel. This refined assessment model better reflects users' actual perception—the passage cost of taking an elevator or escalator is far lower than walking up stairs, and the path planning results tend to recommend mechanical lifting equipment rather than fire escape staircases. This design effectively avoids the unreasonable result of treating all passages the same in existing technologies, which leads to recommending users to walk up multiple floors of stairs. User acceptance of the recommended routes is significantly improved, and satisfaction with finding parking spaces is greatly enhanced. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the overall process of the present invention. Detailed Implementation

[0046] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0047] To facilitate understanding of the present invention, a more comprehensive description of the invention will be given below with reference to the accompanying drawings, and several embodiments of the invention will be provided. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that the disclosure of the invention will be more thorough and complete.

[0048] For examples, please refer to Figure 1 The present invention provides a technical solution: A parking lot intelligent navigation and car-finding system based on offline caching and multi-source fusion positioning includes: The user-side mini-program module provides an interactive interface for scanning and binding parking spaces, manually entering parking spaces, offline car finding, indoor navigation, finding parking spaces upon entry, and online payment. It includes a built-in offline data cache and local storage unit. The offline data cache unit is used when users actively search for and confirm a target parking lot via the mini-program in a network environment (such as at home, on the street, or at the parking lot entrance), or when scanning a QR code at the parking lot entrance, triggering the download and storage of parking lot map data, passage node information, Bluetooth beacon location coordinates, camera installation location mapping table, and parking space mapping relationships to a local database. The user-side mini-program module also supports receiving map update data and real-time parking space status information sent by Bluetooth broadcasting devices within the parking lot via Bluetooth communication. This map update data and real-time parking space status information are periodically distributed by the parking management system through a Bluetooth gateway or Bluetooth broadcasting device. It records and reads parking location information during parking and car finding processes, and performs two-way synchronization of local and cloud data after network recovery. It should be further explained that when a user searches for or selects a target parking lot on the mini-program homepage, the system prompts "Do you want to download the offline map data of this parking lot for use when there is no network?" After the user confirms, the download begins and is stored in the local cache after completion.

[0049] A multi-source fusion positioning module is used to fuse GPS signals, inertial measurement unit (IMU) sensor data, parking lot camera image recognition and parking space occlusion detection data, and Bluetooth beacon positioning data for fusion positioning. It achieves seamless indoor and outdoor positioning and error correction through a Kalman filter algorithm. The IMU sensors include an accelerometer, a gyroscope, and a magnetometer. When the GPS signal is valid, the multi-source fusion positioning module uses GPS positioning as the observation value and IMU estimation as the state prediction for filtering and updating. When the GPS signal is lost, it switches to an indoor positioning mode, which includes the following three sub-modes: (I) Image recognition and parking space occlusion detection and positioning mode: acquire video images of parking space areas collected by monitoring cameras in the parking lot, identify the vehicle occupancy status of each parking space in the image through target detection algorithm, and when a specific parking space is detected to be occluded by a vehicle, combine the pre-calibrated position of the parking space in the physical coordinate system of the parking lot, and use the coordinates of the occluded parking space as the vehicle positioning anchor point. At the same time, calculate the user's rough position estimate by using the pixel position of the vehicle outline in the image and the camera's intrinsic and extrinsic parameter matrix. (ii) Bluetooth beacon positioning mode: By scanning the Bluetooth broadcasting devices (including Bluetooth gate, Bluetooth geomagnetic detector, and Bluetooth beacon) in the parking lot, the signal strength and media access control address of each device are obtained. Combined with the pre-stored Bluetooth device location coordinate map, the trilateration calculation is performed using the signal arrival time difference or signal strength indication value to obtain the estimated user location value. (III) Image recognition and Bluetooth fusion positioning mode: When the surveillance camera recognizes the light spot on the user's mobile terminal screen or the outline of a human body, it combines the Bluetooth beacon positioning results and outputs the fused position estimate through a weighted fusion algorithm. When none of the above-mentioned auxiliary positioning sources are available, the multi-source fusion positioning module enters a pure inertial calculation mode and performs position tracking based on the second integral of acceleration and the gyroscope heading angle. After the GPS signal is lost, the Kalman filter algorithm uses any one or more combinations of the image recognition and parking space occlusion detection positioning results, Bluetooth beacon positioning results, and image recognition and Bluetooth fusion positioning results as observations to periodically correct the cumulative drift of inertial estimation and output a continuous and stable indoor position estimate. The map and path planning module is used to store and load offline map data, and to perform path planning on the same floor and across floors based on the improved A-star algorithm. The cross-floor path planning comprehensively considers the horizontal and vertical movement costs of the passage nodes on each floor. The vertical movement cost is calculated differently according to the type of passage node. The backend service and management backend module is used to provide cloud management of parking lot map data, parking space information, QR code resources, parking orders and charging standards, as well as the functions of adding, deleting, modifying and querying parking space information, QR code resources and parking data, and statistical analysis of parking duration, parking space utilization rate and user behavior. The data synchronization and verification module is used to automatically synchronize local offline cached data with cloud data bidirectionally when the network is restored, and to verify data consistency through conflict resolution strategies based on timestamps and version numbers.

[0050] A parking lot intelligent navigation and vehicle locating method based on offline caching and multi-source fusion positioning, applied to the aforementioned parking lot intelligent navigation and vehicle locating system based on offline caching and multi-source fusion positioning, includes the following steps: S1: Obtaining the vehicle search request and preparing data Users can initiate a car-finding request by scanning any QR code in the parking lot (including parking space QR codes, entrance QR codes, elevator lobby QR codes, etc.) via WeChat mini-program, or by directly clicking the mini-program function entry.

[0051] After receiving a car-finding request, the system first checks the network connection status of the current mobile device. Specifically, if the heartbeat request response time is less than 1 second, it is considered a high-quality network, and full data synchronization in online mode is performed. If the response time is between 1 and 3 seconds, it is considered an available network, and online mode is performed, but the loading of non-critical data (such as historical parking records and statistical data) is postponed until navigation is complete. If there is no response after a timeout or the response time exceeds 3 seconds, it is considered an unavailable network, and the system immediately switches to a completely offline mode without blocking the user's operation process.

[0052] If the network is available, the system will perform online mode operations: send a data request to the cloud server to obtain the complete map data of the current parking lot (including vectorized floor plans of each floor, topology of the traffic path, coordinates of the endpoints of each path segment, node attributes, etc.), the user's latest parking location information (read from the cloud database), and the user's current location information obtained through GPS / base station positioning.

[0053] If the network is unavailable, the system will operate in offline mode: The user-side mini-program module will first attempt to scan for broadcast signals emitted by Bluetooth broadcast devices (including Bluetooth gates, Bluetooth geomagnetic detectors, and Bluetooth beacons) in the parking lot via Bluetooth communication. If a Bluetooth broadcast packet is received, the system will verify the data integrity using a checksum. If the data version number is higher than the local cache version number, the system will decode the data content (including incremental map data or a list of parking space status changes) and incrementally update the local offline cache to ensure that the offline map data and parking space status information are up-to-date and available.

[0054] Then, it reads the pre-stored current parking lot map data (including vectorized floor plans of each floor, traffic path topology, channel node information, Bluetooth device location coordinate map, camera installation location mapping table and internal and external parameter matrix) and user parking location information from the local offline cache database (using SQLite or a similar lightweight embedded database); at the same time, it starts the multi-source fusion positioning module, and uses image recognition and parking space occlusion detection positioning mode, Bluetooth beacon positioning mode or image recognition and Bluetooth fusion positioning mode to estimate the user's current indoor location. The specific positioning mode selection is dynamically determined according to the actual available positioning data sources in the parking lot.

[0055] When a user uses the mini-program for the first time in a network environment and selects a target parking lot, the system automatically initializes the local offline cache in the background, downloads and stores the complete data package for that parking lot. Thereafter, all user operations in that parking lot will prioritize using the locally cached data, with incremental updates only performed when the network is available.

[0056] S2: Vehicle Search Strategy Selection Based on the user's current location and parking location obtained by S1, and combined with the floor information in the parking lot map data, it is determined whether the two are on the same floor. Specifically, the floor field of the user's current location and the floor field of the parking location are read and compared.

[0057] If the two are on the same floor, the same-floor car-finding strategy is adopted. This strategy is based on the improved A-star algorithm to search for the shortest path from the current location to the parking location in the current floor's access path network.

[0058] If the two floors are different, a cross-floor car-finding strategy is adopted. This strategy first reads the access node information of each floor in the parking lot map data (the unique identifier of each access node, node type, node coordinates, the floor it is located on, the list of accessible floors, and the connectivity between access nodes), and then searches for the optimal cross-floor path from the starting floor to the ending floor based on this information.

[0059] S3: Path Planning and Optimal Path Determination Based on the vehicle search strategy type determined by S2, perform the corresponding path planning calculation: For the same-floor car-finding strategy, taking the user's current location as the starting point and the parking location as the ending point, an improved A-star algorithm is executed in the path network of the current floor. The evaluation function of this algorithm is: in: From the starting point to the current node The actual travel path length (obtained by accumulating the lengths of the path segments already traversed in the path network); For the current node Heuristic distance estimation to the target node (using Euclidean distance, i.e.) The algorithm starts from the starting point, maintains an open list and a closed list, and continuously expands the node with the smallest evaluation function value until the target node is reached. Then, it backtracks from the target node to obtain the optimal path.

[0060] For the cross-floor vehicle search strategy, the search space is expanded to multiple floors. Specifically: the starting point is the starting position on the starting floor, and the destination is the destination position on the ending floor. The path switches between floors through passage nodes. The complete cost of the cross-floor path consists of three parts: the horizontal movement cost from the starting position to the candidate passage node within the starting floor, the vertical movement cost between floors corresponding to each passage node, and the horizontal movement cost from the passage exit to the destination position within the ending floor. If intermediate floors are involved, the horizontal movement cost within the intermediate floors also needs to be added.

[0061] For each candidate cross-floor path, calculate the total path cost as described above. The formula for calculating the total path cost is: If a transfer floor is involved, then: After calculating the total cost of all feasible cross-floor candidate paths, sort them by total cost from smallest to largest, and select the path with the smallest total cost as the final vehicle-finding path.

[0062] During the route planning process, all map data reading and route calculations prioritize the use of local offline cached data. Only when the network is available and the local data version lags behind the cloud version will a cloud data update request be initiated.

[0063] S4: Real-time navigation and positioning correction Once the final vehicle location route is determined, the system will present the route as a sequence of line segments on the mini-program map interface and start the real-time navigation process.

[0064] During navigation, the system collects acceleration, gyroscope, and geomagnetic data from the phone's built-in IMU at a fixed frequency (e.g., 20Hz). After preprocessing (including noise reduction, zero-bias correction, and scale factor compensation), the raw data is input into a Kalman filter for multi-source fusion positioning calculation.

[0065] The core calculation process of the Kalman filter is as follows: The state vector is defined as: in( Let be the three-dimensional position coordinates at time k. ) represents the three-dimensional velocity components, ) represents the roll angle, pitch angle, and yaw angle.

[0066] State prediction (time update) is based on IMU acceleration and angular velocity measurements: in, Acceleration and angular velocity measured by the IMU; The control matrix maps IMU measurements to state changes. This is the process noise covariance matrix, which characterizes the IMU measurement noise and model error.

[0067] When the GPS signal strength is higher than a preset threshold (e.g., signal-to-noise ratio > 30dB), the GPS positioning result is used as the observation value input to the filter. The measurement update equation is: in: To observe the noise covariance matrix, Here is the Kalman gain matrix. To estimate the error covariance matrix, It is the identity matrix; The value is dynamically adjusted based on the GPS signal quality, taking a smaller value when the signal quality is good and a larger value when the signal quality is poor.

[0068] When GPS signals are lost, the filter enters a pure inertial estimation mode. The heading angle observations provided by the geomagnetic sensor are partially replaced to correct heading angle drift.

[0069] The optimal position estimate output by the Kalman filter Positional components in , This is the user's current location result.

[0070] The location results are projected onto the traffic path network of the parking lot map, and map matching is performed: the vertical distance from the projection point to each path segment on the final car-finding path is calculated. If the minimum vertical distance is less than a preset matching threshold (e.g., 5 meters), the user's position is corrected to the nearest point on the final car-finding path, achieving position correction. If the minimum vertical distance is greater than the matching threshold, the user is determined to have deviated from the path. When the system determines that the user has deviated from the path, it first starts a deviation confirmation timer. If the user returns to the vicinity of the path within a preset time and the vertical distance is again less than the matching threshold, the deviation prompt is canceled, and navigation continues along the original path. If the user remains in a deviated state for a preset time (e.g., 10 seconds), the system proactively prompts the user with voice and text that they have deviated from the planned path and asks whether to replan an alternative path. If the user confirms the replanning, or if the user continues to walk in the deviated direction for more than a preset distance (e.g., 20 meters) without confirmation, the system automatically replans the optimal alternative path from the current user position to the parking position and updates the navigation guidance. This mechanism aims to avoid unnecessary path replanning triggered by short-term position fluctuations (e.g., the user temporarily avoiding pedestrians or vehicles), while respecting the user's autonomous route selection wishes.

[0071] The user's location is updated in real time on the map, and the path is highlighted. The system guides the user along the path at turning points through voice announcements and text prompts (such as "Turn left in 30 meters ahead").

[0072] In this embodiment, the offline positioning mode of the multi-source fusion positioning module after GPS signal loss is specifically implemented as follows: Taking the B2 level of a three-story underground parking garage as an example, this level has no GPS signal coverage, but the parking garage has deployed surveillance cameras (4-6 per area, covering all parking spaces), Bluetooth barriers (at the entrance / exit), Bluetooth geomagnetic detectors (embedded under each parking space to detect the occupancy status of the parking space), and Bluetooth beacons (evenly distributed at intervals of 8-12 meters, deployed on columns and walls).

[0073] Scenario 1: The user is located within the camera's coverage area and there is a Bluetooth beacon nearby. The multi-source fusion positioning module simultaneously activates two modes: image recognition and parking space occlusion detection positioning, and Bluetooth beacon positioning. Image recognition mode: Acquire video streams of parking space areas captured by parking lot surveillance cameras. An object detection algorithm (using a lightweight YOLO model) identifies the vehicle occupancy status of each parking space in the image in real time. When a user passes a parking space, the camera captures the light spot from the user's phone screen (the light emitted from the screen when the user holds the phone to view a map). Using the camera's intrinsic and extrinsic parameter matrices, the user's image coordinates are mapped to the parking lot's physical coordinate system, outputting a first position estimate, pcam. The positioning frequency is approximately 1-2 Hz, and the positioning accuracy is approximately 1-3 meters.

[0074] Bluetooth Beacon Mode: The user's mobile phone scans for four nearby Bluetooth beacons and obtains the RSSI value and MAC address of each beacon. Based on a pre-stored map of Bluetooth device location coordinates, the system performs trilateration calculations using signal strength indicators and outputs a second location estimate (pble). The positioning frequency is approximately 5Hz, and the positioning accuracy is approximately 3-5 meters.

[0075] Weighted fusion: The system dynamically calculates the fusion weights based on the current image recognition confidence level (based on the category probability value of object detection, ranging from 0 to 1) and the Bluetooth signal strength variance (reflecting signal stability). When the user is directly in front of the camera and the lighting is sufficient, the image recognition confidence level is high (>0.85). Take 0.7, The confidence level is set to 0.3; when the user is in the edge area of ​​the camera or in insufficient lighting, the image recognition confidence level is low (<0.6). Take 0.3, e is set to 0.7. The final output is the fused location estimate. .

[0076] Scenario 2: The user is located in a parking space area, and the location is determined by detecting obstructions in the parking space. When a user parks in a parking space, the surveillance camera above that space detects that the space has changed from "vacant" to "occupied" (the vehicle is obscuring the parking space markings). The multi-source fusion positioning module uses the pre-calibrated coordinates of this parking space (marked during the parking lot map construction phase) as the absolute positioning anchor point for the user's current location. When the user walks between adjacent parking spaces, the system tracks the user's relative displacement using IMU dead reckoning, continuously using changes in parking space occupancy status detected by the camera (obstructions caused by the user passing different parking spaces) as position correction points to eliminate accumulated IMU drift. This mode also applies when the user is searching for their vehicle—when the system detects that the parking space corresponding to the user's parking location is obscured by a vehicle, it uses the coordinates of that parking space as the navigation endpoint for precise guidance.

[0077] Scenario 3: The user is located in an area without camera coverage and the Bluetooth signal is weak. When a user is located in a parking lot corner, stairwell, or other area without camera coverage and with a weak Bluetooth signal (only 1-2 Bluetooth beacons can be received, making trilateration impossible), the system enters a pure inertial estimation mode: gait is detected based on IMU acceleration data (peak detection method), step length is estimated based on step frequency and acceleration amplitude (Weinberg formula), and the heading angle is recursively derived from the gyroscope, starting from the previous frame's effective absolute position (derived from parking space occlusion detection positioning or Bluetooth positioning). Pure inertial estimation results in a positioning drift of approximately 5-8 meters within 30 seconds. When the user moves to an area covered by Bluetooth beacons or cameras, the system automatically uses the absolute positioning results to perform a one-time correction for the accumulated drift.

[0078] Kalman filter fusion strategy: The multi-source fusion positioning module performs Kalman filter time updates (based on IMU state prediction) at a frequency of 20Hz, and performs measurement updates at the actual output frequency of each positioning data source. GPS positioning: 1Hz, as an absolute reference. Image recognition and parking space occlusion detection and localization: 1-2Hz, enabled when confidence level > 0.6; Bluetooth beacon positioning: Enabled at 5Hz, with at least 3 beacons and RSSI fluctuation variance <10dB; Image recognition and Bluetooth fusion positioning: 2-5Hz, enabled when any two positioning sources are available simultaneously; When the information (innovation) between the observation value and the Kalman filter prediction value of a certain positioning source exceeds 3 times the standard deviation, the observation value is determined to be an outlier (such as the image recognition error caused by the user being occluded, or the sudden interference of the Bluetooth signal). The filter automatically reduces the fusion weight of the observation source or temporarily removes the observation source to avoid the erroneous positioning results from polluting the fusion output.

[0079] S5: Arrival Notification and Status Update The system continuously monitors the distance between the user's location and the parking location. When the distance is less than a preset arrival threshold (such as 3 meters), it determines that the user has arrived at the target parking space. The system issues an audio prompt (a beep or voice broadcast "arrived at the target parking space") and a vibration prompt, while simultaneously displaying arrival confirmation information on the mini-program interface.

[0080] Upon arrival confirmation, the system automatically updates the parking status to "vehicle retrieved", records the retrieval timestamp, writes the status change to the local offline cache, and immediately synchronizes it to the cloud server when the network is available. After receiving the status update, the cloud server marks the parking space as "vacant", updates the parking space status data in the management backend, and releases the parking space resource.

[0081] If the user has enabled the payment function, the system will automatically calculate the parking fee based on the parking duration (retrieval time minus parking time) and the parking lot's charging standard after the arrival confirmation, generate a detailed fee list (including parking duration, unit price, discount amount, amount due, etc.), and guide the user to complete the online payment. After the payment is completed, the payment record will be synchronized to the cloud.

[0082] In this embodiment, the Bluetooth communication and offline data transmission mechanism includes: When a user enters the parking lot for the first time and the network is available, the system automatically downloads and stores the complete map data of the current parking lot, information on the passage nodes of each floor, information on the accessible floors corresponding to each passage node, the mapping relationship between QR code identifiers and parking spaces, the location coordinate map of Bluetooth broadcasting devices (including Bluetooth gate barriers, Bluetooth geomagnetic detectors, and Bluetooth beacons), the media access control address and location mapping table of each Bluetooth device, the mapping table of camera numbers and installation locations, and the internal and external parameter matrices of cameras to the local database. The map data includes vectorized floor plans of each floor, the topology of the passageway, parking space numbers and coordinates, and the coordinates and types of passage nodes.

[0083] In addition, the parking management system periodically transmits updated parking map data and real-time parking space status information via Bluetooth gateways or Bluetooth broadcast devices in the form of Bluetooth broadcast packets. The user-side mini-program module periodically scans for Bluetooth broadcast signals when offline. Upon receiving a broadcast packet, it verifies data integrity using a checksum. If the data version number is higher than the local cache version number, it decodes the data content and incrementally updates the local offline cache, ensuring the timeliness of local data even when offline.

[0084] When a user parks, the user-side mini-program module records the parking location information by scanning the parking space QR code or manually entering the parking space number. The parking location information includes the parking space number, floor, coordinates, parking timestamp, and vehicle identifier. It is stored in the local offline cache and asynchronously synchronized to the cloud server by the data synchronization and verification module after the network is restored. When the network is unavailable, all vehicle search requests, map reading, route calculation and navigation operations are completed independently based on local cached data. After the network is restored, the data synchronization and verification module automatically performs bidirectional synchronization and consistency verification between the local offline cached data and the cloud data. The consistency verification adopts a conflict resolution strategy based on timestamp and version number, with the most recently updated data as the standard. When a user enters the parking lot for the first time and the network is available, the user-side mini-program module automatically downloads and stores the complete map data of the current parking lot, the information of each floor passage node, the information of the accessible floors corresponding to each passage node, the mapping relationship between QR code labels and parking spaces, the location coordinate map of Bluetooth devices (Bluetooth gate, Bluetooth geomagnetic detector, Bluetooth beacon), and the mapping table of camera numbers and installation locations to the local database.

[0085] Specifically: A Bluetooth gateway device is deployed in the parking management system (located in the central area of ​​the parking lot or inside the low-voltage wiring shaft). This gateway connects to the backend server via a wired network. The server packages the updated parking map data (including information on newly added parking spaces, lane changes, temporary road closures, etc.) and real-time parking space status information (the vacancy / occupancy status of each parking space, the identifier of the occupying vehicle, etc.) into JSON format and periodically sends it through the Bluetooth gateway using the iBeacon or Eddystone broadcast protocol. The broadcast packet structure is as follows: The client-side mini-program module continuously runs a Bluetooth scanning service in the background (low power mode, scan interval of 2 seconds). Upon receiving a Bluetooth broadcast packet, it first extracts the parking lot ID. If it matches the current parking lot, it extracts the CRC32 checksum to verify data integrity. After successful verification, it extracts the data version number. If the version number is higher than the version number recorded in the local cache, it decodes the data content and performs an incremental update of the local offline cache.

[0086] For example, when a parking space changes from "vacant" to "occupied," the parking management system sends a data packet containing the parking space number and the new status via a Bluetooth gateway. When a user enters the parking lot offline, their app can obtain the latest parking space status information after scanning the Bluetooth broadcast signal, providing real-time parking space distribution information without relying on a mobile network. Actual testing shows that a single broadcast packet is approximately 200-500 bytes in size and can be transmitted and decoded within 2-3 seconds, meeting the real-time requirements for parking space status information.

[0087] This embodiment also includes a dual-fault-tolerant parking recording mechanism, specifically including: (1) QR Code Recording Mode: In normal scenarios, after parking, users scan the QR code on the parking space using a mini-program. The QR code uses QR code encoding and contains the parking lot's unique identifier, parking space number, floor information, coordinate information (longitude and latitude), and a verification code. After scanning, the mini-program parses the QR code content, verifies the data integrity through the verification code, binds the parking space information to the user's account, and records the parking location. This mode is the default recommended method, as it is the most convenient to operate and provides the most accurate information.

[0088] (2) Manual Entry Mode: When the QR code is damaged, soiled, or missing, or when the user's mobile phone camera malfunctions or insufficient light causes scanning failure, the user can switch to manual entry mode. The user enters the parking space number (such as "B2-203" or "203") in the mini-program interface. The system matches and queries the current parking lot ID and the entered number in the local cache or cloud map data to obtain the floor information and coordinate information corresponding to the parking space number. If the match is successful, the parking location is recorded; if the match fails (such as the number format is incorrect or does not exist), the user is prompted to re-enter or a similar number is provided for selection.

[0089] (3) Fuzzy Query Mode: When a user parks without scanning a code or manually entering information (e.g., leaving in a hurry and forgetting to do so), the system cannot directly obtain the parking location when initiating a car search later. In this case, the fuzzy query mode is triggered. The user enters the license plate number (complete or partial) and / or the parking time period and / or the area where the parking space is located (e.g., "B2 area") in the mini-program interface. The system performs a fuzzy matching query based on the local cache or cloud database—partial matching of the license plate number, range filtering of the parking time, and string matching of the area. The query results are presented as a list of candidate parking locations, and the user selects and confirms the correct parking location based on memory. The system records the matching result and optimizes the sorting algorithm for subsequent fuzzy queries.

[0090] The three modes are prioritized in descending order: QR code recording > manual entry > fuzzy search. Users can choose flexibly according to the actual situation to ensure that parking location recording can be completed under any circumstances, achieving full-scene coverage.

[0091] In this embodiment, the planning of cross-floor paths includes: (1) Classification and attribute modeling of passageway nodes: Cross-floor passageway nodes in the parking lot are divided into three categories: Vertical elevators: mechanically driven, requiring no physical exertion from the user, but involving waiting time. The cost of vertical movement is modeled as a distance value converted from a fixed waiting time. in, The average walking speed of users (taken as 1.2 to 1.5 m / s). The average elevator waiting time (10-30 seconds depending on the size of the parking lot) reflects the time cost for users waiting for the elevator.

[0092] Escalators: Continuously operating mechanical ramps that allow users to switch floors simply by standing on them, requiring no physical exertion and requiring no waiting time (or a negligible waiting time). The cost of vertical movement is: .

[0093] Fire escape stairs: Users must walk up and down the stairs, resulting in significant physical exertion and time costs. The cost of vertical movement is: The meanings and values ​​of each parameter are as follows: The structural floor height of a single floor is 3.0m to 3.9m according to the "Code for Design of Parking Garages" JGJ100-2015. This invention takes the measured value or the default value of 3.6m.

[0094] The angle of inclination for fire escape staircases is determined according to the "Code for Fire Protection Design of Buildings" GB50016-2014, which specifies an inclination angle range of 25° to 45°, with a common design value of 30°. This invention uses the measured value or the default value of 30°. The actual walking ramp length of a single-story staircase is .

[0095] This represents the number of floors that need to be traversed on foot. For example, if it requires crossing 3 floors to get from the 1st floor to the 4th floor, then... =3.

[0096] This is a weighting factor used to quantify the differences in physical exertion between floors in different directions. When going uphill, users need to overcome gravity, resulting in greater physical exertion, so this factor is used. =2; When going downhill, the user uses gravity, so the physical exertion is relatively small, so we take... =1. This weighting factor can be dynamically adjusted based on actual parking lot operation data (such as through user surveys or historical navigation data optimization).

[0097] (2) Calculation of total path cost and selection of optimal path: For each candidate cross-floor path, the total path cost is the sum of the costs of each segment: in: The horizontal travel distance from the starting point to the first passage node within the starting floor (calculated by searching the floor plan path network using the A-star algorithm). The cost of vertical movement between floors in the i-th segment (calculated according to the above model based on the channel type); The horizontal travel distance within an intermediate floor from one passageway exit to the next passageway entrance; The horizontal travel distance from the passageway exit to the parking space within the final floor; Calculate all feasible candidate paths Then, the path corresponding to the minimum value is taken as the optimal cross-floor car-finding path.

[0098] (3) Transfer Floor Processing: The transfer floor processing is a further path completion step based on the second type of channel node screening results mentioned above. That is, after determining the channel nodes that can reach intermediate floors, the system continues to search for transfer channel nodes that can reach the destination floor within the intermediate floors, thereby forming a complete cross-floor path. When there are no channel nodes that can directly reach the destination floor from the starting floor, the system automatically searches for transfer floors. Specifically, for each channel node of the starting floor, its list of reachable floors is checked. If there is a floor M in the list that is between the starting floor and the destination floor, then floor M is taken as a transfer floor. In the transfer floor M, transfer channel nodes that can reach the destination floor are searched, and the horizontal movement cost within the transfer floor from the exit of the channel node of the starting floor to the transfer channel node is calculated. The transfer floor is included in the total path cost calculation. If multiple transfer floors are feasible, the total cost of each scheme is calculated separately, and the optimal one is selected.

[0099] In this embodiment, an intelligent positioning function for entry is also included, specifically: (1) When a user drives into the parking lot, they can scan the QR code at the entrance or elevator hall, or click the “Enter and Find Your Parking Space” button on the mini-program interface to trigger the function.

[0100] (2) The system obtains the user's current location (entrance or scanning point location) and requests the real-time status data of available parking spaces in the current parking lot from the cloud server. If the network is unavailable, the system will prioritize reading the available parking space status information from the parking space status update data received via Bluetooth broadcast; if the Bluetooth broadcast data is unavailable, the system will use the most recently updated available parking space status data in the local cache (including the data update timestamp for the user to refer to the data freshness).

[0101] (3) The system iterates through all parking spaces in the state of "idle". Starting from the user's current location and ending at each available parking space, it performs an A-star path search on the current floor plan access path network and calculates the length of the access path from the starting point to each available parking space.

[0102] (4) Sort all available parking spaces in ascending order of entrance path length, select the available parking space with the shortest path length as the recommended parking space, generate an entrance navigation path from the starting point to the parking space, and present it to the user in the form of map highlighting and voice guidance.

[0103] (5) After the user arrives at the recommended parking space along the navigation path and parks, the system automatically recognizes the location arrival event (by determining that the distance between the user's location and the recommended parking space is less than a preset threshold), automatically binds the parking space to the user's account, records the parking location (including parking space number, floor, coordinates, and parking timestamp), and updates the status of the parking space to "occupied" and synchronizes it to the cloud server.

[0104] (6) If the user does not select the recommended parking space and parks in another parking space, the system will detect that the user's location is inconsistent with the recommended parking space through continuous positioning. When the system detects that the time spent in the parking space exceeds the preset threshold (e.g., 2 minutes), it will automatically trigger location re-identification and update the parking space closest to the user's actual location to the bound parking space.

[0105] (7) If the distance between the user's parking position and the recommended parking space is less than the first preset threshold but greater than the second preset threshold, the multi-source fusion positioning module starts the precise positioning confirmation process. It uses the fine-grained displacement detection of the inertial measurement unit to determine whether the user has entered the recommended parking space. If it is determined that the user has entered an adjacent parking space, the nearest parking space is matched and bound according to the user's final parking position.

[0106] In this embodiment, the method for constructing the parking lot map data includes: (1) Based on the parking lot building plan or on-site measured data, construct the access path network for each floor. The access path consists of path segments and nodes. Each path segment is a straight line segment connecting two endpoints. The path segment is located in the center of the access passage and is consistent with the direction of the passage. Adjacent path segments are connected end to end and share endpoints. For two intersecting path segments, a new node is generated at the intersection point, dividing the original two path segments into four path segments to ensure the complete connectivity of the path network.

[0107] (2) The node coordinates are generated using the Mercator projection coordinate system or the universal transverse Mercator projection coordinate system to ensure the absolute positioning accuracy of the coordinate data and the spatial consistency between different floors.

[0108] (3) The passage nodes between each floor are recorded as cross-floor connection points in the path network. Each passage node contains the following attribute fields: node unique identifier (UUID), node type (elevator / escalator / fire escape), node coordinates (x, y), current floor (floor), and list of reachable floors (...). (Stored as a JSON array) and channel node names (such as "Elevator No. 1").

[0109] (4) Each parking space is recorded as an independent node in the path network or associated with the nearest path node. The parking space information includes: unique parking space identifier, parking space number (e.g., "B2-203"), floor, parking space coordinates (x, y), and associated QR code identifier ( Parking space status (vacant / occupied).

[0110] (5) Map data is packaged and stored in a structured file format (GeoJSON or custom binary format), deployed on a cloud server for users to download, and can also be stored in a local offline cache with the same data structure after download.

[0111] It should be noted that the construction of the parking lot map data is an offline preprocessing step completed during the system deployment phase. It is a prerequisite for the initialization of local offline cache data in step S1. The completed parking lot map data is deployed on the cloud server in a structured file format for users to download and store in the local offline cache when they first enter the parking lot.

[0112] This embodiment also includes a data synchronization and consistency guarantee mechanism, specifically: (1) Version number management: Each data table maintains a global version number, which is incremented each time data changes. The local cache records the currently synchronized version number, and the cloud records the latest version number. During synchronization, the version numbers are compared. If the local version number is lower than the cloud version number, the incrementally changed data is retrieved.

[0113] (2) Operation log recording: Each user operation (parking record, status update, etc.) generates an operation log, which includes fields such as operation type, operation object, operation timestamp, and operation result. The operation log is stored in local cache and cloud, and is compared and merged based on the operation log as the basic unit during synchronization.

[0114] (3) Conflict resolution strategy: When the same data is modified simultaneously on both the local machine and the cloud (e.g., user A modifies the parking space status while offline, and user B modifies the same parking space status online), the "last writer wins" strategy is adopted, that is, the timestamps of the two modification records are compared, and the modification result with the newer timestamp is retained. If the timestamps are the same (the probability is extremely low), the user priority is compared or one is randomly selected, and the conflict resolution process is recorded in the synchronization log.

[0115] (4) Synchronization failure retry mechanism: If the data synchronization fails due to network fluctuations or other reasons, the system will automatically enter the retry queue and use an exponential backoff strategy (retry intervals of 5 seconds, 10 seconds, 20 seconds, 40 seconds... with a maximum interval of 5 minutes) to retry until synchronization is successful or the user actively cancels.

[0116] The specific execution process of the bidirectional synchronization is as follows: (1) First, perform data retrieval from cloud to local: The system requests data records that have changed since the last synchronization from the cloud, updates the changed data to the local cache, and ensures that the local data is consistent with the latest state of the cloud; (2) Then execute the local → cloud data upload: the system traverses the operation records marked as 'to be synchronized' in the local cache and uploads them to the cloud one by one; (3) When the same record is modified both locally and in the cloud, the conflict resolution strategy is triggered, and the final version is determined according to the 'last writer wins' rule; (4) After synchronization is complete, update the version numbers on both the local and cloud platforms to ensure that the version numbers on both sides are consistent, and mark all records as 'synchronized'. This process ensures that the synchronization direction is clear, conflicts are resolvable, and eventual consistency is achieved.

[0117] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A parking lot intelligent navigation and vehicle-finding system based on offline caching and multi-source fusion positioning, characterized in that, include: The user-side mini-program module provides functions such as scanning / manually binding parking spaces, offline car finding, indoor navigation, finding parking spaces upon entry, and payment interaction. It has a built-in offline caching unit to pre-store parking lot map data when the network is available. The multi-source fusion positioning module is used to achieve seamless indoor and outdoor positioning through GPS, IMU, camera image recognition and parking space occlusion detection, Bluetooth beacon positioning, and Kalman filtering. When GPS is lost, it switches to indoor mode, which includes image recognition positioning, Bluetooth beacon trilateration positioning, image and Bluetooth weighted fusion positioning, and pure inertial estimation mode, and uses Kalman filtering to periodically correct inertial drift. The map and path planning module is used to perform same-floor and cross-floor path planning based on the parking lot map data and the improved A-star algorithm. When crossing floors, the vertical movement cost is calculated differently according to the elevator, escalator and fire escape types. The backend service and management backend module is used for cloud-based management of maps, parking spaces, QR codes, orders, and pricing standards, and provides data statistical analysis. The data synchronization and verification module is used to automatically synchronize local offline cached data with cloud data bidirectionally when the network is restored, and to verify data consistency through conflict resolution strategies based on timestamps and version numbers.

2. A parking lot intelligent navigation and vehicle locating method based on offline caching and multi-source fusion positioning, applied to the parking lot intelligent navigation and vehicle locating system based on offline caching and multi-source fusion positioning as described in claim 1, characterized in that, Includes the following steps: S1: The user-side mini-program module obtains the user's car-finding request, detects the current network status, and determines the user's current location and parking location. If the network is available, the system retrieves parking lot map data from the cloud server, reads the user's latest parking location information from the cloud database, and obtains the user's current location information via GPS or base station positioning. If the network is unavailable, the user-side mini-program module will first attempt to receive map update data and parking space status information sent by the Bluetooth broadcasting device in the parking lot via Bluetooth communication, perform incremental updates to the local offline cache, and then read the parking lot map data and parking location information from the local offline cache. Among them, the user's current indoor location is estimated by using the multi-source fusion positioning module with image recognition and parking space occlusion detection positioning mode, Bluetooth beacon positioning mode, or image recognition and Bluetooth fusion positioning mode. S2: The map and route planning module determines whether cross-floor car searching is involved based on the user's current location and parking location. If the vehicle is located across multiple floors, the cross-floor vehicle search strategy will be used; otherwise, the same-floor vehicle search strategy will be used. The cross-floor vehicle search strategy calculates the optimal cross-floor path based on the information of each floor's passageway nodes, the vertical movement cost corresponding to the passageway node type, and offline map data. The same-floor vehicle search strategy calculates the optimal same-floor path based on the improved A-star algorithm. S3: The map and path planning module, based on the selected car-finding strategy and combined with parking lot map data, current location and parking location, uses a path planning algorithm to generate at least one candidate car-finding path from the current location to the parking location. The final car-finding path is determined based on the total movement cost of each candidate path, where the total movement cost is obtained by summing the horizontal movement cost and the vertical movement cost. S4: During navigation, the multi-source fusion positioning module collects acceleration, gyroscope and geomagnetic data from the mobile phone's built-in inertial measurement unit in real time, and fuses the multi-source data through the Kalman filter algorithm to dynamically correct the indoor positioning data. The map and path planning module then matches the corrected location information with the final car-finding path on the map. The user-end applet module updates the user's location in real time and guides the user to the parking location along the final car-finding path. S5: Upon arrival at the parking location, the user-side mini-program module sends a notification signal, automatically updates the parking status, and synchronizes it to the cloud server and local offline cache.

3. The parking lot intelligent navigation and vehicle finding method based on offline caching and multi-source fusion positioning according to claim 2, characterized in that, The indoor positioning mode of the multi-source fusion positioning module after GPS signal loss includes: (a) Image recognition and parking space occupancy detection and positioning mode: The monitoring camera is used to identify the occupancy status of the parking space. The pre-calibrated coordinates of the occupied parking space are used as the positioning anchor point. At the same time, the user image information is mapped to the physical coordinate system, and the position estimate is output after multi-frame tracking. (ii) Bluetooth beacon positioning mode: Scan the Bluetooth broadcasting devices in the parking lot to obtain the signal strength and address, and calculate the user's location by combining the pre-stored Bluetooth coordinate map with trilateration. (III) Image recognition and Bluetooth fusion positioning mode: The image recognition positioning result and the Bluetooth positioning result are weighted and fused according to dynamic weights and output; (iv) Pure inertial estimation mode: When none of the above positioning modes are available, dead reckoning is performed based on inertial measurement unit data, and the system automatically switches and corrects the accumulated drift when the external positioning source is detected to be available again. The multi-source fusion positioning module uses the Kalman filter algorithm to take one or more of the above positioning results as observations, periodically corrects the cumulative drift calculated by inertial estimation, and uses GPS observations to correct errors after the GPS signal is recovered.

4. The parking lot intelligent navigation and vehicle finding method based on offline caching and multi-source fusion positioning according to claim 2, characterized in that, The Bluetooth communication and offline data transmission mechanism includes: The parking management system periodically transmits updated parking map data and real-time parking space status information in the form of Bluetooth broadcast packets via Bluetooth gateways or Bluetooth broadcast devices. The broadcast packet format includes data packet type identifier, data version number, data content, and check code. The user-side mini-program module periodically scans for Bluetooth broadcast signals when offline. After receiving a broadcast packet, it verifies the data integrity through a checksum. If the data version number is higher than the local cache version number, it decodes the data content and incrementally updates the local offline cache. When a user parks, the user-side mini-program module records the parking location information by scanning the parking space QR code or manually entering the parking space number. The parking location information includes the parking space number, floor, coordinates, parking timestamp, and vehicle identifier. It is stored in the local offline cache and asynchronously synchronized to the cloud server by the data synchronization and verification module after the network is restored. When the network is unavailable, all vehicle search requests, map readings, route calculations, and navigation operations are completed independently based on locally cached data. Once the network is restored, the data synchronization and verification module automatically performs bidirectional synchronization and consistency verification between the local offline cached data and the cloud data. The consistency verification adopts a conflict resolution strategy based on timestamps and version numbers, with the most recently updated data taking precedence.

5. A parking lot intelligent navigation and vehicle finding method based on offline caching and multi-source fusion positioning according to claim 2, characterized in that, The cross-floor car-finding strategy described in step S2 specifically includes: Determine the starting floor and the coordinates of the starting position within the starting floor based on the user's current location, and determine the ending floor and the coordinates of the ending position within the ending floor based on the parking location; Read all passage nodes in the starting floor map data, and obtain the node type, node location coordinates and reachable floor information list for each passage node. The node types include elevators, escalators and fire escape stairs. Select candidate channel nodes from all channel nodes that meet one of the following conditions: The system will perform transit floor processing in subsequent steps for the second type of channel node, which is either a first type of channel node that contains the destination floor in the reachable floor information list, or a second type of channel node that contains at least one intermediate floor between the starting floor and the destination floor in the reachable floor information list. The total path cost of all candidate channel nodes is compared after the transfer floor processing is completed: the total path cost of the first type of channel nodes is calculated immediately after node screening and temporarily stored in the candidate list; the total path cost of the second type of channel nodes, including the horizontal movement cost of the transfer floor, is calculated after the transfer floor processing is performed to determine the complete transfer path; after the total path cost of all candidate nodes is calculated, they are sorted and the path corresponding to the candidate channel node with the smallest total path cost is selected as the cross-floor vehicle search path.

6. The parking lot intelligent navigation and vehicle finding method based on offline caching and multi-source fusion positioning according to claim 5, characterized in that, The horizontal movement cost is the actual travel path length between the two points, which is obtained by performing the A-star algorithm on the floor plan; The vertical movement cost is determined based on the type of passageway node: when the passageway node type is a vertical elevator, the vertical movement cost is a preset distance value calculated from the elevator waiting time; when the passageway node type is an escalator, the vertical movement cost is zero; when the passageway node type is a fire escape staircase, the vertical movement cost is calculated based on the floor height, the staircase inclination angle, and the number of floors that need to be traversed, using the following formula: in: For the cost of vertical movement, The height of a single floor. The angle of inclination of the fire escape staircase. This represents the number of floors that need to be walked through. As a weighting factor, during upward movement =2, downlink =1; When there is no direct passage node to the destination floor, the cross-floor vehicle search strategy further includes: determining intermediate floors, finding transfer passage nodes in the intermediate floors that can reach the destination floor, calculating the horizontal movement cost from the exit of the candidate passage node at the starting floor to the transfer passage node at the intermediate floor, and including it in the total path cost calculation.

7. A parking lot intelligent navigation and vehicle finding method based on offline caching and multi-source fusion positioning according to claim 2, characterized in that, It also includes a dual-fault-tolerant parking record mechanism, specifically including: The first recording method: After parking, the user scans the QR code on the parking space through the user terminal mini program module. The system parses the parking space number and parking lot sign encoded in the QR code, binds the obtained parking space number to the current user account, and records the parking location. The second recording method: When the QR code scan fails or the QR code is missing, the user can manually enter the parking space number through the user terminal mini program module interface. The system will match the entered parking space number in the local cache or cloud map data to find the corresponding parking space coordinates and complete the parking location recording. The third recording method: When a user neither scans the code nor manually enters information when parking, after the user initiates a car-finding request, they can enter the license plate number, parking time range, or parking space area information in the user-side mini-program module interface. The system will then perform fuzzy matching through local cache or cloud database and return a list of matching parking location candidates for the user to select and confirm. The fuzzy matching supports partial matching of license plate numbers and filtering of time ranges.

8. A parking lot intelligent navigation and vehicle finding method based on offline caching and multi-source fusion positioning according to claim 2, characterized in that, It also includes an intelligent positioning function for entry: Users can scan the QR code at the parking lot entrance or elevator lobby, or trigger an entry and parking space request through the user-end mini-program module interface. The map and route planning module obtains the real-time status information of all available parking spaces in the current parking lot. The real-time status information is obtained from the cloud server or read from the parking space status update data received via Bluetooth broadcast in offline mode. The map and route planning module uses a route planning algorithm to calculate the candidate entry routes from the user's current location to each available parking space and their lengths, based on the user's current location, the coordinates of each available parking space, and the topology of the traffic paths within the parking lot. The available parking spaces are sorted in ascending order of path length. The available parking space with the shortest path length is selected as the recommended parking space. An entry navigation path from the user's current location to the recommended parking space is generated and presented to the user. Once a user arrives at the recommended parking space, the user-side mini-program module automatically binds the parking space to the user's account, records the parking location, and stores it in the local cache and cloud server. If a user does not select a recommended parking space and parks in another parking space, the multi-source fusion positioning module will detect that the user's location is inconsistent with the recommended parking space through continuous positioning. When the detection of the dwell time exceeds a preset threshold, the location re-identification will be automatically triggered, and the parking space closest to the user's actual location will be updated to the bound parking location. If the distance between the user's parking location and the recommended parking space is less than the first preset threshold but greater than the second preset threshold, the multi-source fusion positioning module initiates a precise positioning confirmation process. It uses fine-grained displacement detection by the inertial measurement unit to determine whether the user has entered the recommended parking space. If it is determined that the user has entered an adjacent parking space, it matches and binds the nearest parking space based on the user's final parking location.

9. A parking lot intelligent navigation and vehicle finding method based on offline caching and multi-source fusion positioning according to claim 2, characterized in that, The specific methods for map matching described in step S4 include: The user's location coordinates after Kalman filtering correction are projected onto the traffic path network of the parking lot map, and the vertical distance between the projected point and each path node on the final car-finding path is calculated. If the vertical distance is less than a preset matching threshold, the user's position will be corrected to the nearest node on the final vehicle search path to achieve position correction. If the vertical distance is greater than the preset matching threshold, it is determined that the user has deviated from the final vehicle search path, and the deviation confirmation timer is started first. If the user returns to the vicinity of the path within the preset time and the vertical distance is less than the matching threshold again, the deviation prompt will be canceled and navigation will continue along the original path. If the user remains in a deviated state for a preset time, the user-side mini-program module will prompt the user with voice and text that they have deviated from the planned path and ask whether they want to replan an alternative path. If the user confirms the replanning, or if the user continues to walk beyond the preset distance in the deviated direction without confirmation, the system will automatically replan an alternative car-finding route from the current user location to the parking location and update the navigation guidance. During navigation, the final vehicle-finding route is represented by a sequence of path segments. The user's location is updated in real time and moves along the sequence of path segments. When the user's location crosses the current target path node, it automatically switches to the next path segment and guides the user to turn at path turning points through voice and text prompts.

10. A parking lot intelligent navigation and vehicle finding method based on offline caching and multi-source fusion positioning according to claim 2, characterized in that, Prior to step S1, a pre-construction step for parking lot map data is also included, specifically including: Based on the actual layout of the parking lot, the passage areas of each floor are mapped into a passage path network. The passage path network consists of path segments and nodes. Each path segment is a straight line segment connecting two nodes. The path segment is located in the center of the passage channel. The endpoint coordinates of the path segment are generated using the Mercator coordinate system. For two intersecting path segments, a new node is generated at the intersection point, dividing the original path segment into four path segments. The passage nodes between floors are recorded as cross-floor connection nodes. Each passage node records its node type, node coordinates, current floor, and a list of accessible floors. Complete the parameter calibration of the surveillance cameras in the parking lot, and record the installation location, shooting coverage area, internal and external parameter matrix and corresponding parking space number range of each camera; Complete the parameter calibration of the Bluetooth broadcasting equipment in the parking lot, and record the device type, installation location coordinates, media access control address and broadcast signal coverage of each Bluetooth device. The Bluetooth devices include Bluetooth barrier gates, Bluetooth geomagnetic detectors and Bluetooth beacons. Parking map data is stored in a structured file format and deployed on a cloud server, and can also be downloaded and stored in a local offline cache with the same data structure.

Images