A parking lot tidal parking space prediction and guidance system and method

Abstract

The present application relates to the technical field of intelligent parking management system, and particularly relates to a parking lot tidal parking space prediction and guidance system and method. The method comprises the following steps: acquiring vehicle flow video data of a parking lot entrance, and performing vehicle attribute feature analysis to obtain a real-time vehicle flow vector; performing demand trend prediction on the real-time vehicle flow vector to obtain an expected demand total amount; acquiring and extracting an individual vehicle departure probability table from a current parking space resource state diagram; calculating a type imbalance index according to the expected demand total amount; dynamically generating a reorganization strategy according to the type imbalance index to obtain a space conversion instruction set; performing space identification and boundary reorganization of the parking space according to the space conversion instruction set to obtain an updated parking space resource state diagram, and performing guidance mode decision and parking space allocation to obtain a locked parking space voucher. The present application realizes the maximization of parking lot space resources and time efficiency through a closed-loop control integrating prediction, reorganization and collaborative guidance.

Description

Technical Field

[0001] This invention relates to the field of intelligent parking management system technology, and in particular to a parking lot tidal parking space prediction and guidance system and method. Background Technology

[0002] Traditional parking lots are limited by fixed lane markings, making it difficult to change parking spaces once designated. This makes them unsuitable for handling the fluctuating demand during weekday rush hours and weekend commercial areas, resulting in congestion in some areas while leaving many spaces vacant and wasting valuable space. Existing technologies lack precise supply-demand balancing mechanisms for specific types of parking spaces, particularly for charging stations and shared spaces. Relying solely on static allocation leads to high demand during peak hours and low availability during off-peak periods. Furthermore, they fail to proactively manage parking spaces based on usage duration and availability probability, resulting in inefficient allocation. Parking guidance systems are also inadequate when handling multiple vehicles searching for spaces concurrently. Traditional systems only provide information on the number of available spaces and lack the ability to predict path conflicts. During peak hours, a "swarm effect" often occurs when multiple vehicles enter simultaneously. In addition, weak GPS signals in underground parking lots prevent accurate positioning, leading to frequent vehicle turns, intersection congestion, and increased safety risks.

[0003] In summary, existing technologies suffer from problems such as static resource fixation, imbalance between supply and demand for different types of vehicles, and poor coordination in multi-vehicle guidance, which urgently need to be addressed. Summary of the Invention

[0004] Therefore, it is necessary to provide a parking lot tidal parking space prediction and guidance system and method to solve at least one of the above-mentioned technical problems.

[0005] To achieve the above objectives, a method for predicting and guiding tidal parking spaces in a parking lot includes the following steps:

[0006] Step S1: Obtain video data of traffic flow at the parking lot entrance and perform vehicle attribute feature analysis to obtain real-time traffic flow vectors;

[0007] Step S2: Predict demand trends from real-time traffic flow vectors to obtain the expected total demand; obtain and extract individual vehicle departure probability tables from the current parking space resource status map; calculate the type imbalance index based on the expected total demand.

[0008] Step S3: Dynamically generate a reorganization strategy based on the type imbalance index to obtain a space conversion instruction set; reorganize the spatial identifiers and boundaries of parking spaces based on the space conversion instruction set to obtain an updated parking space resource status map;

[0009] Step S4: Based on the individual vehicle departure probability table and the updated parking space resource status map, make guidance mode decisions and allocate parking spaces to obtain locked parking space certificates, which include regular locked parking space certificates and virtual locked parking space certificates.

[0010] Step S5: Use the locked berth credentials to perform preliminary route planning and generate anchor point sequences to obtain anchor point navigation sequences; based on the anchor point navigation sequences, use cameras to track vehicles at anchor points to obtain real-time vehicle positions and status tables; perform route conflict detection and coordination based on the real-time vehicle positions and status tables to obtain a collaboratively optimized route set; distribute multi-channel guidance beacons based on the collaboratively optimized route set and the locked berth credentials to obtain personalized berth beacons.

[0011] This invention captures high-quality vehicle images using virtual line triggering and sharpness assessment technology. It employs a three-way parallel processing architecture to significantly improve feature extraction efficiency, transforming raw visual information into standardized traffic flow vectors with multi-dimensional attributes, thus fundamentally solving the problem of ambiguous perception of traffic flow types in traditional systems. It innovatively integrates real-time traffic flow with historical baselines through dynamic weighted fusion, achieving accurate demand prediction. Furthermore, it introduces the calculation of the probability of vehicles leaving the parking lot, transforming static parking space inventory into dynamic prediction of available resources. By quantifying the supply-demand relationship through a type imbalance index, it provides bidirectional dynamic data support for spatial reorganization.

[0012] A systematic strategy generation mechanism transforms abstract supply and demand gaps into concrete and executable reconfiguration solutions. By controlling programmable LED markings on the ground and retractable isolation devices, it enables the dynamic reconfiguration of physical parking spaces, fundamentally resolving the tidal supply and demand contradictions caused by the fixed layout of traditional parking lots. The state pre-locking and map solidification update process ensures real-time synchronization and data consistency between the digital map and physical reality, greatly improving the utilization rate of spatial resources.

[0013] The introduction of a dual-mode decision-making mechanism combining conventional guidance and virtual queuing significantly optimizes parking space allocation efficiency. The unique virtual queuing function utilizes high-confidence departure prediction to transform uncertain waiting times into high-probability parking opportunities, effectively reducing ineffective cruising and congestion caused by vehicles without available spaces. Vehicle tracking technology based on camera anchor points solves the challenge of accurate positioning in environments without GPS signals. Kalman filter position prediction and multi-vehicle spatiotemporal trajectory conflict detection achieve a shift from "passive response" to "active prediction." Through dynamic priority assessment and path coordination, it effectively avoids congestion and the "herd effect" at intersections within the parking lot.

[0014] In summary, this invention, through a closed-loop control method integrating prediction, reorganization, and collaborative guidance, utilizes a two-way dynamic supply and demand prediction model, flexible spatial reorganization, and collaborative scheduling based on anchor point calibration to systematically solve the technical problems faced by traditional parking lots, such as static resource fixation, type-based supply and demand imbalance, and poor multi-vehicle guidance coordination, thereby maximizing the spatial resources and time efficiency of parking lots.

[0015] Preferably, the present invention also provides a parking lot tidal parking space prediction and guidance system for executing the parking lot tidal parking space prediction and guidance method described above, the parking lot tidal parking space prediction and guidance system comprising:

[0016] The vehicle flow profile acquisition module is used to acquire video data of vehicle flow at the parking lot entrance and perform vehicle attribute feature analysis to obtain real-time vehicle flow vectors.

[0017] The demand pressure forecasting module is used to predict demand trends from real-time traffic flow vectors to obtain the expected total demand; obtain and extract individual vehicle departure probability tables from the current parking space resource status map; and calculate the type imbalance index based on the expected total demand.

[0018] The spatial flexible reorganization module is used to dynamically generate reorganization strategies based on the type imbalance index, resulting in a spatial conversion instruction set; and to reorganize the spatial identifiers and boundaries of parking spaces based on the spatial conversion instruction set, resulting in an updated parking space resource status map.

[0019] The intelligent parking space allocation module is used to make guidance mode decisions and allocate parking spaces based on the individual vehicle departure probability table and the updated parking space resource status map, and obtain the locked parking space certificate, which includes the regular locked parking space certificate and the virtual locked parking space certificate.

[0020] The multi-vehicle collaborative guidance module is used to perform preliminary route planning and anchor point sequence generation using locked berth credentials to obtain an anchor point navigation sequence; based on the anchor point navigation sequence, it tracks anchor point vehicles through cameras to obtain a real-time vehicle position and status table; based on the real-time vehicle position and status table, it performs route conflict detection and coordination to obtain a collaboratively optimized route set; based on the collaboratively optimized route set and locked berth credentials, it distributes guidance beacons through multiple channels to obtain personalized berth beacons.

[0021] The parking lot tidal parking space prediction and guidance system of this invention achieves closed-loop intelligent optimization of parking lot resource management through the coordinated operation of its five major modules, resulting in significant benefits. This system fundamentally solves the problem of tidal resource mismatch caused by the fixed layout of traditional parking lots through precise perception by the vehicle flow profile acquisition module, forward-looking two-way supply and demand analysis by the demand pressure prediction module, and dynamic physical space adaptation by the spatial flexible reorganization module. Furthermore, the innovative virtual queuing mechanism of the intelligent parking space allocation module, combined with the conflict-free path planning of the multi-vehicle collaborative guidance module, not only effectively utilizes resources to be released when parking spaces are saturated, but also transforms disordered individual car-finding behavior within the parking lot into a globally coordinated and orderly traffic flow. Ultimately, this maximizes the utilization of both parking lot space and time resources, significantly improving operational efficiency, traffic safety, and user parking experience. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating the steps of a parking lot tidal parking space prediction and guidance method. Detailed Implementation

[0023] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0024] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0025] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0026] To achieve the above objectives, please refer to Figure 1This invention provides a method for predicting and guiding tidal parking spaces in a parking lot, comprising the following steps:

[0027] Step S1: Obtain video data of traffic flow at the parking lot entrance and perform vehicle attribute feature analysis to obtain real-time traffic flow vectors;

[0028] In this embodiment of the invention, high-definition images of vehicles crossing the line at the parking lot entrance are captured using a preset virtual trigger line. The clearest frame is then selected as the original image of the vehicle to be identified using a Laplacian operator. This original image is distributed to three parallel processing units, which respectively perform license plate and character recognition through a convolutional neural network, search for new energy and shared vehicle identifiers through color thresholding and template matching, and estimate the physical dimensions of the vehicle body through edge detection and reference object calibration, thereby generating a discrete feature set. Finally, the system summarizes these discrete features into standardized vehicle type and size attributes based on a fixed decision tree rule base, and encapsulates them together with the capture timestamp and license plate number into a real-time traffic flow vector in JSON format.

[0029] Step S2: Predict demand trends from real-time traffic flow vectors to obtain the expected total demand; obtain and extract individual vehicle departure probability tables from the current parking space resource status map; calculate the type imbalance index based on the expected total demand.

[0030] In this embodiment of the invention, the real-time traffic flow vectors of the most recent 5 minutes are aggregated, and the number of vehicles entering the parking lot for each type is statistically analyzed to form the immediate demand intensity. Subsequently, the system extracts the average entry rate of the same period over the past four weeks from the historical database as the historical baseline for the same period's traffic flow. By weighting the immediate intensity and the historical baseline with 0.7 and 0.3 respectively, and multiplying by a time period coefficient, the expected total demand for each type of vehicle in the next 30 minutes is predicted. Simultaneously, the system calculates the dwell time for each vehicle in the parking lot, and uses a Sigmoid function that combines the historical average dwell time with contextual coefficients such as the current time period and vehicle type to accurately calculate an individual vehicle departure probability table, and summarizes it to obtain a parking space release prediction table. Finally, using the formula: Imbalance Index = Expected Total Demand - (Current Available Parking Spaces + Predicted Parking Space Releases), the net shortage of each type of parking space is calculated to form a type imbalance index.

[0031] Step S3: Dynamically generate a reorganization strategy based on the type imbalance index to obtain a space conversion instruction set; reorganize the spatial identifiers and boundaries of parking spaces based on the space conversion instruction set to obtain an updated parking space resource status map;

[0032] In this embodiment of the invention, parking space types are divided into demand queues and redundancy queues based on the type imbalance index. Then, continuous areas consisting of redundant types, in an available state, and possessing the physical basis for conversion are selected from the current parking space resource status diagram according to priority. The system matches these areas with a template library containing multiple conversion schemes and uses a greedy algorithm to generate a regional conversion quota table containing specific areas and conversion templates, aiming to fill gaps most efficiently. This table is ultimately formatted as a space conversion instruction set. During execution, the system first pre-locks the parking space status of the target area as "reconstruction in progress" in memory. Then, it sends instructions to the physical controller to drive the ground LED light strips to redraw parking space markings and raise physical isolation bollards. After receiving confirmation signals of successful hardware execution, the system formally deletes old parking spaces and creates new parking spaces with new attributes through a database transaction, thus solidifying the updated parking space resource status diagram.

[0033] Step S4: Based on the individual vehicle departure probability table and the updated parking space resource status map, make guidance mode decisions and allocate parking spaces to obtain locked parking space certificates, which include regular locked parking space certificates and virtual locked parking space certificates.

[0034] In this embodiment of the invention, based on the real-time traffic flow vector of newly entering vehicles, all available parking spaces with matching attributes are retrieved from the updated parking space resource status graph to form a candidate parking space list. The system decides whether to execute the "regular guidance" or "virtual queuing" mode based on whether the list is empty. In the regular mode, the system calculates and selects the parking space with the optimal path, locks its status, and reserves it for a limited time, generating a regular locked parking space certificate. In the virtual queuing mode, the system filters out opportunity parking spaces with a departure probability higher than 95% from the individual vehicle departure probability table. After obtaining the user's consent, a virtual queuing guidance certificate is generated to guide the vehicle to a waiting point near the opportunity parking space. While guiding the vehicle to the waiting point, the system monitors the status of the opportunity parking space with high priority. Once it is released, it is immediately locked and seamlessly switched to a virtual locked parking space certificate, completing the allocation.

[0035] Step S5: Use the locked berth credentials to perform preliminary route planning and generate anchor point sequences to obtain anchor point navigation sequences; based on the anchor point navigation sequences, use cameras to track vehicles at anchor points to obtain real-time vehicle positions and status tables; perform route conflict detection and coordination based on real-time vehicle positions and status tables to obtain a collaboratively optimized route set; distribute multi-channel guidance beacons based on the collaboratively optimized route set and the locked berth credentials to obtain personalized berth beacons.

[0036] In this embodiment of the invention, the A* algorithm is used to plan a preliminary path for vehicles holding locked parking space credentials, and this path is decomposed into a series of anchor point navigation sequences monitored by cameras. As the vehicle moves, each anchor point camera detects the vehicle's passage using virtual coil technology and confirms its identity using license plate recognition, thereby obtaining a series of accurate vehicle calibration position records. The system uses these discrete calibration records as observations input into a Kalman filter to continuously predict and update the vehicle's position and speed between two anchor points, forming a real-time vehicle position and status table. Based on this table, the system projects the future spatiotemporal trajectories of all vehicles on the road, detects potential path intersection conflicts, and coordinates according to preset traffic priority rules (such as distance from the target, main road priority). To avoid vehicles, a new path including deceleration instructions is planned, ultimately forming a conflict-free collaboratively optimized path set, and personalized parking space beacons are distributed to the driver.

[0037] Preferably, step S1 includes:

[0038] Acquire and capture instantaneous images of vehicles from the real-time video stream at the parking lot entrance to form the original image of the vehicle to be identified;

[0039] The original image of the vehicle to be identified is distributed to three parallel processing units for preliminary parallel feature extraction to obtain a discrete feature set.

[0040] By using a pre-defined rule base, the discrete feature set is logically summarized and labeled to obtain a standardized attribute dataset;

[0041] The standardized attribute dataset is encapsulated into traffic flow vectors to generate real-time traffic flow vectors.

[0042] In some embodiments, at the parking lot entrance, when the complete outline of a vehicle crosses a preset virtual trigger line on the ground, the system automatically triggers high frame rate capture, extracting five consecutive images from the video stream. Subsequently, the system uses the Laplacian operator to calculate the sharpness of these five images, selecting the frame with the highest operator value as the original image of the vehicle to be identified. This image is a JPEG file with a resolution of not less than 1920×1080 pixels.

[0043] The original image of the vehicle to be recognized is sent into three independent parallel processing units simultaneously. The first unit is the license plate location and character recognition unit, which quickly locates the rectangular area of the license plate through a cascaded classifier, and then uses a convolutional neural network model to perform optical character recognition on the characters within the area, outputting the license plate number. The second unit is the prominent logo search unit, which uses the color space threshold segmentation method to detect the unique gradient green of new energy vehicle license plates, and uses a template matching algorithm based on feature points to compare a database containing ten shared car platform logos in the preset front and side door areas of the vehicle. The third unit is the vehicle body contour measurement unit, which uses the Canny edge detection algorithm to outline the two-dimensional contour of the vehicle, calculates its pixel width and length, and based on the pre-calibrated ground standard width line at the entrance (such as 2.5 meters) as a reference, converts the pixel size into a physical size estimate, and finally forms a discrete feature set containing license plate text, prominent color blocks, matching logos, and contour estimates.

[0044] The system analyzes the discrete feature set according to a fixed decision tree rule library. In the type judgment branch, if the prominent color block is green, the vehicle type is marked as new energy; otherwise, if the matching logo is any of the logos in the database, it is marked as shared; otherwise, it is marked as ordinary. In the size judgment branch, if the estimated length value of the contour is less than 4.2 meters, the vehicle size is marked as compact; if the length value is between  4.2 meters and 5.0 meters, it is marked as standard; otherwise, it is marked as large. This process generates a structured standardized attribute data set, such as {type: new energy, size: standard}.

[0045] Finally, the system creates a data entry, using the system timestamp (accurate to seconds) when capturing the original image of the vehicle to be recognized as the first field, the license plate text in the discrete feature set as the vehicle unique identifier as the second field, and the complete standardized attribute data set as the third field. These three fields are jointly encapsulated into a JSON format string, constituting a real-time traffic flow vector, such as [2023-10-27 09:01:15, Beijing AXXXXX, {type: new energy, size: standard}], and this vector is immediately pushed to the subsequent processing module.

[0046] Preferably, step S2 includes:

[0047] Aggregate the recent traffic flow characteristics of the real-time traffic flow vector to obtain the immediate demand intensity;

[0048] Extract the historical same-period flow baseline from the preset historical database according to the immediate demand intensity;

[0049] Perform trend weighted fusion prediction on the immediate demand intensity and the historical same-period flow baseline to obtain the expected total demand;

[0050] Extract current available and occupied parking space data of various types from the current parking space resource status map, calculate the probability of vehicles leaving the parking lot, and obtain a parking space release prediction table and an individual vehicle departure probability table. The calculation of the probability of vehicles leaving the parking lot includes: statistically analyzing the vehicle dwell time of the occupied parking space data to obtain a vehicle dwell status table; performing categorized dwell pattern matching on the vehicle dwell status table to obtain a basic departure probability table; adjusting the contextual factor weights of the basic departure probability table to obtain an individual vehicle departure probability table; and predicting the total number of parking spaces released from the individual vehicle departure probability table to obtain a parking space release prediction table.

[0051] Based on the parking space release forecast table and the expected total demand, the net value of the supply-demand gap is calculated to obtain the type imbalance index.

[0052] In some embodiments, real-time traffic flow vector data streams within the last 5 minutes are captured, and the number of new energy vehicles, shared vehicles, standard vehicles, and compact vehicles entering the market within this time window is statistically summarized by classification to form a quantitative real-time demand intensity. Its data structure is {new energy: 5, shared vehicles: 2, standard vehicles: 12, compact vehicles: 1}.

[0053] Based on the current timestamp, for example, 9:05 AM on Tuesday, the system automatically queries a historical database containing data from the past four weeks to extract all entry records for each type of vehicle during the Tuesday morning to 9:30 AM time period. By averaging these records, the system generates a historical traffic baseline representing the general entry rate for that time period, with the data structure being {New Energy Vehicles: 3 vehicles / 5 minutes, Shared Vehicles: 1 vehicle / 5 minutes, Standard Vehicles: 15 vehicles / 5 minutes}.

[0054] A weighted average algorithm is used to fuse immediate demand intensity and historical baseline traffic for the same period to calculate the expected total demand within the next 30-minute scheduling cycle. The calculation formula is as follows:

[0055] Expected demand = (Immediate demand intensity × W_r + Historical baseline traffic volume × W_h) × (Scheduling cycle length / Statistical window length);

[0056] Where W_r is the weight of immediate demand, set to 0.7; and W_h is the weight of historical baseline, set to 0.3. Taking new energy vehicles as an example, the expected demand is (5 × 0.7 + 3 × 0.3) × (30 / 5) = 26.4, rounded to 26 vehicles. This calculation is applied to all vehicle types, ultimately generating the total expected demand: {New Energy: 26, Shared Vehicles: 9, Standard Vehicles: 81}.

[0057] The system iterates through the currently occupied parking space data, calculates the dwell time for each vehicle, and creates a vehicle dwell status table. Then, it compares the dwell time of each vehicle with the historical average dwell time corresponding to its type, and calculates a basic departure probability using a preset sigmoid function. This probability approaches 1 from 0 as the dwell time approaches the average dwell time, forming a basic departure probability table. Specifically, during system initialization, a general departure probability baseline model based on statistical analysis of large amounts of public parking lot data is loaded. After the system goes live, it continuously collects real-world data from the parking lot, uses machine learning to train and iterate a dedicated departure probability model for each vehicle type, and adaptively adjusts the key parameters of the sigmoid function.

[0058] Next, the system applies scenario adjustment coefficients to correct the base departure probability: for example, during the period from 17:00 to 19:00, the departure probability of all vehicles is multiplied by a coefficient of 1.2; for shared vehicles, the departure probability is further multiplied by a coefficient of 1.15. These coefficients are not fixed values, but are dynamically generated by performing regression analysis on historical data to quantify the impact of specific time periods, weather, and other factors on the average parking time. The corrected results constitute an individual vehicle departure probability table. To further improve accuracy, the system also introduces a "user profile adjustment coefficient." With user authorization, the system identifies whether a vehicle is a registered monthly parking vehicle or a frequent user by its license plate number, and adjusts the departure probability individually based on its inherent entry and exit habits. Finally, the system sums all individual departure probabilities for each type of vehicle to obtain a parking space release prediction table that predicts the number of parking spaces available for each type in the next 30 minutes, for example, {New Energy: 5, Shared Vehicles: 4, Standard Vehicles: 13}.

[0059] The formula for calculating the net supply-demand gap is as follows:

[0060] Imbalance Index = Total Expected Demand - (Current Available Parking Spaces + Forecasted Parking Space Release);

[0061] The system queries in real time and finds that there are currently 18 available new energy vehicle parking spaces. Taking new energy vehicles as an example, their imbalance index = 26 - (18 + 5) = +3. A positive value indicates that there is a predicted shortage of 3 parking spaces. This calculation is applied to all parking space types, and finally outputs a type imbalance index that quantifies future supply and demand pressure: {New Energy: +3, Shared Vehicles: -1, Standard Vehicles: -24}.

[0062] Preferably, step S3, which dynamically generates a reorganization strategy based on the type imbalance index, includes:

[0063] Based on the type imbalance index, the imbalance types are prioritized to obtain the parking space type conversion priority queue;

[0064] Based on the parking space type conversion priority queue, the reconfigurable areas are filtered in the current parking space resource status graph to obtain a list of candidate reconfigurable areas;

[0065] The candidate reconstruction region list is transformed and adapted for matching to obtain the region template matching results;

[0066] Based on the regional template matching results, the conversion quantity is optimized and calculated to obtain the regional conversion quota table;

[0067] The spatial conversion instruction set is obtained by formatting the spatial conversion instruction based on the regional conversion quota table.

[0068] In some embodiments, upon receiving the type imbalance index {New Energy: +8, Shared Vehicle: +2, Standard: -11, Compact: -3}, priority sorting is first performed. Positive values ​​(shortage) are arranged in descending order, and negative values ​​(redundancy) are arranged in ascending order, generating two independent queues: the demand queue is [New Energy, Shared Vehicle], and the redundancy queue is [Standard, Compact]. This is the parking space type conversion priority queue.

[0069] Based on the parking space type conversion priority queue, the system filters the current parking space resource status graph. It first targets the highest priority type "Standard" in the redundant queue, searching for contiguous areas consisting of parking spaces of this type where all spaces are currently "available." These areas must also meet an additional condition: they must have pre-installed charging interfaces serving the highest priority type "New Energy" in the demand queue. The filtering process yields a list of candidate reconfigurable areas, such as ['A01', 'C05', 'D12'], where each ID represents a physical area that meets the criteria.

[0070] For each region in the candidate reconstruction region list, an adaptation match is performed from a fixed transformation template library. This template library contains entries such as {Template ID:'3 Standard Transformer 4 Compact New Energy', Input Type:'Standard', Input Quantity: 3, Output Type:'New Energy', Output Quantity: 4}. The system detects that region 'A01' contains 6 consecutive standard parking spaces, so it can be matched twice with the '3 Standard Transformer 4 Compact New Energy' template; region 'C05' contains 3 standard parking spaces, so it can be matched once. The final region template matching result is generated as [{Region ID:'A01', Applicable Template:'3 Standard Transformer 4 Compact New Energy'}, {Region ID:'C05', Applicable Template:'3 Standard Transformer 4 Compact New Energy'}].

[0071] To address the shortage of 8 "new energy" parking spaces, the system performs an optimized calculation of the conversion quantity. It employs a greedy algorithm, prioritizing a template conversion in region 'A01' with the most matching attempts, generating 4 new energy parking spaces, leaving a shortage of 4 spaces. Next, a template conversion is allocated again in the remaining candidate region 'C05', generating another 4 new energy parking spaces, completely filling the gap. Based on this, the system generates a region conversion quota table: [{Region ID:'A01', Template ID:'3 Standard to 4 Compact New Energy', Execution Count: 1}, {Region ID:'C05', Template ID:'3 Standard to 4 Compact New Energy', Execution Count: 1}].

[0072] Based on the regional conversion quota table, its contents are formatted into a set of standardized instructions that can be parsed by the physical execution layer. Each instruction specifies the operation location, operation basis, and operation target, ultimately generating a spatial conversion instruction set [{Region ID:'A001', Conversion Template:'3 Standard Transformer 4 Compact New Energy', Target Quantity: 4}, {Region ID:'C05', Conversion Template:'3 Standard Transformer 4 Compact New Energy', Target Quantity: 4}].

[0073] Preferably, step S3, which involves reorganizing the spatial identification and boundaries of parking spaces according to the spatial transformation instruction set, includes:

[0074] Based on the spatial conversion instruction set, the target area status is pre-locked to obtain the transitional parking space status diagram;

[0075] The physical identifier is redrawn according to the spatial transformation instruction set to obtain the identifier execution confirmation signal;

[0076] Deploy boundary isolation devices according to the spatial conversion instruction set and obtain an isolation execution confirmation signal;

[0077] Based on the transitional parking space status map, the sign execution confirmation signal, and the isolation execution confirmation signal, the resource map is solidified and updated to obtain the updated parking space resource status map.

[0078] In some embodiments, based on the region ID 'A01' in the spatial conversion instruction set, the status of the original parking space IDs 'A01-1', 'A01-2', and 'A01-3' contained in the current parking space resource status map is uniformly changed from "available" to "reconstructing" in a memory copy of the current parking space resource status map. This modified memory copy constitutes the transitional parking space status map and is immediately synchronized to the guidance system, making these parking spaces appear unavailable to the outside world.

[0079] A command is sent to the ground marking controller in target area 'A01'. The controller first extinguishes the white embedded LED light strips that form the boundaries of the original three standard parking spaces. Then, based on the coordinate data embedded in the '3 markings become 4 compact new energy parking spaces' template in the command, it illuminates a new combination of green LED light strips, precisely forming the visual boundaries and dedicated markings for the four compact new energy parking spaces. Upon completion of the operation, the controller returns a marking execution confirmation signal containing a status code: {'A01': 'Redrawing successful'}.

[0080] Upon receiving the confirmation signal for the identification execution, the system sends a command to the physical isolation controller in area 'A01'. This controller then drives eight retractable bollards installed on the newly defined parking space boundary line to rise 30 centimeters from ground level, creating a physical separation. After the operation is complete, the controller returns an isolation execution confirmation signal: {'A01': 'Isolation deployment successful'}.

[0081] After confirming the successful receipt of both the identification execution confirmation signal and the isolation execution confirmation signal for area 'A01', a database transaction operation is performed. This operation, based on the transitional parking space status diagram, logically deletes parking space IDs 'A01-1', 'A01-2', and 'A01-3' from the database and creates four new parking space IDs 'A01-N1' to 'A01-N4'. The system assigns these new parking spaces new attributes (type: new energy, size: compact) and their precise boundary coordinates, sets their initial state to "available," and ultimately generates an authoritative, updated parking space resource status diagram.

[0082] Preferably, step S4 includes:

[0083] Based on the real-time traffic flow vector, the updated parking space resource status map is searched and matched for target parking spaces to obtain a candidate parking space list.

[0084] Based on the candidate berth list, a guidance mode decision is made to obtain the guidance mode decision result, where the value of the guidance mode decision result is either regular guidance or attempting virtual queuing;

[0085] When the guidance mode decision result is regular guidance, the candidate berth list is locked and reserved to obtain a regular locked berth certificate;

[0086] When the guidance mode decision result is to attempt virtual queuing, high-confidence opportunity parking spaces are identified based on the individual vehicle departure probability table to obtain a list of opportunity parking spaces.

[0087] A virtual queuing plan is generated based on the list of available parking spaces and confirmed by the user to obtain a virtual queuing guidance voucher.

[0088] The virtual queuing guidance credentials are used for queuing guidance and status monitoring to obtain a parking space release confirmation event.

[0089] Based on the confirmation event of parking space release, the guidance right is switched and finally locked, and a virtual parking space lock certificate is obtained.

[0090] In some embodiments, after receiving the real-time traffic flow vector of a "new energy" type and "standard" size vehicle, the system uses this as a query condition to filter all parking spaces in the updated parking space resource status map that are "available". The system selects parking spaces that match both type and size, forming a candidate parking space list containing parking space IDs and their coordinates, for example, [{ID:'A01-N2', coordinates:[x1,y1]},{ID:'B03-N5', coordinates:[x2,y2]}].

[0091] Check the candidate berth list. If the list is not empty, the system generates a guidance mode decision of "normal guidance". If the list is empty, the system generates a guidance mode decision of "attempt virtual queuing".

[0092] If the guidance mode decision result is "normal guidance", the system calculates the path distance from the vehicle's current location to each candidate parking space in the list and selects the parking space 'A01-N2' with the shortest distance. Subsequently, the system updates the status of 'A01-N2' to "reserved" in the updated parking space resource status diagram, and binds the vehicle's ID with a 120-second reservation period, finally generating a normal parking space lock certificate containing the parking space ID and the reservation period.

[0093] If the guidance mode decision result is "attempt virtual queuing", the system scans the individual vehicle departure probability table, filters out all occupied parking spaces of the same type with a departure probability exceeding the 95% threshold, and forms a list of available parking spaces, such as [{ID:'C07', probability:97%}].

[0094] Based on the list of available parking spaces, the system selects the highest-probability space 'C07' and matches it with the physically nearest temporary waiting area. The system sends a message to the driver via the display screen at the entrance: "This type of parking space is full, but space C07 is expected to become available in 2 minutes. Would you like to join the queue in the nearby waiting area?" When multiple vehicles simultaneously request to "attempt virtual queuing," the system establishes a strict "first-come, first-served" queue based on the timestamps of the vehicles' arrival at the entrance trigger line. Only the vehicle at the front of the queue receives the current optimal queuing plan, ensuring fairness and orderliness in allocation. If the driver confirms their agreement via touchscreen, the system generates a virtual queuing guidance credential containing the target available parking space ID and the coordinates of the waiting area.

[0095] On one hand, the system guides vehicles to designated waiting points; on the other hand, it monitors the status of parking space 'C07' in real time with the highest priority. Simultaneously, a "maximum waiting timeout timer" is activated. If the timer expires and the available parking space remains unreleased, the system automatically triggers a "queue failure" event and immediately re-searches for the next optimal queuing target for the waiting vehicle. If no available opportunity is found after two attempts, the system prompts the user and guides them to leave, avoiding indefinite waiting. Once the status of 'C07' changes from "occupied" to "available," the system immediately generates a parking space release confirmation event.

[0096] Upon receiving the parking space release confirmation event, the system immediately performs a locking operation, setting the status of parking space 'C07' to "Reserved" and binding it to the ID of the vehicle currently in the queue. Simultaneously, the system invalidates the original virtual queuing guidance credential and generates a new virtual parking space lock credential pointing to 'C07,' completing a seamless transfer of guidance rights.

[0097] Preferably, step S4, which involves locking and reserving berths in the candidate berth list, includes:

[0098] The path distance is calculated based on the candidate berth list to obtain the berth path distance table;

[0099] Based on the berth route distance table, a comprehensive score of the berths is calculated to obtain a preferred berth ranking table;

[0100] The reservation time is dynamically calculated based on the berth preference ranking table to obtain the reservation time parameter;

[0101] Based on the reserved duration parameter, the updated parking space resource status map is used to lock the parking space status, and the parking space locking status is obtained.

[0102] Configure a failure policy for the parking space locking status to obtain the reserved policy configuration;

[0103] Generate a regular parking space locking certificate based on the reserved strategy configuration and parking space locking status.

[0104] In some embodiments, starting from the vehicle's current location, the shortest driving path distance to each parking space in the candidate parking space list is calculated using the A* algorithm based on a pre-set high-precision road network map in the parking lot, generating a parking space path distance table containing each parking space ID and its corresponding distance value.

[0105] Each parking space in the parking space route distance table is given a comprehensive score, and the scoring formula is as follows:

[0106] Overall score = W_d × (1 / path distance) + W_e × (1 / distance to exit);

[0107] The path distance starts at the vehicle's current location (e.g., just entering the parking lot) and ends at a candidate parking space. The exit distance starts at a candidate parking space and ends at the parking lot's vehicle exit. W_d is the path distance weight, set to 0.6, and W_e is the exit distance weight, set to 0.4. The system sorts all candidate parking spaces in descending order based on the calculated comprehensive score, forming a parking space preference ranking table.

[0108] Select the first parking space from the preferred parking space ranking list and dynamically calculate its reservation time. The calculation formula is:

[0109] Reserved time = (Optimal route distance / Average driving speed in the parking lot) + Parking buffer time;

[0110] The average driving speed within the parking lot is a preset value of 5 meters per second, and the parking buffer time is a fixed value of 60 seconds. The calculated result, for example, 96 seconds, is the reserved time parameter.

[0111] Based on the reservation duration parameter and the selected optimal parking space ID, a database update operation is performed in the updated parking space resource status diagram to change the parking space status from "available" to "reserved," and the vehicle ID that initiated the reservation and a reservation expiration timestamp accurate to the second are recorded. The status after this operation is completed is the parking space locked status.

[0112] Configure an invalidation policy for the parking space's locked status. This policy includes two rules: First, if the current system time exceeds the reserved expiration time stamp of the parking space and the vehicle has not arrived, the parking space status will automatically revert to "available"; second, if the system receives a cancel parking command from the user's terminal of the vehicle, the parking space status will immediately revert to "available". This is the reserved policy configuration.

[0113] The locked parking space ID, reserved vehicle ID, and reserved expiration timestamp are encapsulated into a JSON data object. This object is the regular parking space locking certificate, which is used for subsequent route guidance steps.

[0114] Preferably, step S5, which involves tracking the anchor point vehicle using a camera based on the anchor point navigation sequence, includes:

[0115] Based on the anchor point navigation sequence, virtual anchor point vehicle passing detection is performed using the camera video stream at the anchor point and the pre-set parking lot road network map to obtain the original list of vehicle passing events;

[0116] The vehicle identity is matched against the original list of vehicle passage events using a pre-set vehicle information database in the guidance process to obtain vehicle passage records with confirmed identities.

[0117] Generate vehicle calibration location records based on vehicle pass records verified by identity and anchor point navigation sequences;

[0118] The vehicle's dynamic driving parameters are obtained by dynamically calculating the driving parameters based on the vehicle calibration location records.

[0119] The vehicle's location is continuously updated based on dynamic driving parameters to obtain a real-time vehicle location and status table.

[0120] In some embodiments, a virtual detection line is defined in the camera monitoring screen corresponding to each anchor point, based on a preset parking lot road network map. When the video stream analysis module detects that the outline of a moving object completely crosses the detection line through background subtraction and centroid tracking algorithms, a vehicle passage event is triggered. The system records the anchor point ID, a timestamp accurate to milliseconds, and captures the image of the object, forming an entry containing {anchor point ID, timestamp, vehicle image}, which is then summarized into the original vehicle passage event list.

[0121] The system iterates through the original list of vehicle passage events, applying an optical character recognition (OCR) algorithm to extract the license plate number from the vehicle image in each entry. Then, it compares this license plate number with a "guiding vehicle information database" containing the IDs of all vehicles currently receiving guidance and their navigation sequences. If a match is found, the event is confirmed as a valid vehicle passage, and the system generates an identity-verified vehicle passage record with the structure {Vehicle ID, Anchor ID, Actual Arrival Time}.

[0122] Based on the vehicle pass records verified by identity, the system extracts the high-precision physical coordinates corresponding to the anchor point ID from the parking lot road network map. The system combines these coordinates with the actual arrival time in the record to generate a vehicle calibration location record {Vehicle ID, Calibration Location Coordinates, Calibration Time}. This record represents that the vehicle's actual position at that moment was precisely calibrated to that anchor point.

[0123] The system retrieves the last vehicle calibration location record for this vehicle ID. Using the location coordinates from both calibration records, the system obtains the precise road segment distance between the two anchor points from the road network map; using the timestamps from both calibration records, the system calculates the actual travel time. Based on this, the system calculates the vehicle's average speed on that road segment and compares it with the speed on the previous road segment to calculate the acceleration, forming the vehicle's dynamic driving parameters {Vehicle ID, Current Speed, Current Acceleration}.

[0124] A Kalman filter state is maintained for each guided vehicle. Using the dynamic driving parameters calculated in the previous step as input, and the vehicle calibration position records as observations, the filter is updated. As the vehicle travels between two anchor points, the system continuously predicts the vehicle's position and speed using this filter at a frequency of 100 milliseconds, and updates the prediction results in real-time to a centralized real-time vehicle position and state table. This table provides accurate spatiotemporal data for all guided vehicles for subsequent path conflict detection.

[0125] Preferably, step S5, which involves path conflict detection and coordination based on the real-time vehicle location and status table, includes:

[0126] Spatiotemporal trajectory projection is performed on the real-time vehicle position and status table and anchor point navigation sequence to obtain a multi-vehicle spatiotemporal trajectory set;

[0127] Spatiotemporal conflict regions are detected on a set of spatiotemporal trajectories of multiple vehicles to obtain a detailed list of conflict points;

[0128] Based on the list of conflict point details, the vehicle passage priority is assessed to obtain a passage priority decision table;

[0129] A collaborative optimization path set is generated based on the traffic priority decision table and the spatiotemporal trajectory set of multiple vehicles.

[0130] In some embodiments, for each vehicle receiving guidance, a spatiotemporal trajectory projection is performed over the next 60 seconds based on its real-time vehicle position, current position, speed, and acceleration in the status table, and the predetermined route defined by its anchor point navigation sequence. This projection generates a series of (x, y, t) data points accurate to coordinates and time with a time step of 0.5 seconds. The trajectory projections of all vehicles together constitute a multi-vehicle spatiotemporal trajectory set.

[0131] The system pairs and compares the trajectories of any two vehicles in a multi-vehicle spatiotemporal trajectory set. By calculating the Euclidean distance between the two vehicles at the same time step t, it detects whether there are overlapping situations where the distance is less than 5 meters (one vehicle's safe distance) and the duration exceeds 1 second. If such situations exist, the system records them as a conflict event and extracts the geographical location of the conflict (e.g., intersection ID 'J03'), the vehicle IDs involved, and the predicted conflict time to form a detailed list of conflict points.

[0132] Each event in the conflict point details list is evaluated based on a three-tier decision-making rule to determine traffic priority. The first tier rule prioritizes vehicles closer to the target parking space. The second tier rule prioritizes vehicles on the main road over those on side roads if they are close. The third tier rule prioritizes vehicles that arrive at the conflict point first if the first two criteria are indistinguishable. For pre-identified complex conflict areas with multiple vehicles intertwined (such as intersections and spiral ramps), the system also activates a "regional resource lock" mechanism. Vehicles must apply for and obtain a unique traffic lock for that area within a specific time window during route planning. Other vehicles will automatically avoid the locked time-space resources, thus achieving refined time-division reuse of key traffic areas and alleviating complex congestion. The evaluation results form a decision table that clearly indicates the traffic priority of priority vehicles and vehicles that need to yield.

[0133] The paths for yielding vehicles are dynamically adjusted based on the traffic priority decision table. For priority vehicles, their anchor point navigation sequence remains unchanged. For yielding vehicles, such as vehicle B, the system recalculates their target speed for the segment preceding the conflict point 'J03', ensuring that their arrival time at 'J03' is 3 seconds later than the time when priority vehicle A safely passes that point. This adjustment generates a new anchor point navigation sequence that includes deceleration instructions. The final, conflict-free navigation sequences for all vehicles collectively constitute the collaboratively optimized path set.

[0134] Finally, the collaboratively optimized path set is encapsulated into personalized parking space beacons and distributed through a three-tiered priority channel: First, centimeter-level indoor navigation commands are pushed to users with the accompanying mobile application installed via Bluetooth Low Energy beacons; second, path data is sent to connected vehicle navigation systems via cloud API; third, as a general backup solution, the license plate number of the target vehicle and clear turn arrows are displayed on the information display screen at key intersections to ensure that all vehicles can receive guidance information.

[0135] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0136] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for predicting and guiding tidal parking spaces in a parking lot, characterized in that, Includes the following steps: Step S1: Obtain video data of traffic flow at the parking lot entrance and perform vehicle attribute feature analysis to obtain real-time traffic flow vectors; Step S2: Predict demand trends from real-time traffic flow vectors to obtain the expected total demand; obtain and extract individual vehicle departure probability tables from the current parking space resource status map; predict the total parking space release from the individual vehicle departure probability tables to obtain a parking space release prediction table; calculate the net supply-demand gap based on the parking space release prediction table and the expected total demand. This calculation is applied to parking space types, and finally outputs a type imbalance index that quantifies future supply and demand pressure. Step S3: Dynamically generate a reorganization strategy based on the type imbalance index. Specifically, the parking space types are divided into demand queues and redundant queues according to the type imbalance index. Then, continuous areas consisting of redundant types, in an available state, and with a physical basis for conversion are selected from the current parking space resource status map according to priority. These areas are matched with a template library containing multiple conversion schemes. A greedy algorithm is used to generate a regional conversion quota table containing specific areas and conversion templates with the goal of filling the gap in the most efficient way. Finally, the table is formatted as a space conversion instruction set, which includes the area ID, conversion template, and target quantity. Based on the spatial transformation instruction set, the spatial identification and boundary of the parking space are reorganized to obtain the updated parking space resource status map; Step S4 includes: performing target parking space retrieval and matching on the updated parking space resource status map based on the real-time traffic flow vector to obtain a candidate parking space list; making a guidance mode decision based on the candidate parking space list to obtain a guidance mode decision result, where the value of the guidance mode decision result is either regular guidance or virtual queuing attempt; when the guidance mode decision result is regular guidance, locking and reserving parking spaces on the candidate parking space list to obtain a regular locked parking space certificate; when the guidance mode decision result is virtual queuing attempt, identifying high-confidence opportunity parking spaces based on an individual vehicle departure probability table to obtain an opportunity parking space list; generating a virtual queuing scheme and obtaining user confirmation based on the opportunity parking space list to obtain a virtual queuing guidance certificate; performing queuing guidance and status monitoring on the virtual queuing guidance certificate to obtain a parking space release confirmation event; and switching guidance rights and finally locking the parking space based on the parking space release confirmation event to obtain a virtual locked parking space certificate. Step S5: Use the locked berth credentials to perform preliminary route planning and generate the anchor point sequence to obtain the anchor point navigation sequence; Based on the anchor point navigation sequence, virtual anchor point vehicle passing detection is performed using the camera video stream at the anchor point and the pre-set parking lot road network map to obtain the original vehicle passing event list; the vehicle identity is matched against the original vehicle passing event list using the pre-set vehicle information database in the guidance to obtain the identity-confirmed vehicle passing record; and the vehicle calibration position record is generated based on the identity-confirmed vehicle passing record and the anchor point navigation sequence. The vehicle's dynamic driving parameters are obtained by dynamically calculating the driving parameters based on the vehicle calibration location records. Based on the vehicle's dynamic driving parameters, position prediction is continuously updated to obtain a real-time vehicle position and status table; based on the real-time vehicle position and status table, path conflict detection and coordination are performed to obtain a collaboratively optimized path set; based on the collaboratively optimized path set and the locked parking space credentials, multi-channel guidance beacons are distributed to obtain personalized parking space beacons.

2. The parking lot tidal parking space prediction and guidance method according to claim 1, characterized in that, Step S1 includes: Acquire and capture instantaneous images of vehicles from the real-time video stream at the parking lot entrance to form the original image of the vehicle to be identified; The original image of the vehicle to be identified is distributed to three parallel processing units for preliminary parallel feature extraction, specifically: The first unit is the license plate location and character recognition unit, which quickly locates the rectangular area of ​​the license plate through a cascaded classifier, and then uses a convolutional neural network model to perform optical character recognition on the characters in the area to output the license plate number. The second unit is the salient identifier search unit, which uses the color space threshold segmentation method to detect the gradient green unique to new energy vehicle license plates, and uses a template matching algorithm based on feature points to compare the vehicle's front face and side door areas with a database containing ten car-sharing platform logos. The third unit is the vehicle body contour measurement unit, which uses the Canny edge detection algorithm to outline the two-dimensional contour of the vehicle, calculates its pixel width and length, and uses a pre-calibrated standard width line at the entrance as a reference to convert the pixel size into a physical size estimate, ultimately forming a discrete feature set containing license plate text, salient color blocks, matching logos, and contour estimates. The discrete feature set is logically summarized and labeled using a pre-defined rule base, specifically as follows: If the prominent color block is green, the vehicle type is labeled as new energy; otherwise, if the matching logo is any logo in the database, it is labeled as shared; otherwise, it is labeled as ordinary. In the size determination branch, if the estimated length value of the outline is less than 4.2 meters, the vehicle size is labeled as compact; if the length value is between 4.2 meters and 5.0 meters, it is labeled as standard; otherwise, it is labeled as large, resulting in a standardized attribute dataset, which includes type and size. The standardized attribute dataset is encapsulated into traffic flow vectors to generate real-time traffic flow vectors.

3. The parking lot tidal parking space prediction and guidance method according to claim 1, characterized in that, Step S2 includes: Aggregate recent traffic flow features from real-time traffic flow vectors to obtain immediate demand intensity; Based on the intensity of immediate demand, extract the historical traffic baseline from the preset historical database; The expected total demand is obtained by performing a trend-weighted fusion forecast of the immediate demand intensity and the historical baseline of the same period's flow. Extract the current available parking space data and occupied parking space data of each type from the current parking space resource status map, calculate the probability of vehicles leaving the site, and obtain the parking space release prediction table and the individual vehicle departure probability table. The calculation of the probability of vehicles leaving the site includes: statistical analysis of vehicle dwell time on the occupied parking space data to obtain the vehicle dwell status table. Perform categorized parking pattern matching on the vehicle parking status table, specifically as follows: The dwell time of each vehicle is compared with the historical average dwell time corresponding to its type. The basic departure probability is calculated by a preset sigmoid function. This probability approaches 1 from 0 as the dwell time approaches the average dwell time, forming a basic departure probability table. The contextual factor weights of the basic exit probability table are adjusted as follows: The application scenario adjustment coefficient corrects the basic departure probability. During the period from 17:00 to 19:00, the departure probability of all vehicles is multiplied by a coefficient of 1.2; for shared vehicles, the departure probability is multiplied by an additional coefficient of 1.

15. These coefficients are not fixed values, but are dynamically generated by regression analysis of historical data to quantify the impact of specific time periods and weather factors on the average parking time. The corrected results constitute an individual vehicle departure probability table.

4. The parking lot tidal parking space prediction and guidance method according to claim 1, characterized in that, Step S3 involves dynamically generating a reorganization strategy based on the type imbalance index, including: The imbalance type priority is sorted according to the type imbalance index. Specifically, after receiving the type imbalance index, the priority is sorted, positive values ​​are arranged in descending order and negative values ​​are arranged in ascending order, generating two independent queues, including the demand queue and the redundancy queue, which is the parking space type conversion priority queue. Based on the parking space type conversion priority queue, the reconfigurable areas are filtered in the current parking space resource status graph to obtain a list of candidate reconfigurable areas; For each region in the candidate reconstruction region list, an adaptation and matching process is performed from a fixed transformation template library to obtain the region template matching result. The transformation template library contains entries for template ID, input type, number of inputs, and number of outputs, and the region template matching result contains region ID and applicable template. The conversion quantity is optimized based on the regional template matching results, specifically as follows: A greedy algorithm is used to first allocate a template transformation in the region with the most matching times, and then allocate a template transformation again in the remaining candidate regions to obtain a region transformation quota table, which contains region ID, template ID and execution count; The spatial conversion instruction set is obtained by formatting the spatial conversion instruction based on the regional conversion quota table.

5. The parking lot tidal parking space prediction and guidance method according to claim 1, characterized in that, Step S3, which involves reorganizing the spatial identification and boundaries of parking spaces according to the spatial transformation instruction set, includes: Based on the spatial conversion instruction set, the target area status is pre-locked to obtain the transitional parking space status diagram; The physical identifier is redrawn according to the spatial transformation instruction set to obtain the identifier execution confirmation signal; Deploy boundary isolation devices according to the spatial conversion instruction set and obtain an isolation execution confirmation signal; Based on the transitional parking space status map, the sign execution confirmation signal, and the isolation execution confirmation signal, the resource map is solidified and updated to obtain the updated parking space resource status map.

6. The parking lot tidal parking space prediction and guidance method according to claim 5, characterized in that, Step S4, which involves locking and reserving berths in the candidate berth list, includes: The path distance is calculated based on the candidate berth list to obtain the berth path distance table; Based on the berth route distance table, a comprehensive score of the berths is calculated to obtain a preferred berth ranking table; The reservation time is dynamically calculated on the parking space optimization ranking table to obtain the reservation time parameter. The calculation formula for the dynamic calculation of the reservation time is: reservation time = (optimal path distance / average driving speed in the parking lot) + parking buffer time. Based on the reserved duration parameter, the updated parking space resource status map is used to lock the parking space status, and the parking space locking status is obtained. Configure a failure policy for the parking space locking status to obtain a reservation policy configuration. The failure policy is that if the current system time exceeds the reserved expiration time stamp of the parking space and the vehicle has not arrived, the parking space status will automatically be restored to available; if the system receives a cancel parking command sent from the user terminal of the vehicle, the parking space status will be restored to available immediately. Generate a regular parking space locking certificate based on the reserved strategy configuration and parking space locking status.

7. The parking lot tidal parking space prediction and guidance method according to claim 1, characterized in that, Step S5, which involves path conflict detection and coordination based on real-time vehicle location and status tables, includes: Spatiotemporal trajectory projection is performed on the real-time vehicle position and status table and anchor point navigation sequence to obtain a multi-vehicle spatiotemporal trajectory set; Spatiotemporal conflict regions are detected on a set of spatiotemporal trajectories of multiple vehicles to obtain a detailed list of conflict points; Based on the list of conflict point details, the vehicle passage priority is assessed to obtain a passage priority decision table; A collaborative optimization path set is generated based on the traffic priority decision table and the spatiotemporal trajectory set of multiple vehicles.

8. A parking lot tidal parking space prediction and guidance system, characterized in that, For executing the parking lot tidal parking space prediction and guidance method as described in claim 1, the parking lot tidal parking space prediction and guidance system includes: The vehicle flow profile acquisition module is used to acquire video data of vehicle flow at the parking lot entrance and perform vehicle attribute feature analysis to obtain real-time vehicle flow vectors. The demand pressure forecasting module is used to predict demand trends from real-time traffic flow vectors to obtain the expected total demand; obtain and extract individual vehicle departure probability tables from the current parking space resource status map; and calculate the type imbalance index based on the expected total demand. The spatial flexible reorganization module is used to dynamically generate reorganization strategies based on the type imbalance index, resulting in a spatial conversion instruction set; and to reorganize the spatial identifiers and boundaries of parking spaces based on the spatial conversion instruction set, resulting in an updated parking space resource status map. The intelligent parking space allocation module is used to make guidance mode decisions and allocate parking spaces based on the individual vehicle departure probability table and the updated parking space resource status map, and obtain the locked parking space certificate, which includes the regular locked parking space certificate and the virtual locked parking space certificate. The multi-vehicle collaborative guidance module is used to perform preliminary route planning and anchor point sequence generation using locked berth credentials to obtain an anchor point navigation sequence; based on the anchor point navigation sequence, it tracks anchor point vehicles through cameras to obtain a real-time vehicle position and status table; based on the real-time vehicle position and status table, it performs route conflict detection and coordination to obtain a collaboratively optimized route set; based on the collaboratively optimized route set and locked berth credentials, it distributes guidance beacons through multiple channels to obtain personalized berth beacons.

Images