Method and system for predicting idle parking spaces of parking lot

Abstract

The invention provides a method and a system for predicting idle parking spaces of a parking lot. The method comprises the steps that ramp curvature parameters extracted from topological structure data of the special-shaped parking lot and historical vehicle turning radius data are fused, and a spatial-temporal topological graph of the special-shaped parking lot is generated; aggregating the parking space occupancy state data of all the special-shaped parking lots and the special-shaped parking lot space-time topological graph through a federal transfer learning framework to obtain target feature extraction model parameters; combining the time change characteristics of the channel flow of the special-shaped parking lot in the spatial-temporal topological graph of all the special-shaped parking lots and the target characteristic extraction model parameters, and generating a dynamic parking space distribution strategy; and mapping the dynamic parking space distribution strategy to a real-time navigation path of the special-shaped parking lot space-time topological graph, so that the to-be-parked vehicle arrives at the idle parking space through the real-time navigation path. According to the invention, the dynamic parking space distribution strategy is generated by fusing the topological structure and historical data of the special-shaped parking lot and real-time navigation is carried out, so that the parking space utilization rate and the parking efficiency are optimized.

Description

Technical Field

[0001] This application relates to the field of computer application technologies, and particularly to a method and system for predicting available parking spaces in a parking lot. Background Art

[0002] In irregular parking lots (such as spiral ramp parking lots, multi-story stereoscopic parking lots, etc.), due to reasons such as curved lanes, complex ramps, dynamic changes in vehicle flow, and difficulty in visually identifying parking spaces, the time for vehicles to search for available parking spaces increases, thereby affecting the overall operation efficiency of the parking lot. To address this challenge, a technical solution that can monitor and analyze the occupancy status of each parking space in the parking lot in real time is needed to improve the utilization rate of parking spaces, reduce the detention time of vehicles in the parking lot, and optimize the allocation of space resources in the parking lot.

[0003] Existing solutions use sensor technology and wireless communication networks to monitor the occupancy of parking spaces. Specifically, existing solutions can collect real-time data on the occupancy status of parking spaces by installing geomagnetic sensors or ultrasonic sensors in each parking space, and transmit these data to the central control system for processing.

[0004] Existing solutions usually can only provide data on the occupancy status of parking spaces, resulting in insufficient responsiveness to dynamic traffic flow changes. And existing solutions are difficult to meet the requirements of joint management of multi-irregular parking lots, unable to adjust the parking space allocation strategy and navigation path in real time, resulting in low efficiency for waiting vehicles to search for available parking spaces during peak hours or emergencies. Summary of the Invention

[0005] This application provides a method for predicting available parking spaces in a parking lot to solve the problems in the prior art that the response to dynamic traffic flow is not sensitive, the joint management of multi-irregular parking lots is difficult, and the real-time adjustment of the parking space allocation strategy and navigation path is insufficient, resulting in low efficiency for waiting vehicles to search for available parking spaces during peak hours or emergencies.

[0006] In a first aspect, this application provides a method for predicting available parking spaces in a parking lot, including: obtaining occupancy status data of parking spaces in multiple irregular parking lots and topological structure data of the irregular parking lots; For each irregular parking lot, by fusing the ramp curvature parameters extracted from the topological structure data of the irregular parking lot and historical vehicle turning radius data, generating a spatio-temporal topological map of the irregular parking lot; Through a federated transfer learning framework, aggregating the occupancy status data of all the irregular parking lots and the spatio-temporal topological map of the irregular parking lot to obtain target feature extraction model parameters; Combining the time-varying characteristics of the channel flow of the irregular parking lots in all the spatio-temporal topological maps of the irregular parking lots and the target feature extraction model parameters to generate a dynamic parking space allocation strategy; Map the dynamic parking space allocation strategy to the real-time navigation path of the special-shaped parking lot spatio-temporal topology map, so that the vehicle to be parked can reach the idle parking space through the real-time navigation path.

[0007] Optionally, aggregating the parking space occupancy status data of all the special-shaped parking lots and the special-shaped parking lot spatio-temporal topology map through the federated transfer learning framework to obtain the target feature extraction model parameters, including: Input the parking space occupancy status data of each special-shaped parking lot and the special-shaped parking lot spatio-temporal topology map into the corresponding local nodes in the federated transfer learning framework, so as to store the parking space occupancy status data and the special-shaped parking lot spatio-temporal topology map of the corresponding special-shaped parking lot in each local node, and one local node corresponds to one special-shaped parking lot; In each local node, extract the spiral ramp geometric parameters in the special-shaped parking lot spatio-temporal topology map, and the spiral ramp geometric parameters include lane curvature; Based on the parking space occupancy status data and the spiral ramp geometric parameters, train the local feature extraction model on the local node to obtain the local feature extraction model parameters related to the lane curvature; In the central server in the federated transfer learning framework, aggregate all the local feature extraction model parameters to generate the global feature extraction model parameters; Perform convergence processing on the global feature extraction model parameters to obtain the target feature extraction model parameters.

[0008] Optionally, the aggregating all the local feature extraction model parameters to generate the global feature extraction model parameters includes: Adopt a weighted average strategy to fuse all the local feature extraction model parameters to generate intermediate feature extraction model parameters; wherein, the weighted average strategy includes a weight allocation scheme determined based on the data volume, data quality or model performance index of each local node; Verify the intermediate feature extraction model parameters, and determine the verified intermediate feature extraction model as the global feature extraction model parameters.

[0009] Optionally, combining the time variation characteristics of the channel flow of the special-shaped parking lots in all the special-shaped parking lot spatio-temporal topology maps and the target feature extraction model parameters to generate a dynamic parking space allocation strategy, including: Use a pre-trained machine learning algorithm to extract the time variation characteristics of the channel flow of the corresponding special-shaped parking lot from each special-shaped parking lot spatio-temporal topology map, and the time variation characteristics include the vehicle flow data of each channel in the corresponding special-shaped parking lot at different time periods; Input the time-varying characteristics of the traffic flow in all irregular parking lots into a target feature extraction model that adopts target feature extraction model parameters to obtain an integrated spatio-temporal feature representation, where the integrated spatio-temporal feature representation is used to describe the traffic flow patterns and their dynamic change trends of each irregular parking lot; Obtain the real-time traffic flow data at the current moment, and combine it with the integrated spatio-temporal feature representation to predict the distribution of available parking spaces in each irregular parking lot within a future time period, obtaining a distribution map of available parking spaces, where the distribution map of available parking spaces contains the location information and the estimated available time of each available parking space; Based on the distribution map of available parking spaces and in combination with the spiral ramp geometric parameters, calculate the accessibility score value of each available parking space; According to the accessibility score value of each available parking space, perform load balancing analysis on all the irregular parking lots to obtain a list of candidate parking spaces for the vehicles to be parked; Obtain the driving trajectory information of the vehicle to be parked, and generate a dynamic parking space allocation strategy according to the driving trajectory information and the list of candidate parking spaces for the vehicle to be parked.

[0010] Optionally, training the local feature extraction model on the local node based on the parking space occupancy status data and the spiral ramp geometric parameters to obtain local feature extraction model parameters related to the lane curvature, including: Initialize the local feature extraction model; Based on the parking space occupancy status data and the spiral ramp geometric parameters, construct an input data set for the local feature extraction model; Input the input data set into the initialized local feature extraction model, and through training, obtain local feature extraction model parameters related to the lane curvature.

[0011] Optionally, the obtaining of the parking space occupancy status data of multiple irregular parking lots includes: Obtain the pressure distribution data of each parking space in the corresponding irregular parking lot collected by each flexible pressure film sensor array in real time. One irregular parking lot corresponds to one flexible pressure film sensor array. Each parking space in the irregular parking lot is provided with a flexible pressure film sensor, and each flexible pressure film sensor is provided on the curved lane in the irregular parking lot. The flexible pressure film sensor array is arranged in a conforming manner according to the contour of the parking spaces and the curved lane contour of the irregular parking lot; Convert the pressure distribution data of the parking space into parking space occupancy status data.

[0012] Optionally, the converting of the pressure distribution data of the parking space into parking space occupancy status data includes: According to a preset parsing rule, parse the pressure distribution data to obtain the vehicle parking position and weight distribution information; Generate a parking space occupancy identifier based on the vehicle parking position and the weight distribution information; Combine the parking space occupancy identifier with the vehicle weight distribution information to generate parking space occupancy status data.

[0013] In a second aspect, the present application provides a remote monitoring system for the operating state of a deluge alarm valve group, including: An acquisition module, configured to acquire parking space occupancy status data and special-shaped parking lot topological structure data of multiple special-shaped parking lots; A generation module, configured to generate a spatio-temporal topological map of a special-shaped parking lot for each special-shaped parking lot by fusing ramp curvature parameters extracted from the special-shaped parking lot topological structure data and historical vehicle turning radius data; An extraction module, configured to aggregate the parking space occupancy status data of all the special-shaped parking lots and the spatio-temporal topological map of the special-shaped parking lot through a federated transfer learning framework to obtain target feature extraction model parameters; A generation module, configured to generate a dynamic parking space allocation strategy by combining the time variation characteristics of the traffic flow of the passageways of the special-shaped parking lots in all the spatio-temporal topological maps of the special-shaped parking lots and the target feature extraction model parameters; A mapping module, configured to map the dynamic parking space allocation strategy to the real-time navigation path of the spatio-temporal topological map of the special-shaped parking lot, so that the vehicle to be parked can reach the available parking space through the real-time navigation path.

[0014] In a third aspect, the present application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for predicting available parking spaces in a parking lot as described in any item of the first aspect.

[0015] In a fourth aspect, the present application provides a computer storage medium, storing a computer program, and when the computer program is executed by a computer, it implements a method for predicting available parking spaces in a parking lot as described in any item of the first aspect.

[0016] In this application, a method for predicting the available parking spaces in a parking lot is provided. The method includes: obtaining the parking space occupancy status data and the topological structure data of the special-shaped parking lot for multiple special-shaped parking lots; for each special-shaped parking lot, generating a spatio-temporal topological map of the special-shaped parking lot by fusing the ramp curvature parameters extracted from the topological structure data of the special-shaped parking lot and the historical vehicle turning radius data; aggregating the parking space occupancy status data of all the special-shaped parking lots and the spatio-temporal topological map of the special-shaped parking lot through a federated transfer learning framework to obtain the target feature extraction model parameters; combining the time variation characteristics of the channel flow of the special-shaped parking lots in all the spatio-temporal topological maps of the special-shaped parking lots with the target feature extraction model parameters to generate a dynamic parking space allocation strategy; mapping the dynamic parking space allocation strategy to the real-time navigation path of the spatio-temporal topological map of the special-shaped parking lot, so that the vehicle to be parked can reach the available parking space through the real-time navigation path.

[0017] This application realizes the comprehensive integration and utilization of multi-source data by obtaining the parking space occupancy status data and topological structure data of multiple irregular parking lots; generates an accurate spatio-temporal topological map of irregular parking lots by fusing ramp curvature parameters and historical vehicle turning radius data, providing a reliable basis for dynamic parking space allocation; aggregates the data of multiple parking lots through a federated transfer learning framework to generate target feature extraction model parameters, supporting the collaborative management and resource optimization of multiple irregular parking lots. Combining the time-varying characteristics of channel traffic flow with the target feature extraction model parameters, a dynamic parking space allocation strategy is generated to improve the real-time response ability to traffic flow changes. Mapping the dynamic parking space allocation strategy to the real-time navigation path enables the vehicle to be parked to quickly reach the available parking space, improving the parking efficiency, especially excellent performance during peak hours or emergencies. By adjusting the parking space allocation strategy and navigation path in real time, it effectively responds to the complex management requirements of multiple irregular parking lots and improves the overall intelligence level of the system. Further, the parking space occupancy status data and spatio-temporal topological map of each irregular parking lot are input into the local nodes of the federated transfer learning framework. Each local node stores the data of the corresponding parking lot and extracts the spiral ramp geometric parameters (such as lane curvature) in the spatio-temporal topological map; based on the parking space occupancy status data and spiral ramp geometric parameters, a local feature extraction model is trained to obtain local model parameters related to lane curvature; in the central server, all local model parameters are aggregated using a weighted average strategy to generate intermediate feature extraction model parameters, and the global feature extraction model parameters are determined through verification, and finally the target feature extraction model parameters are converged. Through the federated transfer learning framework, the distributed storage and collaborative training of multi-irregular parking lot data are realized. Combining the spiral ramp geometric parameters and parking space occupancy status data improves the accuracy and adaptability of the feature extraction model; using the weighted average strategy to aggregate local model parameters ensures the robustness and generalization ability of the global model. The finally generated target feature extraction model parameters can effectively support the optimization of dynamic parking space allocation and navigation strategies, improving the intelligence level and efficiency of parking management.

[0018] These aspects or other aspects of this application will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of this application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 It is a flowchart of a method for predicting available parking spaces in a parking lot provided by an embodiment of this application; Figure 2Schematic structural diagram of a prediction system for available parking spaces in a parking lot provided by an embodiment of the present application; Figure 3 Schematic structural diagram of a computing device provided by an embodiment of the present application. Detailed implementation manners

[0021] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application.

[0022] In some processes described in the specification and claims of the present application and the above-mentioned accompanying drawings, a plurality of operations appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order in which they appear herein or may be executed in parallel. The serial numbers of the operations, such as 11, 12, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0023] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0024] Figure 1 Flowchart of a method for predicting available parking spaces in a parking lot provided by an embodiment of the present application, as Figure 1 shown, the method includes: S11. Obtain the parking space occupancy status data and the topological structure data of the special-shaped parking lot for multiple special-shaped parking lots.

[0025] Among them, the parking space occupancy status data includes the real-time occupancy status of each parking space in the parking lot (such as free, occupied, reserved, etc.), which is usually collected through sensors, cameras or Internet of Things devices. The topological structure data of the special-shaped parking lot is the physical structure information describing the parking lot, including geometric features such as lanes, ramps, turning areas, and entrance and exit positions, specifically including ramp curvature parameters, etc., which is usually obtained through architectural design drawings or 3D modeling. The ramp can refer to the lane passed by the entrance and exit, and can also refer to the parking ramp with parking spaces.

[0026] In the embodiments of the present application, in this embodiment, a flexible pressure film sensor array laid conformally in a special-shaped parking lot can be used to collect parking space pressure data in real time, and multi-modal data fusion can be performed in combination with cameras and Internet of Things devices to further determine the real-time occupancy status of the parking space. For example, if the pressure data corresponding to the parking space is greater than or equal to the preset pressure threshold, it means that there is already a vehicle on the parking space, so the real-time occupancy status of the parking space is occupied; if the pressure data corresponding to the parking space is less than the preset pressure threshold, it means that there is no vehicle on the parking space, and the photo of the parking space taken by the camera can be combined to confirm that there is indeed no vehicle on the parking space. Then, in combination with the Internet of Things device, it can be checked whether the parking space is reserved during the current time period. If so, the real-time occupancy status is reserved, otherwise it is idle. The embodiments of the present application can generate a digital twin model of the parking lot based on building information modeling or three-dimensional laser scanning. The embodiments of the present application can use the least squares method to fit an arc to determine the lane curvature radius, and determine the slope of the ramp, the position of the entrance and exit, etc. according to the triangulation method.

[0027] Exemplarily, a certain commercial complex includes 3 special-shaped parking lots (A, B, and C), and flexible pressure film sensors are installed conformally on their parking spaces. The resolution of the flexible pressure film sensor is 5 cm × 5 cm. The spiral ramp of Parking Lot A is modeled to determine the curvature radius of 3.2 m and the slope of 12°. After the sensor data is parsed by the edge computing node, it is determined that the pressure value of Parking Space A-07 at 12:00 is 2100 kg (exceeding the threshold), and it is marked as "occupied by a large vehicle". The data is uploaded to the cloud through the protocol and jointly constitutes the initial data set with the data of Parking Lots B and C.

[0028] S12. For each special-shaped parking lot, by fusing the ramp curvature parameters extracted from the special-shaped parking lot topology structure data and the historical vehicle turning radius data, a spatio-temporal topology map of the special-shaped parking lot is generated.

[0029] Among them, the ramp curvature parameter is used to describe the geometric characteristics of the parking lot ramp, including the curvature radius, slope, length, etc. of the ramp, and is used to evaluate the difficulty and safety of vehicle driving. The historical vehicle turning radius data is used to record the actual turning radius of different types of vehicles in the parking lot, and is used to optimize path planning and parking space allocation. The spatio-temporal topology map of the special-shaped parking lot is a dynamic graph structure that combines the physical topology information of the parking lot and the traffic flow change characteristics in the time dimension, and is used to describe the possible driving paths and bottleneck areas of vehicles in the parking lot.

[0030] In the embodiments of the present application, in order to integrate the ramp curvature parameter and the historical vehicle turning radius data, the radius of curvature, slope, and length can be encoded as vectors, for example, 3.2m, 12°, 50m. In order to eliminate the dimension, normalization processing can also be performed to obtain [0.32, 0.12, 0.5]. This embodiment can also classify vehicles into small cars (turning radius 4m), medium-sized cars (6m), and large cars (8m) based on K-means clustering. If the ramp radius of curvature ≥ the vehicle turning radius, the vehicle is marked as "passable". Then, this embodiment can use a graph database to store the dynamic graph structure, where the nodes are parking spaces, channels, etc., and the edge weights are the real-time passing times. The real-time passing time of this ramp provides data support for the subsequent calculation of road congestion and road scoring.

[0031] Exemplarily, the spiral ramp with a radius of curvature of 3.2m in Parking Lot A matches the historical vehicle data and only allows small cars with a turning radius ≤ 3.5m to pass. In the spatio-temporal topology graph, the passing time of this ramp at 12:00 is the channel length 50m ÷ the reference vehicle speed 1.39m / s × (1 + congestion sensitivity parameter 0.8 × current traffic flow 15 vehicles / channel capacity 20 vehicles) = 48 seconds. Correspondingly, the congestion coefficient is 15 / 20 = 0.6, so dynamic weight adjustment can be triggered, resulting in a decrease in the score of this path.

[0032] S13. Aggregate the parking space occupancy status data of all irregular parking lots and the spatio-temporal topology graph of the irregular parking lots through the federated transfer learning framework to obtain the target feature extraction model parameters.

[0033] Among them, the federated transfer learning framework is a distributed machine learning framework that allows multiple local nodes to collaboratively train a model without sharing the original data, protecting data privacy. The target feature extraction model parameters are the global model parameters generated through federated learning and are used to extract the common features of parking lot data to support dynamic parking space allocation and path planning.

[0034] In this embodiment, each parking lot serves as a federated learning node for distributed training. The input of the graph attention network is the spatio-temporal topology graph, and the output is the correlation features between lane curvature and parking space occupancy. The loss function of this network can be the sum of regularization and mean square error. The central server adopts a weighted average strategy and can determine the weight of the model parameters provided by this federated learning node according to the data volume, data quality, and model accuracy. Exemplarily, in terms of quantity, Parking Lot A has 1000 pieces, B has 800 pieces, and Parking Lot C has 1200 pieces. In terms of data quality, the data quality of Parking Lot A is 0.9, the data quality of Parking Lot B is 0.7, and the data quality of Parking Lot C is 0.8. In terms of model accuracy, the model accuracy of Parking Lot A is 95%, the model accuracy of Parking Lot B is 90%, and the model accuracy of Parking Lot C is 92%. Finally, the weight corresponding to Parking Lot A is 0.35. The central server aggregates the model parameters of all parking lots to generate global model parameters, and the validation set accuracy is improved, supporting cross-parking lot feature generalization.

[0035] S14. Combine the time-varying characteristics of the channel traffic of the special-shaped parking lots in all spatio-temporal topology graphs of special-shaped parking lots with the target feature extraction model parameters to generate a dynamic parking space allocation strategy.

[0036] Among them, the time-varying characteristics of the channel traffic describe the traffic flow distribution in different time periods within the parking lot, such as the congested areas during peak hours and the idle areas during off-peak hours. The dynamic parking space allocation strategy dynamically adjusts the parking space allocation plan according to the real-time traffic flow and the target feature extraction model parameters to optimize the parking efficiency and resource utilization rate.

[0037] Exemplarily, when the score of the idle parking space 12 in Parking Lot C is 3.0 and the score of the idle parking space 05 in Parking Lot B is 2.2, the dynamic allocation strategy preferentially guides the vehicle to be parked to Parking Lot C.

[0038] S15. Map the dynamic parking space allocation strategy to the real-time navigation path of the spatio-temporal topology graph of the special-shaped parking lot so that the vehicle to be parked can reach the idle parking space through the real-time navigation path.

[0039] Among them, the real-time navigation path is the vehicle driving path generated according to the dynamic parking space allocation strategy, including turning prompts, ramp prompts, and parking space guidance information.

[0040] In the embodiment of the present application, the improved A* algorithm can be used, and the cost function is the sum of the actual travel time, the estimated remaining time, and the path reachability score.

[0041] Exemplarily, after the vehicle enters parking lot C, the ramp passing time is determined to be 48 seconds based on the real-time topology map, and the planned path is the entrance, ramp passage C3, until the vacant parking space 12. The navigation instructions can be updated synchronously through the display screen and in-vehicle terminal, prompting "Spiral ramp ahead, speed limit 5 km / h".

[0042] The following is a specific example: Suppose there are 3 irregularly shaped parking lots (A, B, C) in a certain city. The topological structure of each parking lot is complex, including multiple spiral ramps and narrow turning areas. During peak hours, the occupancy rates of parking lots A and B are relatively high, while parking lot C is relatively vacant. The occupancy status data of the 3 parking lots are obtained through the sensor network, and the topological structure data are generated through digital modeling. The ramp curvature parameters of parking lots A, B, and C are extracted, and combined with the historical vehicle turning radius data to generate their respective spatio-temporal topology maps. Through the federated transfer learning framework, the data of the 3 parking lots are aggregated to train the target feature extraction model parameters. Combining the time-varying features in the spatio-temporal topology map (such as the congestion conditions of parking lots A and B during peak hours), a dynamic parking space allocation strategy is generated to preferentially guide vehicles to parking lot C. The dynamic parking space allocation strategy is mapped into the real-time navigation path, and the driver is guided by the navigation system to efficiently reach the vacant parking space in parking lot C.

[0043] By executing S11~S15, the solution of this embodiment of the present application realizes the collaborative management and resource optimization of multiple irregularly shaped parking lots through the deep integration of multi-source heterogeneous data, the federated transfer learning framework, and the dynamic path planning algorithm. The system can real-time sense and respond to dynamic traffic flow changes, and based on the global feature extraction model parameters and spatio-temporal topology map, dynamically generate and adjust the parking space allocation strategy and navigation path, thereby improving the parking efficiency and resource utilization rate. Especially during peak hours or in emergency scenarios, the system effectively alleviates the parking lot congestion problem through the intelligent path guidance and parking space allocation mechanism, providing a more efficient, intelligent, and convenient parking service experience for drivers.

[0044] In a possible embodiment, in S13, the occupancy status data of all irregularly shaped parking lots and the spatio-temporal topology map of the irregularly shaped parking lots are aggregated through the federated transfer learning framework to obtain the target feature extraction model parameters, including: Step 131: Input the occupancy status data of each irregularly shaped parking lot and the spatio-temporal topology map of the irregularly shaped parking lot into the corresponding local node in the federated transfer learning framework, so as to store the occupancy status data of the corresponding irregularly shaped parking lot and the spatio-temporal topology map of the irregularly shaped parking lot in each local node, and one local node corresponds to one irregularly shaped parking lot.

[0045] Among them, the local node is a distributed computing unit in the federated transfer learning framework, and each node is responsible for storing and processing the data of one irregularly shaped parking lot to ensure data privacy and security.

[0046] In practical applications, each parking lot is deployed with a corresponding edge computing node and configured with a corresponding encrypted communication module. Embodiments of the present application can also isolate the data processing environments of different parking lots through virtualization technology. The central server issues initial model parameters, and each node loads local models with the same structure.

[0047] Step 132: In each local node, extract the geometric parameters of the spiral ramp in the spatio-temporal topology map of the special-shaped parking lot. The geometric parameters of the spiral ramp include lane curvature.

[0048] Among them, the geometric parameters of the spiral ramp are used to describe the geometric characteristics of the spiral ramp in the parking lot, including lane curvature, slope, length, etc., and are used to evaluate the difficulty and safety of vehicle driving. Lane curvature is a parameter used to describe the degree of bending of the ramp, which directly affects the turning radius and driving speed of the vehicle.

[0049] Embodiments of the present application can extract ramp geometric data through an analysis tool, fit the center line arc of the ramp by the least squares method, calculate the radius of curvature, and normalize the curvature, slope, and length to the range of [0, 1]. The center line of the spiral ramp in Parking Lot A is fitted to obtain a radius of curvature r = 3.2 m ( ), a slope of 12°, and a length of 50 m. After normalization, it is encoded as a vector [0.3125, 0.6, 0.5].

[0050] Step 133: Based on the parking space occupancy status data and the geometric parameters of the spiral ramp, train the local feature extraction model on the local node to obtain local feature extraction model parameters related to the lane curvature.

[0051] Among them, the local feature extraction model is a machine learning model (such as a convolutional neural network or a graph neural network) used to extract features from the parking space occupancy status data and geometric parameters. The local feature extraction model parameters are the weight and bias parameters obtained after training the local feature extraction model, and are used to describe the relationship between the lane curvature and the parking space occupancy status.

[0052] In embodiments of the present application, the model architecture on each node is a 3-layer graph attention network, with 128 hidden units in each layer, and the activation function is ReLU. The time series data of parking space occupancy in the past 7 days (at 5-minute intervals) is concatenated with the ramp parameters as the graph node features, and then used as the training data. An optimizer with a learning rate of 0.001 is used to train for 50 rounds until convergence. Exemplarily, the local model of Parking Lot A can be trained to obtain parameters including the attention layer weights and the output layer bias, and the output layer bias is used to adjust the prediction value baseline.

[0053] Step 134: In the central server within the federated transfer learning framework, aggregate all local feature extraction model parameters to generate global feature extraction model parameters.

[0054] Among them, the global feature extraction model parameters are global model parameters generated by aggregating the model parameters of multiple local nodes, and are used to extract the common features of parking lot data. The weighted average strategy is a model parameter aggregation method that assigns weights according to the data volume, data quality, or model performance of each local node to ensure the robustness of the global model. Exemplarily, parameters such as convolutional kernel weights and attention coefficients are weighted and averaged according to the weights to generate global parameters.

[0055] Step 135: Perform convergence processing on the global feature extraction model parameters to obtain the target feature extraction model parameters.

[0056] Among them, the convergence processing is to make the model parameters reach a stable state through iterative optimization to ensure the generalization ability and performance of the model. The target feature extraction model parameters are the final model parameters after convergence processing and are used to support dynamic parking space allocation and path planning.

[0057] The following is a specific example: Suppose there are 3 irregular parking lots (A, B, C) in a certain city. The topological structure of each parking lot is complex, including multiple spiral ramps and narrow turning areas. The parking space occupancy status data and spatio-temporal topology maps of parking lots A, B, and C are respectively input into 3 local nodes of the federated transfer learning framework, and each node stores the data corresponding to the parking lot. In each local node, the spiral ramp geometric parameters (such as lane curvature, slope) in the spatio-temporal topology map are extracted. Based on the parking space occupancy status data and spiral ramp geometric parameters, a local feature extraction model is trained to obtain local model parameters related to lane curvature. In the central server, the model parameters of the 3 local nodes are aggregated using the weighted average strategy to generate global feature extraction model parameters. Convergence processing is performed on the global feature extraction model parameters to obtain the target feature extraction model parameters, which are used to support dynamic parking space allocation and path planning.

[0058] By executing steps 131 to 135, the embodiment of the present application realizes the distributed training and collaborative optimization of multi-irregular parking lot data by introducing the federated transfer learning framework, effectively solving the data island problem in traditional centralized learning. By fusing the spiral ramp geometric parameters (such as lane curvature, slope) with the real-time parking space occupancy status data, the system can accurately capture the dynamic features and spatial constraints of the parking lot, and then generate high-precision target feature extraction model parameters. These parameters can not only support the real-time optimization of the dynamic parking space allocation strategy, but also realize intelligent path planning in combination with the spatio-temporal topology map, improving the parking efficiency and resource utilization rate. Especially during peak hours or in emergency scenarios, the system effectively alleviates the parking lot congestion problem by adaptively adjusting the parking space allocation and navigation path, providing a more efficient, intelligent, and convenient parking service experience for drivers, while ensuring data privacy and security.

[0059] In a possible embodiment, step 134, aggregating all local feature extraction model parameters to generate global feature extraction model parameters, includes: Step a1, performing a fusion process on all local feature extraction model parameters using a weighted average strategy to generate intermediate feature extraction model parameters.

[0060] Among them, the weighted average strategy is a model parameter aggregation method that calculates the weighted average by assigning weights to the model parameters of each local node to generate global model parameters. The local feature extraction model parameters are the model weights and bias parameters obtained after training at each local node and are used to describe the features of local data. The intermediate feature extraction model parameters are the preliminary global model parameters generated by the weighted average strategy and have not been verified and optimized yet.

[0061] Step a2, where the weighted average strategy includes a weight assignment scheme determined based on the data volume, data quality, or model performance metrics of each local node.

[0062] Among them, the data volume is based on the data scale of the parking space occupancy status data and the spatio-temporal topology map stored in the local node. The data quality is evaluated according to the accuracy, integrity, and timeliness of the data, usually through data cleaning and verification metrics. The model performance metrics refer to the performance of the local model on the validation set, such as accuracy, recall, F1 score, etc. The weight assignment scheme assigns higher weights to local nodes with larger data volumes. For example, if the data volume of node A is twice that of node B, then the weight of node A is twice that of node B. The data quality is evaluated through data cleaning and verification metrics (such as the proportion of missing values, the proportion of outliers), and higher-quality nodes are assigned higher weights. Weights are assigned according to the performance of the local model on the validation set (such as accuracy), and nodes with better performance are assigned higher weights. Finally, the weight assignment scheme is normalized to ensure that the sum of all weights is 1.

[0063] Step a3, verifying the intermediate feature extraction model parameters and determining the verified intermediate feature extraction model as the global feature extraction model parameters.

[0064] Among them, verification is to evaluate the performance of the intermediate feature extraction model parameters through a validation set to ensure their generalization ability and robustness. Verifying the intermediate feature extraction model parameters is to evaluate the performance of the intermediate model by using an independent validation set (data not participating in training), and calculating metrics such as accuracy, recall, F1 score, etc. If the model performance does not reach the preset threshold, return to step a1, adjust the weight assignment scheme or re-aggregate the local model parameters. If the model performance meets the standard, then determine the intermediate feature extraction model parameters as the global feature extraction model parameters for subsequent dynamic parking space allocation and path planning.

[0065] Exemplarily, by fusing multi-node features through weighted averaging and combining dynamic weight allocation of data quality and model performance, the system generates intermediate parameters that take into account both global commonalities and local characteristics. Then, through the stress test of an independent validation set, a globally robust model is selected. For example, in a certain federated learning, the accuracy rates of the local models of 3 parking lots are 92%, 85%, and 90% respectively. The central server generates intermediate parameters through weighted averaging (weights 0.4, 0.2, 0.4), and the accuracy rate of the validation set is increased to 93%. Finally, it supports the improvement of the parking space allocation accuracy in the spiral ramp area of Parking Lot C by 20% and the improvement of the vehicle diversion efficiency during peak hours by 35%.

[0066] The following is a specific example: Suppose there are 3 irregular-shaped parking lots (A, B, C) in a certain city, and the data volume and model performance of each parking lot are as follows: Parking Lot A: data volume of 1000 pieces, model accuracy rate of 95%. Parking Lot B: data volume of 800 pieces, model accuracy rate of 90%. Parking Lot C: data volume of 1200 pieces, model accuracy rate of 92%. The weighted average strategy is used to fuse the model parameters of 3 local nodes. Weights are allocated based on the data volume and model accuracy rate. For example, the weight of Parking Lot A is 0.35, the weight of Parking Lot B is 0.25, and the weight of Parking Lot C is 0.40. Verify the intermediate feature extraction model parameters. If the accuracy rate on the validation set reaches 93%, then determine it as the global feature extraction model parameters.

[0067] By executing steps a1~a3, the present application embodiment realizes the efficient fusion of local model parameters of multiple irregular-shaped parking lots through the weighted average strategy, combines the data volume, data quality, and model performance indicators, and generates high-precision global feature extraction model parameters. After verification and optimization, these parameters can effectively support dynamic parking space allocation and path planning, improve parking efficiency and resource utilization rate, and perform excellently especially during peak hours or in case of emergencies, providing a more intelligent and convenient parking experience for drivers. In a possible embodiment, S14, combining the time-varying characteristics of the channel traffic of the irregular-shaped parking lots in all irregular-shaped parking lot spatio-temporal topology maps with the target feature extraction model parameters, generates a dynamic parking space allocation strategy, including: Step 141, using a pre-trained machine learning algorithm, extract the time-varying characteristics of the channel traffic of the corresponding irregular-shaped parking lots from the spatio-temporal topology maps of the irregular-shaped parking lots. The time-varying characteristics include the vehicle traffic data of each channel in the corresponding irregular-shaped parking lot at different time periods.

[0068] Among them, the time-varying characteristics of the channel traffic of the irregular-shaped parking lots include the vehicle traffic data of each channel at different time periods, which are used to describe the traffic pattern and dynamic change trend of the parking lot.

[0069] Step 142: Input the time-varying characteristics of the traffic flow in all irregular parking lots into the target feature extraction model with the target feature extraction model parameters to obtain the integrated spatio-temporal feature representation, which is used to describe the traffic flow patterns and their dynamic change trends of each irregular parking lot.

[0070] Among them, the integrated spatio-temporal feature representation integrates the time-varying characteristics of the traffic flow in each irregular parking lot to form a unified spatio-temporal feature representation, which is used to describe the traffic flow patterns and their dynamic change trends of each parking lot.

[0071] Step 143: Obtain the real-time traffic flow data at the current moment, and combine it with the integrated spatio-temporal feature representation to predict the distribution of available parking spaces in each irregular parking lot in the future time period, and obtain the available parking space distribution map, which contains the location information and the estimated available time of each available parking space.

[0072] Among them, the real-time traffic flow data is the vehicle traffic flow data of each channel at the current moment. The available parking space distribution map contains the location information and the estimated available time of each available parking space. Generating the available parking space distribution map is to generate an available parking space distribution map containing the location of each available parking space and the estimated available time according to the prediction result.

[0073] Step 144: Based on the available parking space distribution map, combine the geometric parameters of the spiral ramp to calculate the accessibility score value of each available parking space.

[0074] Among them, the accessibility score value is based on the location of the available parking space and the geometric parameters of the spiral ramp to calculate the accessibility score value of each available parking space, which is used to evaluate the difficulty of a vehicle reaching the available parking space. Parameter acquisition is to obtain the geometric parameters of the spiral ramp from the parking lot design drawings or the management system. Accessibility calculation is to combine the available parking space distribution map and the geometric parameters of the spiral ramp, and use an algorithm to calculate the accessibility score value of each available parking space.

[0075] Step 145: According to the accessibility score value of each available parking space, perform load balancing analysis on all irregular parking lots to obtain a list of candidate parking spaces for the vehicles to be parked.

[0076] Among them, the list of candidate parking spaces is to screen out the list of candidate parking spaces for the vehicles to be parked according to the accessibility score value. The accessibility score value is to sort all available parking spaces according to the accessibility score value to obtain a list of available parking spaces with scores from high to low. Load balancing is to further screen the list of candidate parking spaces in combination with the current load conditions of each parking lot (such as the number of parked vehicles, traffic congestion conditions, etc.) to ensure load balancing.

[0077] Step 146: Obtain the driving trajectory information of the vehicle to be parked, and generate a dynamic parking space allocation strategy according to the driving trajectory information and the list of candidate parking spaces for the vehicle to be parked.

[0078] Among them, the generation of the dynamic parking space allocation strategy is to generate a dynamic parking space allocation strategy based on the optimal parking path, including specific parking space locations and estimated arrival times.

[0079] Exemplarily, input the channel traffic data every 15 minutes in the past 7 days, use the long short-term memory network to capture periodic patterns, such as the weekday morning peak traffic is 20% higher than that on weekends, and output time-encoded features, such as peak period label vectors. Concatenate the outputs of the long short-term memory networks of each parking lot with geometric parameters such as lane curvature and slope, and input them into the graph convolutional network to generate a 128-dimensional feature vector to represent the global traffic law and spatial constraints. Input the real-time channel traffic (such as 3 vehicles per minute at the current entrance) and the spatio-temporal feature vector into the temporal convolutional network, output the free probability and duration of each parking space, and generate a distribution map in the form of a heat map. In the path from the entrance to the target parking space, accumulate the reciprocal of the curvature and the slope ratio of each ramp. The smaller the curvature (the sharper the turn) or the larger the slope, the lower the score. According to the formula: Sort. If the load rate of parking lot A is 90% and the load rate of parking lot C is 70%, then preferentially allocate high-score parking spaces in parking lot C. According to the real-time vehicle position, the candidate parking space list, and the real-time topological map (such as sudden congestion in channel C2), the model selects the optimal path with the shortest time consumption and a score ≥ 3.5, and pushes an instruction through the navigation system: "Turn right into ramp P1 and go straight for 50 meters to reach parking space C-08".

[0080] The following is a specific example: In the irregular parking lot of a large commercial complex, the system needs to predict the distribution of free parking spaces in each parking lot in real time and allocate the best parking spaces for the vehicles entering the parking lot. Through sensors installed in each channel, vehicle traffic data is collected in real time and cleaned and preprocessed. Use a pre-trained model to extract the time-varying features of channel traffic from the spatio-temporal topological map and integrate them into a unified spatio-temporal feature representation. Combine the real-time channel traffic data and the integrated spatio-temporal feature representation, and use a deep learning model to predict the distribution of free parking spaces in the future time period. According to the free parking space distribution map and the spiral ramp geometric parameters, calculate the accessibility score value of each free parking space. Combine the current load conditions of each parking lot to screen out the candidate parking space list for the vehicle to be parked. By obtaining the driving trajectory information of the vehicle to be parked, generate a specific parking space allocation strategy to guide the vehicle to park in the best free parking space.

[0081] By executing steps 141 to 146, the system of the embodiment of the present application can predict the distribution of free parking spaces in each irregular parking lot in real time, and dynamically allocate the best parking spaces according to the driving trajectory and parking needs of the vehicle. This not only improves the utilization rate of the parking lot, reduces the time for vehicles to search for parking spaces, but also effectively avoids congestion in the parking lot and improves the user experience.

[0082] In a possible embodiment, in step 133, based on the parking space occupancy status data and the spiral ramp geometric parameters, train the local feature extraction model on the local node to obtain the local feature extraction model parameters related to the lane curvature, including: Step b1: Initialize the local feature extraction model.

[0083] Among them, initialization is to assign initial values to the weight and bias parameters of the model. Usually, random initialization or pre-trained model parameters are used for initialization to ensure the stability in the initial stage of training. The initialized model provides a basis for subsequent training.

[0084] In the embodiment of the present application, the local feature extraction model can be built using the PyTorch framework. The model structure is a 3-layer fully connected network (input layer: 256 dimensions, hidden layer: 128 dimensions, output layer: 64 dimensions). The activation function uses ReLU, and the output layer is connected to the Sigmoid function to generate the parking space occupancy probability. The hidden layer is initialized using the Xavier normal distribution to avoid gradient disappearance. The bias is initialized to zero. If there is a historical model, download the pre-trained parameters (such as the feature extraction layer of ResNet-18) from the central server, and freeze the bottom layer parameters and only fine-tune the top layer.

[0085] Exemplarily, after the local model of parking lot A is initialized, the input layer receives 256-dimensional features (128-dimensional parking space status time series + 128-dimensional ramp geometric parameters). The dimension of the weight matrix of the hidden layer is 128×256, and the dimension of the bias vector is 128×1.

[0086] Step b2: Based on the parking space occupancy status data and the spiral ramp geometric parameters, construct the input data set of the local feature extraction model.

[0087] Among them, the input data set is to perform feature engineering processing on the parking space occupancy status data and the spiral ramp geometric parameters to construct the input data required for model training. The construction of the input data set is to extract the parking space occupancy status data from the local node, convert it into time series features (such as the change in parking space occupancy in the past 1 hour). Extract the spiral ramp geometric parameters and perform normalization processing on them (such as normalizing to the range [0,1]). Concatenate the parking space occupancy status data and the spiral ramp geometric parameters to form the input feature vector of the model. Exemplarily, read the parking space status in the past 1 hour from the local database (1 record per second), and count the occupancy rate change in a 5-minute window to generate a 128-dimensional time series vector (such as the occupancy rate drops from 80% to 70% within the window, encoded as 0.8, 0.78,..., 0.7). Extract the curvature , Gradient (5° - 15°) and length (20 - 100m) are normalized to [0, 1] through MinMax and concatenated into a 128 - dimensional vector. For example, curvature 0.3 is converted to 0.3 / 0.5 = 0.6, and gradient 12° is converted to 12 / 15 = 0.8. The parking space status vector and the ramp parameter vector are concatenated by column to form a 256 - dimensional input feature. Then, the input dataset is divided into a training set, a validation set, and a test set in the ratio of 70%:15%:15%. Another example, among 1000 data of parking lot A, 700 are used for training, 150 for validation, and 150 for testing. A single data sample: input feature = [0.8, 0.78,..., 0.7 (time series), 0.6, 0.8,... (ramp parameters)], label = 1 (parking space occupied).

[0088] Step b3: Input the input dataset into the initialized local feature extraction model, and obtain the local feature extraction model parameters related to the lane curvature through training.

[0089] Among them, training adjusts the model parameters through an optimization algorithm (such as the gradient descent method) to minimize the loss function (such as mean squared error or cross - entropy loss). Specifically, in the training process, the input dataset is input into the initialized local feature extraction model, and the model calculates the predicted value through forward propagation. Calculate the loss between the predicted value and the true value (such as using the mean squared error loss function), calculate the gradient through the backpropagation algorithm, and update the model parameters using an optimizer (such as the Adam optimizer). Repeat the above process until the performance of the model on the validation set reaches the preset threshold or the training round ends. Finally, the local feature extraction model parameters related to the lane curvature are obtained, and these parameters will be uploaded to the central server for aggregation.

[0090] Exemplarily, after 50 rounds of training of the local model in parking lot A, the accuracy of the validation set reaches 95%, and the loss function converges to 0.12. The extracted model parameters include: some values of the first - layer weight W1 are 0.32, - 0.15, 0.28,..., showing the negative impact of the lane curvature parameters (the last 128 dimensions of the input feature) on the activation value of the hidden layer. The output - layer bias b3 = 0.18, which is used to adjust the prediction baseline.

[0091] The following is a specific example: Assume that the parking space occupancy status data of Parking Lot A includes the following: The parking space identity identifier (Identifier, ID) is 001, the status is idle, the timestamp is: 2023-10-01, 12:00:00, the parking space ID is 002, the status is occupied, the timestamp is 2023-10-01, 12:00:00. The spiral ramp geometric parameters of Parking Lot A include the following: The ramp ID is 001, the lane curvature is 0.5, the slope is 10%, and the length is 50 meters. Initialize the local feature extraction model (such as a neural network), construct the input data set using the initialization method, splice the parking space occupancy status data and the spiral ramp geometric parameters to form a feature vector. Input the input data set into the model, and obtain the local feature extraction model parameters related to the lane curvature through training.

[0092] By performing steps b1~b3, the solution of this embodiment of the present application initializes the local feature extraction model, constructs the input data set and conducts training, generating local feature extraction model parameters related to the lane curvature. These parameters can effectively capture the dynamic features and spatial constraints of the parking lot data, provide a high-quality basis for subsequent federated learning and global model optimization, and improve the intelligent level and efficiency of parking management.

[0093] In a possible embodiment, S11. Obtain the parking space occupancy status data of multiple irregularly shaped parking lots, including: Step 111. Obtain the pressure distribution data of each parking space in the corresponding irregularly shaped parking lot collected by each flexible pressure film sensor array in real time. One irregularly shaped parking lot corresponds to one flexible pressure film sensor array. Flexible pressure film sensors are arranged on each parking space in the irregularly shaped parking lot, and flexible pressure film sensors are also arranged on the curved lanes in the irregularly shaped parking lot. The flexible pressure film sensor array is arranged in a conforming manner according to the parking space contour and the curved lane contour of the irregularly shaped parking lot.

[0094] Among them, the flexible pressure film sensor array is an array composed of multiple flexible pressure film sensors for detecting the pressure distribution. Each sensor can be bent and attached to the parking spaces and curved lanes of the irregularly shaped parking lot. The pressure distribution data is the pressure data of each parking space and lane collected by the sensors in real time, reflecting the vehicle's position and weight distribution. The sensor arrangement is to install flexible pressure film sensors on each parking space and curved lane of the irregularly shaped parking lot to ensure that the sensor array is arranged in a conforming manner according to the contours of the parking spaces and lanes. The data collection is to collect the pressure data of the sensors in real time and transmit it to the central data processing unit through wireless communication.

[0095] Step 112. Convert the pressure distribution data of the parking space into parking space occupancy status data.

[0096] Among them, the parking space occupancy status data represents the status information of whether each parking space is currently occupied, usually represented in binary form (0 indicates free, 1 indicates occupied). Data conversion is to convert the judgment result into parking space occupancy status data and store it in the database for subsequent use.

[0097] The following is a specific example: In a special-shaped parking lot of a large commercial complex, the system needs to monitor the occupancy status of each parking space in real time for parking space management and navigation. Flexible pressure film sensors are installed on each parking space and curved lanes in the parking lot to ensure that the sensor array is laid in a conforming manner according to the contours of the parking spaces and lanes. The sensors collect pressure data in real time and transmit it to the central data processing unit through wireless communication. The central unit cleans and preprocesses the received data to remove noise and outliers. The central processing unit analyzes the pressure distribution data of each parking space, sets a pressure threshold, and determines whether each parking space is occupied. The judgment result is converted into parking space occupancy status data and stored in the database for use by the parking space management and navigation system.

[0098] By executing steps 111 to 112, the embodiment of the present application enables the system to obtain the pressure distribution data of each parking space in the special-shaped parking lot in real time and convert it into parking space occupancy status data. This not only improves the efficiency and accuracy of parking space management, but also provides real-time parking space navigation services for drivers, enhancing the user experience.

[0099] In a possible embodiment, step 112, converting the pressure distribution data of the parking space into parking space occupancy status data, includes: Step c1, analyzing the pressure distribution data according to a preset parsing rule to obtain vehicle parking position and weight distribution information.

[0100] Among them, the preset parsing rules can be a set of predefined data processing rules and algorithmic logics for extracting vehicle parking position and weight distribution information from the original pressure data. The rule composition can include data filtering rules, outlier removal rules, coordinate transformation rules, pressure peak detection rules, weight calculation rules, etc. Among them, the data filtering rules can be used to filter invalid signals (such as instantaneous pressures generated by environmental vibrations and pedestrian tramplings); the outlier removal rules can refer to identifying abnormal data through standard deviation analysis or box plot method; the coordinate transformation rules can refer to mapping the physical coordinates of the sensor array to the global coordinate system of the parking lot; the pressure peak detection rules can refer to identifying local pressure maximum points through a sliding window (such as a 5×5 grid); the weight calculation rules can refer to calculating the total vehicle weight based on pressure integration and sensor calibration parameters. The vehicle parking position is the coordinate of the vehicle center point (such as X / Y axis coordinates) located by the sensor array coordinates. The weight distribution information is the pressure data distribution map of the vehicle exerted on the sensor, reflecting the wheel positions and weight concentration areas. The weight estimation is to sum the pressure values within the pressure distribution area through an integration algorithm and convert it into the total vehicle weight in combination with the sensor sensitivity parameter (such as 0.5V / N). The position parsing is based on a pressure peak detection algorithm (such as a convolutional neural network), identifying the local maximum points in the pressure distribution map and calculating the vehicle center position in combination with the sensor grid coordinates.

[0101] Step c2: Generate a parking space occupancy identifier based on the vehicle parking position and weight distribution information.

[0102] Among them, the parking space occupancy identifier is a data structure containing binary status (0 / 1), timestamp, and confidence score, used to indicate whether the parking space is occupied. The weight status determination adopts a dynamic threshold method (such as the mean of the historical data sliding window ± 3σ). When the weight exceeds the threshold (such as >50kg) and the duration is >5 seconds, it is marked as occupied. The generation process of the parking space occupancy identifier is as follows: Match the corresponding parking space ID according to the vehicle parking position. Determine the vehicle type (such as small vehicle, large vehicle) based on the weight distribution information, and generate the parking space occupancy identifier in combination with the preset rules. For example, if the weight exceeds the threshold, it is marked as "occupied by large vehicle". Store the parking space occupancy identifier in the parking space status database for subsequent use.

[0103] Step c3: Combine the parking space occupancy identifier with the vehicle weight distribution information to generate parking space occupancy status data.

[0104] Among them, the parking space occupancy status data is in the form of a structured data packet, including dimensions such as occupancy status, weight distribution heat map, and time series change curve. The combination of the parking space occupancy identifier and the vehicle weight distribution information is achieved by spatio-temporal alignment, binding the occupancy identifier and the weight data through the timestamp and the parking space number to ensure the matching of the two types of data at the same moment and the same parking space. For example, when the occupancy identifier of parking space B-07 is 1, the corresponding weight data needs to match the sensor acquisition range of this parking space, and is structured and encapsulated to generate a data packet containing multiple-dimensional fields, such as "parking space number": "B-07", "occupancy status": 1, "confidence level": 98%, "total weight": 2100 kg, "weight distribution heat map": "Base64-encoded matrix data", "timestamp": "2025-03-24T10:15:30". Then, through dynamic verification, the reliability of the occupancy identifier is verified through the weight data. For example, if the identifier is "occupied" but the weight is lower than the threshold (such as <50 kg), an anomaly detection mechanism is triggered. The deep combination of the parking space occupancy identifier and the weight distribution information can avoid misjudgment (such as a shopping cart parked briefly), and the accuracy rate is increased to over 99%. Combining the weight data to optimize the vehicle transfer strategy (such as the logic of the skateboard device transferring the vehicle to the conveyor belt in Web Page 1); storing the complete weight heat map to assist in determining the liability for accidents (such as restoring the vehicle position when a scrape occurs).

[0105] The following is a specific example: A three-dimensional parking garage at an airport uses a conformable flexible pressure sensor array (resolution 5 cm × 5 cm) to collect pressure data of 2000 special-shaped parking spaces in real time, with a sampling frequency of 10 Hz. The system uses a preprocessing module to filter out temperature interference (adapting to extreme environments from -20°C to 60°C), and accurately identifies the pressure distribution pattern of four-wheel vehicles through a model, outputting structured data in a lightweight data exchange format (such as vehicle coordinates (x = 3.2 m, y = 5.7 m) and weight 1.8 t). The dynamic threshold algorithm combines the pressure distribution characteristics to determine that parking space A12 in Area B is in the occupied state (confidence level 98%), generating a scalable vector icon with a timestamp to cover the electronic map in real time, supporting visual tracking of the parking space status. The data integration module constructs a comprehensive data packet containing a 256-color scale heat map, occupancy duration, and vehicle weight distribution, and pushes it to the central dispatching system through a protocol, with the communication delay controlled within 200 ms, meeting the high real-time requirements. Through multi-dimensional data fusion and low-latency transmission, this system realizes the accurate monitoring and efficient dispatching of special-shaped parking spaces.

[0106] By performing steps c1 to c3, the embodiments of the present application achieve accurate extraction of vehicle parking positions and weight distribution information through the analysis and processing of pressure distribution data, and generate real-time parking space occupancy status data in combination with parking space occupancy identifiers. This data can effectively support dynamic parking space allocation and route planning, improve the intelligent level and efficiency of parking management, and perform particularly well during peak hours or in case of emergencies, providing a more intelligent and convenient parking experience for drivers.

[0107] Figure 2 FIG. is a schematic structural diagram of a prediction system for vacant parking spaces in a parking lot provided by an embodiment of the present application, as Figure 2 shown, the system includes: An acquisition module 21, configured to acquire parking space occupancy status data and special-shaped parking lot topological structure data of multiple special-shaped parking lots.

[0108] A generation module 22, configured to generate a spatio-temporal topological map of a special-shaped parking lot for each special-shaped parking lot by fusing ramp curvature parameters extracted from the special-shaped parking lot topological structure data and historical vehicle turning radius data.

[0109] An extraction module 23, configured to aggregate the parking space occupancy status data and the spatio-temporal topological map of all special-shaped parking lots through a federated transfer learning framework to obtain target feature extraction model parameters.

[0110] A generation module 24, configured to generate a dynamic parking space allocation strategy by combining the time-varying characteristics of the traffic flow in the channels of the special-shaped parking lots in the spatio-temporal topological maps of all special-shaped parking lots and the target feature extraction model parameters.

[0111] A mapping module 25, configured to map the dynamic parking space allocation strategy to the real-time navigation path in the spatio-temporal topological map of the special-shaped parking lot, so that the vehicle to be parked can reach the vacant parking space through the real-time navigation path.

[0112] Figure 2 The described prediction method system for vacant parking spaces in a parking lot can execute Figure 1 the prediction method for vacant parking spaces in a parking lot described in the embodiments shown, and its implementation principle and technical effects will not be elaborated again. For the prediction system for vacant parking spaces in a parking lot in the above embodiments, the specific manners in which each module and unit perform operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0113] In a possible design, Figure 2 the prediction system for vacant parking spaces in a parking lot in the embodiments shown can be implemented as a computing device, as Figure 3 shown, the computing device may include a storage component 31 and a processing component 32.

[0114] The storage component 31 stores one or more computer instructions, and the one or more computer instructions are called and executed by the processing component 32.

[0115] The processing component 32 is configured to: obtain the parking space occupancy status data and the special-shaped parking lot topology data of multiple special-shaped parking lots; for each special-shaped parking lot, generate a special-shaped parking lot spatio-temporal topology map by fusing the ramp curvature parameters extracted from the special-shaped parking lot topology data and the historical vehicle turning radius data; aggregate the parking space occupancy status data and the special-shaped parking lot spatio-temporal topology map of all special-shaped parking lots through a federated transfer learning framework to obtain the target feature extraction model parameters; combine the time variation characteristics of the channel flow of the special-shaped parking lots in all special-shaped parking lot spatio-temporal topology maps and the target feature extraction model parameters to generate a dynamic parking space allocation strategy; map the dynamic parking space allocation strategy to the real-time navigation path of the special-shaped parking lot spatio-temporal topology map, so that the vehicle to be parked can reach the idle parking space through the real-time navigation path.

[0116] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0117] The storage component 31 is configured to store various types of data to support the operations of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Random Access Memory (RAM), Static Random-Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), magnetic memory, flash memory, magnetic disks or optical discs.

[0118] Of course, the computing device may necessarily further include other components, such as input / output interfaces, display components, communication components, etc.

[0119] The input / output interface provides an interface between the processing component and the peripheral interface module, and the peripheral interface module may be an output device, an input device, etc.

[0120] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0121] Among them, the computing device may be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device may refer to a cloud server, and the above-mentioned processing component, storage component, etc. may be basic server resources leased or purchased from a cloud computing platform.

[0122] The embodiment of the present application further provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it can implement the above-mentioned Figure 1 prediction method for idle parking spaces in a parking lot shown in the embodiment.

[0123] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments and will not be described herein again.

[0124] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative work.

[0125] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or equivalently replace some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A method for predicting vacant parking spaces in a parking lot, characterized in that: include: Obtaining parking space occupancy status data and topological structure data of multiple special-shaped parking lots; For each special-shaped parking lot, a spatiotemporal topological map of the special-shaped parking lot is generated by fusing the ramp curvature parameters extracted from the topological structure data of the special-shaped parking lot with the historical vehicle turning radius data; Aggregating the parking space occupancy status data of all the special-shaped parking lots and the spatiotemporal topology map of the special-shaped parking lots through a federated transfer learning framework to obtain target feature extraction model parameters; Combining the time-varying characteristics of the channel flow of all the irregular parking lots in the spatiotemporal topology of the irregular parking lots with the target feature extraction model parameters, a dynamic parking space allocation strategy is generated; The dynamic parking space allocation strategy is mapped to the real-time navigation path of the spatiotemporal topological map of the special-shaped parking lot, so that the vehicles to be parked can reach the vacant parking spaces through the real-time navigation path.

2. The method according to claim 1, characterized in that The method of aggregating the parking space occupancy status data of all the special-shaped parking lots and the spatiotemporal topology map of the special-shaped parking lots through the federated transfer learning framework to obtain target feature extraction model parameters includes: Inputting the parking space occupancy status data of each of the special-shaped parking lots and the spatiotemporal topology map of the special-shaped parking lot into the corresponding local node in the federated transfer learning framework, so as to store the parking space occupancy status data and the spatiotemporal topology map of the corresponding special-shaped parking lot in each of the local nodes, and one local node corresponds to one special-shaped parking lot; In each of the local nodes, extracting geometric parameters of the spiral ramp in the spatiotemporal topological map of the special-shaped parking lot, the geometric parameters of the spiral ramp including the curvature of the lane; Based on the parking space occupancy status data and the spiral ramp geometric parameters, training a local feature extraction model on a local node to obtain local feature extraction model parameters related to lane curvature; In a central server within the federated transfer learning framework, aggregating all the local feature extraction model parameters to generate global feature extraction model parameters; The global feature extraction model parameters are subjected to convergence processing to obtain target feature extraction model parameters.

3. The method according to claim 2, characterized in that The aggregating all the local feature extraction model parameters to generate global feature extraction model parameters includes: A weighted average strategy is used to fuse all the local feature extraction model parameters to generate intermediate feature extraction model parameters; wherein the weighted average strategy includes a weight allocation scheme determined based on the data volume, data quality or model performance index of each local node; The intermediate feature extraction model parameters are verified, and the verified intermediate feature extraction model is determined as the global feature extraction model parameters.

4. The method according to claim 1, characterized in that: The method combines the time variation characteristics of the channel flow of all the irregular parking lots in the spatiotemporal topology of the irregular parking lots with the target feature extraction model parameters to generate a dynamic parking space allocation strategy, including: Using a pre-trained machine learning algorithm, extract the time-varying characteristics of the channel flow of the corresponding special-shaped parking lot from the spatiotemporal topology map of each special-shaped parking lot, wherein the time-varying characteristics include the vehicle flow data of each channel in the corresponding special-shaped parking lot in different time periods; Input the time variation characteristics of the channel flow of all special-shaped parking lots into the target feature extraction model using the target feature extraction model parameters to obtain an integrated spatiotemporal feature representation, which is used to describe the flow pattern of each special-shaped parking lot and its dynamic change trend; Acquire the real-time channel flow data at the current moment, combine it with the integrated spatiotemporal feature representation, predict the distribution of free parking spaces in each special-shaped parking lot in the future time period, and obtain a free parking space distribution map, wherein the free parking space distribution map includes the location information and the estimated available time of each free parking space; Based on the free parking space distribution map and in combination with the spiral ramp geometric parameters, calculating the accessibility score value of each free parking space; According to the accessibility score of each vacant parking space, a load balancing analysis is performed on all the special-shaped parking lots to obtain a list of candidate parking spaces for vehicles to be parked; Acquire the driving trajectory information of the vehicle to be parked, and generate a dynamic parking space allocation strategy according to the driving trajectory information and the candidate parking space list of the vehicle to be parked.

5. The method according to claim 2, characterized in that: The method of training a local feature extraction model on a local node based on the parking space occupancy status data and the spiral ramp geometric parameters to obtain local feature extraction model parameters related to lane curvature includes: Initialize the local feature extraction model; Based on the parking space occupancy status data and the spiral ramp geometric parameters, construct an input data set for a local feature extraction model; The input data set is input into the initialized local feature extraction model, and local feature extraction model parameters related to lane curvature are obtained through training.

6. The method according to claim 1, characterized in that The method of obtaining the parking space occupancy status data of a plurality of special-shaped parking lots includes: Obtain the pressure distribution data of each parking space in the corresponding special-shaped parking lot collected in real time by each flexible pressure film sensor array. One special-shaped parking lot corresponds to one flexible pressure film sensor array. Each parking space in the special-shaped parking lot is provided with a flexible pressure film sensor, and a flexible pressure film sensor is provided on a curved lane in the special-shaped parking lot. The flexible pressure film sensor array is arranged in a fitting manner according to the contour of the parking space and the contour of the curved lane of the special-shaped parking lot. The pressure distribution data of the parking space is converted into parking space occupancy status data.

7. The method according to claim 6, characterized in that The step of converting the pressure distribution data of the parking space into parking space occupancy status data comprises: Analyze the pressure distribution data according to preset analysis rules to obtain the parking position and weight distribution information of the vehicle; Generate a parking space occupancy mark according to the vehicle parking position and the weight distribution information; The parking space occupancy mark is combined with the vehicle weight distribution information to generate parking space occupancy status data.

8. A system for predicting vacant parking spaces in a parking lot, characterized in that: include: An acquisition module, used to acquire parking space occupancy status data and topological structure data of a plurality of special-shaped parking lots; A generation module, for generating a spatiotemporal topological map of the special-shaped parking lot by fusing the ramp curvature parameters extracted from the topological structure data of the special-shaped parking lot with the historical vehicle turning radius data; An extraction module, used for aggregating the parking space occupancy status data of all the special-shaped parking lots and the spatiotemporal topology map of the special-shaped parking lots through a federated transfer learning framework to obtain target feature extraction model parameters; A generation module, for combining the time variation characteristics of the channel flow of all the irregular parking lots in the spatiotemporal topology map of the irregular parking lots with the target feature extraction model parameters to generate a dynamic parking space allocation strategy; The mapping module is used to map the dynamic parking space allocation strategy into the real-time navigation path of the spatiotemporal topological map of the special-shaped parking lot, so that the parked vehicle can reach the vacant parking space through the real-time navigation path.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for predicting vacant parking spaces in a parking lot as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, a method for predicting vacant parking spaces in a parking lot as described in any one of claims 1 to 7 is implemented.

Images