Collaborative parking system and method based on distributed wireless sensor network

Abstract

The invention provides a collaborative parking system based on a distributed wireless sensor network. The collaborative parking system comprises a distributed sensing module, a visual positioning module, a parking space distribution module, a path planning module and a joint simulation module. All the modules are connected in sequence to form a complete technical closed loop of sensing, positioning, decision making, planning and verification. According to the method, the visual SLAM based on natural features and the correction technology based on the artificial tag AprilTags are creatively fused, on the premise that the infrastructure cost is hardly increased, the problem of accurate vehicle positioning in the underground parking lot in the GPS-free environment is solved, the positioning accuracy meets the practical requirement, and the high deployment cost of a laser or UWB scheme is avoided; a parking space distribution optimization model based on saturation dynamic adjustment and an efficient integer coding particle swarm solving algorithm are provided, and global coordination of multi-vehicle parking space requirements can be completed within second-level time.

Description

Technical Field

[0001] This invention relates to the field of intelligent connected vehicles and automatic parking technology, specifically to a collaborative parking system and method based on a distributed wireless sensor network. It is applicable to closed or semi-closed scenarios with limited satellite navigation signals and dense traffic, such as large underground parking lots and shopping mall garages. The aim is to achieve high-precision vehicle positioning, intelligent parking space allocation, and efficient path planning through multimodal perception, collaborative computing, and global optimization. Background Technology

[0002] With the acceleration of urbanization and the continuous growth of car ownership, "parking difficulties" have become a common problem plaguing urban traffic. Statistics show that underground parking lots in large commercial complexes and office areas often experience a simultaneous shortage of parking spaces and internal traffic congestion during peak hours, while some spaces remain vacant due to a lack of transparency in information. Existing solutions have the following limitations: 1. High cost and limited field of vision for single-vehicle intelligent parking (AVP): Relying on ultrasonic radar, surround view cameras and even lidar on the vehicle itself, it not only significantly increases the cost per vehicle, but also has a limited perception range. It cannot obtain the overall distribution of available parking spaces and real-time traffic situation in the parking lot, which can easily lead to multiple vehicles blindly patrolling and exacerbate congestion in the parking lot.

[0003] 2. Bottlenecks exist in underground environment positioning technology: Global Navigation Satellite System (GNSS) signals are completely ineffective in underground parking lots. Alternative solutions such as Ultra Wideband (UWB) or LiDAR SLAM can provide high-precision positioning, but they suffer from high infrastructure deployment costs and complex maintenance, making large-scale deployment difficult.

[0004] 3. Simplified parking space allocation strategy: Existing parking guidance systems or basic vehicle-to-everything (V2X) applications mostly adopt simple strategies such as "nearest parking space priority" or "zone guidance", which fail to consider the overall efficiency of the system and lack consideration for users' personalized preferences (such as proximity to elevator entrance, need for charging piles, etc.).

[0005] 4. Path planning does not take into account the driving experience in the parking lot: Traditional path planning algorithms (such as Dijkstra's algorithm and the basic A* algorithm) often generate too many unnecessary right-angle turns or detours on the parking lot grid map, which not only increases the driving time and energy consumption, but also affects the smoothness and comfort of driving.

[0006] In summary, existing technologies struggle to achieve high-precision, low-cost vehicle positioning in underground parking scenarios while maintaining controllable costs, and further hinder collaborative parking decision-making and planning that considers multi-objective optimization. Therefore, there is an urgent need for a collaborative parking system integrating advanced sensing, positioning, decision-making, and planning technologies to improve the overall operational efficiency of parking lots and the user parking experience. Summary of the Invention

[0007] This invention is made to solve the above-mentioned problems, and aims to provide a cooperative parking system and method based on a distributed wireless sensor network.

[0008] This invention provides a collaborative parking system based on a distributed wireless sensor network, characterized by comprising: a distributed perception module, a visual positioning module, a parking space allocation module, a path planning module, and a co-simulation module. The distributed perception module includes cameras, physiological signal acquisition devices, eye trackers, audio acquisition equipment, and vehicle operation sensors. It is used to collect visual data of the parking lot environment, driver physiological data, driver eye movement data, in-vehicle audio data, and vehicle operation data. The visual positioning module is used to extract and match features from images captured by the camera, and combines real-time localization and mapping algorithms with AprilTags visual reference labels to achieve precise vehicle positioning within the parking lot. The parking space allocation module is used to construct an optimization model with the objective of minimizing the total parking time based on parking lot saturation and user preference information. It then uses an integer-encoded particle swarm optimization algorithm to solve the model and output the globally optimal parking space allocation scheme. The path planning module uses an improved A* search algorithm that combines a weighted heuristic function and a multi-parent node strategy to plan the optimal driving path from the current location to the assigned parking space for each vehicle, based on a grid map containing multi-layered information. The co-simulation module is used to integrate and verify the functionality and performance of the vision positioning module, berth allocation module, and path planning module in a co-simulation environment of the robot operating system and MATLAB.

[0009] The collaborative parking system based on a distributed wireless sensor network provided by this invention also has the following features: a 1080p resolution, 30fps RGB camera; a wearable electrocardiograph with a 250Hz sampling rate for acquiring physiological signals; a telemetry eye tracker with a 120Hz sampling rate and a gaze tracking error ≤1.2°; an audio acquisition device with a sampling rate of 16000Hz; and vehicle operation sensors including a steering wheel / pedal force sensor with a 500Hz sampling rate.

[0010] The cooperative parking system based on a distributed wireless sensor network provided by this invention also has the following features: the feature extraction performed by the visual positioning module includes: constructing a multi-layer Gaussian pyramid for the input image, dividing the pixel layer into grids to extract ORB feature points, and using a quadtree algorithm to filter the feature points to achieve uniform distribution; the feature matching includes: using a fast nearest neighbor search algorithm for coarse matching, and using a random sampling consensus algorithm for fine matching to eliminate mismatched points.

[0011] The collaborative parking system based on a distributed wireless sensor network provided by this invention also has the following features: In the parking space allocation module, the optimization model is a 0-1 integer programming model, the objective function of which is to minimize the total weighted parking time of all vehicles, and the constraint condition ensures that each vehicle is allocated only one parking space and each parking space is allocated to at most one vehicle; In the integer-encoded particle swarm optimization algorithm, the position of each particle represents a complete parking space allocation sequence.

[0012] The cooperative parking system based on a distributed wireless sensor network provided by this invention also has the following features: In the path planning module, the cost function of the improved A* search algorithm is f(n) = g(n) + w * h(n), where g(n) is the actual cost from the starting point to node n, h(n) is the Manhattan distance heuristic value from node n to the endpoint, and w is an adjustable weight coefficient; the multi-parent node strategy means that when expanding nodes, if there are multiple parent nodes that make g(n) equal, the parent node that makes the path straighter is selected first.

[0013] The collaborative parking system based on a distributed wireless sensor network provided by this invention also has the following features: the co-simulation module is built on the Ubuntu operating system, the robot operating system is responsible for running sensor-driven, real-time localization and map building nodes and publishing vehicle pose topics, and the MATLAB environment is responsible for subscribing to pose topics and executing parking space allocation and path planning algorithms.

[0014] The collaborative parking system based on a distributed wireless sensor network provided by this invention also has the following features: In the visual positioning module, the detection and recognition process of the AprilTags visual reference tags includes: image grayscale conversion and binarization, connected component clustering and quadrilateral fitting, Hamming distance-based encoding and decoding, and sub-pixel-level corner point positioning.

[0015] The collaborative parking system based on a distributed wireless sensor network provided by this invention also has the following features: the calculation of the total weighted parking time is dynamically adjusted based on the parking lot saturation P: when the saturation is lower than or equal to a preset threshold s, the parking time of a single vehicle includes driving time and walking time; when the saturation is higher than the threshold s, the parking time of a single vehicle only considers walking time.

[0016] The collaborative parking system based on a distributed wireless sensor network provided by this invention also includes a data preprocessing and fusion unit, which performs time synchronization, filtering and noise reduction, and feature-level fusion on the multimodal data collected by the distributed sensing module, providing unified and high-quality input data for subsequent modules.

[0017] The present invention also provides a cooperative parking method based on a distributed wireless sensor network, applicable to the system of any one of claims 1 to 9, characterized by comprising the following steps: S1: Synchronously collect multimodal data of the parking lot environment and vehicles through distributed sensing modules; S2: The visual positioning module processes visual data to achieve accurate vehicle positioning on the parking lot map; S3: The parking space allocation module calculates and outputs the optimal parking space allocation scheme based on the real-time parking space status, vehicle location, and user preferences; S4: The route planning module plans the optimal driving route for each vehicle that has been assigned a parking space. S5: Verify and evaluate the overall system performance under different parking lot saturation scenarios in the co-simulation module.

[0018] The role and effect of invention The cooperative parking system and method based on a distributed wireless sensor network according to the present invention achieves the following significant beneficial effects through the implementation of the above technical solutions: 1. Achieved high-precision, low-cost indoor positioning: Creatively integrates visual SLAM based on natural features with correction technology based on artificial tags AprilTags, solving the problem of accurate vehicle positioning in underground parking lots without GPS without increasing infrastructure costs. The positioning accuracy meets the requirements for practical application and avoids the high deployment costs of laser or UWB solutions.

[0019] 2. Achieving a balance between global optimization and user experience: The proposed parking space allocation optimization model based on dynamic saturation adjustment and the efficient integer-encoded particle swarm optimization algorithm can complete the global coordination of parking space demand for multiple vehicles within seconds. It can respect individual user preferences and improve satisfaction when parking spaces are plentiful, while prioritizing overall system traffic efficiency and reducing congestion when parking spaces are scarce, thus significantly reducing total parking time (simulations show a reduction of 15%-20%).

[0020] 3. Generated smoother and more efficient driving paths: The improved A* search algorithm balances search efficiency and path quality through a weighted heuristic function, and effectively reduces redundant inflection points in the path through a multi-parent node strategy. The planned path is more in line with the vehicle's motion characteristics, which not only shortens the travel time but also improves the smoothness and comfort of driving, while reducing the vehicle's steering losses.

[0021] 4. A flexible and reliable development and verification platform has been built: The ROS-MATLAB co-simulation architecture decomposes complex systems into loosely coupled modules, facilitating independent algorithm development, testing, and integration. This platform can simulate various extreme and typical scenarios, providing powerful tools for algorithm performance evaluation and system optimization, greatly accelerating the product development process and reliability verification.

[0022] 5. Excellent scalability and practicality: The "distributed perception + centralized collaborative decision-making" architecture reduces the intelligence requirements for individual vehicles, facilitating the retrofitting and promotion of existing vehicles and parking lots. The modular design of the system also makes it easy to integrate more functions in the future, such as linkage with parking fee collection systems and electric vehicle charging management systems, demonstrating broad prospects for industrial application. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the distributed sensing module in this invention; Figure 2 This is a flowchart illustrating the visual positioning module in this invention; Figure 3 This is a schematic diagram of the quadtree algorithm in this invention; Figure 4 This is a flowchart illustrating the improved A* search algorithm with multiple parent nodes in this invention. Detailed Implementation

[0024] To make the technical means, creative features, objectives and effects of this invention easier to understand, the following embodiments are described in detail with reference to the accompanying drawings. Example

[0025] This invention achieves precise vehicle positioning in the absence of GPS, intelligent parking space allocation considering global efficiency and user preferences, and smooth and efficient path planning through distributed, multimodal data acquisition combined with advanced computer vision and optimization algorithms, thereby systematically improving parking lot throughput and service levels.

[0026] This embodiment provides a collaborative parking system based on a distributed wireless sensor network, including: a distributed perception module, a visual positioning module, a parking space allocation module, a path planning module, and a co-simulation module. These modules are connected sequentially to form a complete technical closed loop of "perception-positioning-decision-planning-verification".

[0027] In this embodiment, the distributed perception module includes a camera, a physiological signal acquisition device, an eye tracker, an audio acquisition device, and vehicle operation sensors. It is used to comprehensively collect visual data of the parking lot environment, driver physiological state data, driver attention (eye movement) data, in-vehicle audio environment data, and operational data from the vehicle's steering wheel and pedals. This multimodal data provides rich contextual information for subsequent analysis. Specifically, the camera is a 1080p resolution, 30fps RGB camera; the physiological signal acquisition device is a wearable electrocardiogram with a 250Hz sampling rate; the eye tracker is a telemetry eye tracker with a 120Hz sampling rate and a gaze tracking error ≤1.2°; the audio acquisition device has a sampling rate of 16000Hz; and the vehicle operation sensors include a high-precision steering wheel / pedal force sensor with a 500Hz sampling rate.

[0028] In this embodiment, the visual positioning module processes the continuous image stream captured by the camera to achieve centimeter-level precise positioning of the vehicle within the parking lot. Its core technology lies in combining visual SLAM (Simultaneous Localization and Mapping) based on natural features with absolute pose correction based on artificial tags (AprilTags). First, the module constructs a multi-layer Gaussian pyramid for the input image, extracts ORB feature points by dividing the base layer into a grid, and uses a quadtree algorithm for filtering to ensure that feature points are evenly distributed in the image, thereby improving the robustness and accuracy of subsequent matching. Next, the Fast Approximate Nearest Neighbor (FLANN) algorithm is used for rapid matching of feature descriptors, and the Random Sample Consensus (RANSAC) algorithm is used to eliminate mismatched point pairs, resulting in stable feature matching relationships. Simultaneously, the module detects AprilTags visual reference labels deployed in real time at key locations such as parking lot pillars and walls. The AprilTags detection process includes image grayscale conversion, adaptive threshold binarization, connected component clustering and quadrilateral fitting, Hamming distance-based encoding and decoding, and sub-pixel-level corner point precise positioning. Finally, the continuous relative pose provided by ORB-SLAM2 and the discrete absolute pose provided by AprilTags are fused to output the vehicle's global coordinates and pose in a pre-built high-precision parking map. This fusion scheme effectively overcomes the cumulative error problem of pure visual SLAM, and in tests on public datasets (such as TUM), the average trajectory error is reduced by more than 30% compared to traditional methods.

[0029] In this embodiment, the parking space allocation module is the "decision-making brain" of the system. Its task is to calculate a globally optimal or near-optimal parking space allocation scheme for all vehicles requesting parking based on the real-time parking lot status (distribution of available parking spaces, vehicle positions within the lot) and user-submitted preference information (such as the desire to park in a certain area, the need for charging stations, etc.). The core of this module is a dynamic optimization model. The model introduces "parking lot saturation P" as a key parameter and sets a preference consideration threshold s. When the parking lot is relatively empty (P ≤ s), the system has sufficient space to meet users' personalized needs. At this time, the single-vehicle parking time model considers both the "driving time" from the current location to the candidate parking space and the "walking time" from the parking space to the destination (such as the elevator entrance). When the parking lot approaches saturation (P > s), to avoid vehicles lingering in the lot for a long time and causing congestion, the model will focus on optimizing the "walking time," guiding vehicles to park in the available parking space closest to their destination, while weakening the personalized preference for the driving route. The objective function is to minimize the total weighted parking time of all vehicles, constituting a typical 0-1 integer programming problem. To efficiently solve this NP-hard problem, this invention designs an integer-encoded particle swarm optimization algorithm. In this algorithm, the position vector of each "particle" is directly encoded as a parking space allocation sequence. For example, the sequence [18, 5, 33] represents allocating vehicle 1 to parking space 18, vehicle 2 to parking space 5, and vehicle 3 to parking space 33. The algorithm iteratively searches by simulating the social learning and individual experience learning mechanisms of bird flocks and introduces a special repair operator to handle encoding conflicts (such as repeated allocation of the same parking space), ultimately converging to a high-quality solution that satisfies the "one vehicle, one parking space" constraint.

[0030] In this embodiment, after obtaining the parking space allocation results, the path planning module is responsible for planning an optimal driving path for each vehicle from its current location to its allocated parking space. This module operates based on a grid map containing three layers of semantic information: the bottom layer is a static obstacle layer (such as walls, pillars, and fixed facilities); the middle layer is a traffic rule layer (such as lane lines, no-parking zones, and directional indicators); and the top layer is a dynamic occupancy layer (real-time updated positions of parked vehicles). On this map, the module employs an improved A* search algorithm. The improvements are mainly reflected in two aspects: First, a variable weight coefficient w is introduced into the heuristic function, with the cost function expressed as f(n) = g(n) + w * h(n), where g(n) is the actual cost and h(n) is the Manhattan distance heuristic value. By adjusting w (typically between 0.8 and 1.5), a balance can be achieved between "search speed" (when w is large) and "path length optimality" (when w is small). Second, a multi-parent node selection strategy is implemented. In the standard A* algorithm, each node has only one parent node. In this invention, when expanding nodes, if the g(n) values ​​reaching the current node n through different parent nodes are found to be equal, the invention does not simply record one parent node. Instead, it evaluates which parent node makes the path from the grandparent node to the current node straighter (i.e., the current node, parent node, and grandparent node are more likely to be collinear) and prioritizes that parent node. This strategy effectively reduces unnecessary right-angle inflection points in the final planned path, making the path smoother and beneficial for vehicle control. Experiments show that this strategy can reduce the number of path inflection points by an average of 15%-25%.

[0031] In this embodiment, the co-simulation module is used to integrate, verify, and evaluate the performance of the aforementioned positioning, allocation, and planning algorithms before system deployment. This module is built on the Ubuntu operating system and employs a co-simulation framework combining the Robot Operating System (ROS) and MATLAB. The ROS side is responsible for running near real-time tasks, such as camera image publishing, ORB-SLAM2 positioning nodes, and vehicle control interfaces, and publishes vehicle pose data through a topic mechanism. The MATLAB side subscribes to these topics and utilizes its powerful mathematical computation and optimization toolbox to run the integer programming model for parking space allocation and the improved A* algorithm for path planning. This architecture ensures the real-time requirements of perception and positioning while fully leveraging MATLAB's advantages in algorithm prototype verification and complex computation. This module can flexibly simulate various scenarios from low saturation (sufficient vacant parking spaces) to extremely high saturation (e.g., P=0.96, nearly full), verifying the effectiveness and robustness of the system strategy under different pressures.

[0032] In this embodiment, the invention also includes a data preprocessing and fusion unit. This unit is located after the distributed sensing module and is responsible for timestamping, filtering and denoising (such as bandpass filtering of ECG signals and denoising of audio signals) and removing outliers from the raw multimodal data collected by each sensor. It also fuses the relevant data at the feature level (such as mapping the eye-tracking gaze coordinates to the visual image coordinate system) to provide a unified, high-quality, and semantically rich input data stream for subsequent modules.

[0033] This embodiment also provides a cooperative parking method based on a distributed wireless sensor network, applied to the aforementioned system, including the following steps: S1: Through the distributed sensing module, multimodal data of the environment and vehicles are collected synchronously in the target parking lot scenario; S2: The visual positioning module processes the visual data stream and outputs the precise pose of all vehicles in the parking lot map through feature extraction, matching and fusion. S3: The parking space allocation module receives vehicle position, available parking space information and user preferences, calculates and outputs the globally optimal parking space allocation scheme; S4: The route planning module plans the optimal smooth path from the starting point to the destination for each vehicle based on the allocation scheme and the real-time map. S5: In the co-simulation module, typical scenarios under different saturation levels are constructed to simulate and verify the overall system solution and evaluate its performance.

[0034] The role and effect of the embodiments The cooperative parking system and method based on a distributed wireless sensor network according to the present invention achieves the following significant beneficial effects through the implementation of the above technical solutions: 1. Achieved high-precision, low-cost indoor positioning: Creatively integrates visual SLAM based on natural features with correction technology based on artificial tags AprilTags, solving the problem of accurate vehicle positioning in underground parking lots without GPS without increasing infrastructure costs. The positioning accuracy meets the requirements for practical application and avoids the high deployment costs of laser or UWB solutions.

[0035] 2. Achieving a balance between global optimization and user experience: The proposed parking space allocation optimization model based on dynamic saturation adjustment and the efficient integer-encoded particle swarm optimization algorithm can complete the global coordination of parking space demand for multiple vehicles within seconds. It can respect individual user preferences and improve satisfaction when parking spaces are plentiful, while prioritizing overall system traffic efficiency and reducing congestion when parking spaces are scarce, thus significantly reducing total parking time (simulations show a reduction of 15%-20%).

[0036] 3. Generated smoother and more efficient driving paths: The improved A* search algorithm balances search efficiency and path quality through a weighted heuristic function, and effectively reduces redundant inflection points in the path through a multi-parent node strategy. The planned path is more in line with the vehicle's motion characteristics, which not only shortens the travel time but also improves the smoothness and comfort of driving, while reducing the vehicle's steering losses.

[0037] 4. A flexible and reliable development and verification platform has been built: The ROS-MATLAB co-simulation architecture decomposes complex systems into loosely coupled modules, facilitating independent algorithm development, testing, and integration. This platform can simulate various extreme and typical scenarios, providing powerful tools for algorithm performance evaluation and system optimization, greatly accelerating the product development process and reliability verification.

[0038] 5. Excellent scalability and practicality: The "distributed perception + centralized collaborative decision-making" architecture reduces the intelligence requirements for individual vehicles, facilitating the retrofitting and promotion of existing vehicles and parking lots. The modular design of the system also makes it easy to integrate more functions in the future, such as linkage with parking fee collection systems and electric vehicle charging management systems, demonstrating broad prospects for industrial application.

[0039] The above embodiments are preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention.

Claims

1. A cooperative parking system based on a distributed wireless sensor network, characterized in that, include: The system includes a distributed sensing module, a visual positioning module, a berth allocation module, a path planning module, and a co-simulation module. The distributed sensing module includes a camera, a physiological signal acquisition device, an eye tracker, an audio acquisition device, and vehicle operation sensors, used to collect visual data of the parking lot environment, driver physiological data, driver eye movement data, in-vehicle audio data, and vehicle operation data. The visual positioning module is used to extract and match features from the images captured by the camera, and combines real-time localization and mapping algorithms with AprilTags visual reference labels to achieve precise vehicle positioning within the parking lot. The parking space allocation module is used to construct an optimization model with the objective of minimizing the total parking time based on parking lot saturation and user preference information, and to solve the model using an integer-encoded particle swarm optimization algorithm, outputting the globally optimal parking space allocation scheme. The path planning module is used to plan the optimal driving path from the current location to the assigned parking space for each vehicle based on a grid map containing multi-layered information and an improved A* search algorithm that combines a weighted heuristic function and a multi-parent node strategy. The co-simulation module is used to integrate and verify the functions and performance of the visual positioning module, the berth allocation module, and the path planning module in a co-simulation environment of the robot operating system and MATLAB.

2. The cooperative parking system based on a distributed wireless sensor network according to claim 1, characterized in that: in, The camera is a 1080p resolution, 30fps RGB camera; the physiological signal acquisition device is a wearable electrocardiograph with a 250Hz sampling rate; the eye tracker is a telemetry eye tracker with a 120Hz sampling rate and a gaze tracking error ≤1.2°; the audio acquisition device has a sampling rate of 16000Hz; and the vehicle operation sensor includes a steering wheel / pedal force sensor with a 500Hz sampling rate.

3. The cooperative parking system based on a distributed wireless sensor network according to claim 1, Its features are: in, The feature extraction performed by the visual positioning module includes: constructing a multi-layer Gaussian pyramid for the input image, dividing the pixel layer into a grid to extract ORB feature points, and using a quadtree algorithm to filter the feature points to achieve uniform distribution; The feature matching includes coarse matching using a fast nearest neighbor search algorithm and fine matching using a random sampling consensus algorithm to eliminate mismatches.

4. The cooperative parking system based on a distributed wireless sensor network according to claim 1, characterized in that: in, In the berth allocation module, the optimization model is a 0-1 integer programming model, whose objective function is to minimize the total weighted parking time of all vehicles. The constraint condition ensures that each vehicle is allocated only one berth and each berth is allocated to at most one vehicle. In the integer-encoded particle swarm optimization algorithm, the position of each particle represents a complete berth allocation sequence.

5. The cooperative parking system based on a distributed wireless sensor network according to claim 1, characterized in that: in, In the path planning module, the cost function of the improved A* search algorithm is f(n) = g(n) + w * h(n), where g(n) is the actual cost from the starting point to node n, h(n) is the Manhattan distance heuristic value from node n to the destination, and w is an adjustable weight coefficient; the multi-parent node strategy refers to the fact that when expanding nodes, if there are multiple parent nodes that make g(n) equal, the parent node that makes the path straighter is selected first.

6. The cooperative parking system based on a distributed wireless sensor network according to claim 1, characterized in that: in, The co-simulation module is built on the Ubuntu operating system. The robot operating system is responsible for running sensor-driven, real-time localization and map building nodes and publishing vehicle pose topics. The MATLAB environment is responsible for subscribing to the pose topics and executing the parking space allocation and path planning algorithms.

7. The cooperative parking system based on a distributed wireless sensor network according to claim 1, characterized in that: in, In the visual positioning module, the detection and recognition process of the AprilTags visual reference labels includes image grayscale conversion and binarization, connected component clustering and quadrilateral fitting, Hamming distance-based encoding and decoding, and sub-pixel-level corner point positioning.

Images