Parking berth detection method and system based on network model and multi-source data fusion

Abstract

The invention relates to the technical field of traffic management, and provides a parking space detection method and system based on a network model and multi-source data fusion, and the method comprises the steps: building a network model according to a parking space state node set, building directed branches between nodes according to an undetermined state transmission probability matrix, and obtaining a parking space combination state network model; collecting image information of each node in the parking space combination state network model, and preprocessing and enhancing the collected image information; carrying out pixel classification on the preprocessed and enhanced image information to identify a parking space state, and updating an undetermined state transmission probability in the parking space combination state network model; optimizing the updated parking space combination state network model according to the maximum posterior probability, and performing decomposition and parallel processing on a model optimization task; and analyzing the multi-source data by adopting the optimized parking space combination state network model, and issuing parking space state information in real time. According to the invention, the detection precision and real-time performance can be improved.

Description

Technical Field

[0001] This invention relates to the field of traffic management technology, and in particular to a parking space detection method and system based on network model and multi-source data fusion. Background Technology

[0002] With the continuous improvement of people's living standards, holiday travel has become an important part of daily life. The increasingly normalized concentrated holiday patterns such as Golden Week and short holidays have led to an explosive growth in the number of tourists. While this trend has effectively boosted domestic demand and enriched people's cultural lives, it has also brought severe challenges to popular scenic spots. Large-scale traffic congestion and excessive crowds are frequent occurrences, seriously affecting the tourist experience and posing potential risks to road traffic safety and public order. Against this backdrop, achieving scientific assessment and accurate prediction of the gathering of people and vehicles in target areas, thereby preventing large-scale congestion and reducing public safety risks, has become a major and urgent task for management departments.

[0003] Among the many aspects of scenic area management, parking management holds a central position. Currently, many scenic areas still lack efficient and accurate technical support for parking space management, making it impossible to obtain real-time and accurate information on parking space occupancy. This results in tourists spending a significant amount of time searching for available parking spaces, exacerbating traffic congestion and chaos on surrounding roads, and reducing overall traffic efficiency. Therefore, developing and applying an intelligent system capable of real-time and accurate detection of parking space status is of significant practical necessity for improving scenic area management and tourist satisfaction.

[0004] Among existing parking space detection technologies, the most widely used include ultrasonic parking space detectors and infrared parking space detectors. Ultrasonic detectors emit ultrasonic waves and receive the echoes, using the time difference principle to measure the distance between the vehicle and the detector, thus determining the parking space status; this is a non-contact measurement method. These devices are typically installed directly above the parking space or at a suitable height. Infrared detectors are divided into active and passive types. Active infrared detectors emit and receive infrared beams, determining vehicle entry and exit based on whether the beam is blocked; passive infrared detectors detect moving objects by sensing changes in infrared radiation within a designated area.

[0005] However, all of the aforementioned existing technologies have significant limitations. Ultrasonic detectors are susceptible to interference from foreign objects adhering to the probe surface or strong ambient sound waves, leading to ranging errors or even misjudgments. Their installation and debugging processes are also relatively complex, requiring on-site calibration by professionals, resulting in high maintenance costs. Infrared detectors are highly sensitive to environmental factors; changes in ambient temperature and strong light exposure can cause false alarms. Furthermore, they are ineffective at detecting stationary vehicles and struggle to accurately identify vehicles parked for extended periods, limiting their applicability in real-time parking space status monitoring.

[0006] It is particularly noteworthy that smart scenic area scenarios have their unique characteristics: the environment is complex, and vegetation and decorations can easily obstruct vision; visitor flow fluctuates greatly, and the real-time and accuracy requirements of the detection system are extremely high during peak periods; in addition, it is often necessary to integrate multi-source data such as cameras and people flow counting equipment to achieve comprehensive management. These characteristics make it difficult for traditional detection methods to meet the needs of smart scenic areas for efficient, reliable, and integrated parking management.

[0007] Chinese patent CN116108389A discloses a reliable perception method for urban parking across the entire area based on crowdsourced data. This method utilizes crowdsourced data and simulation modeling to achieve reliable perception of the status of roadside parking spaces at the urban level. However, in practical applications, this solution has the following drawbacks: 1. The core analysis and model training of the solution heavily rely on a pre-set simulation environment (such as PreScan) and idealized parameter distributions (such as Poisson distribution), which differ significantly from the complex and ever-changing real-world road environment and traffic flow characteristics. 2. The data freshness function is based on the strong assumption that "vehicle arrivals and departures both follow a Poisson distribution" and ignores higher-order terms. This assumption often does not hold true in real-world urban traffic systems (especially affected by traffic lights, congestion, and unexpected events), potentially leading to distorted freshness assessments and affecting the reliability of the fused data. 3. The solution involves multi-level models (detection model, XGBoost classification, probabilistic fusion calculation) and iterative processes, resulting in high computational overhead. Especially when performing city-level full-area perception, it requires processing massive amounts of crowdsourced data and completing fusion calculations in real time, placing extremely high demands on system computing power and communication bandwidth, making it difficult to meet the high real-time requirements of practical applications.

[0008] Therefore, there is an urgent need for an intelligent parking space detection method that can overcome the above-mentioned defects, adapt to complex environments, support multi-source data fusion, and has high real-time performance, so as to improve the efficiency of parking management in scenic areas, alleviate traffic pressure, and provide effective technical support for comprehensive management decisions in scenic areas. Summary of the Invention

[0009] In view of this, in order to overcome the shortcomings of the prior art, the present invention aims to provide a parking space detection method and system based on network model and multi-source data fusion.

[0010] According to a first aspect of the present invention, a parking space detection method based on network model and multi-source data fusion is provided, the method comprising: Step S1: Construct a network model based on the set of parking space state nodes, establish directed branches between nodes based on the probability matrix of undetermined states, and obtain the parking space combination state network model. Step S2: Collect image information of each node in the parking space combination state network model, and perform preprocessing and enhancement processing on the collected image information; Step S3: Identify parking space status by performing pixel classification on the pre-processed and enhanced image information, and update the undetermined state transmission probability in the parking space combination state network model; Step S4: Optimize the updated parking space combination state network model based on the maximum posterior probability, and decompose and parallelize the model optimization task; Step S5: Analyze the multi-source data using the optimized parking space combination status network model and release parking space status information in real time.

[0011] Optionally, in the parking space detection method based on network model and multi-source data fusion of the present invention, in step S1, the parking space status node set includes the parking space with car status and no car status nodes. The elements of the undetermined state transmission probability matrix are used to represent the undetermined state transmission probability of parking space status change. The undetermined state transmission probability is set as the directed branch between the two nodes before and after the state change.

[0012] Optionally, in the parking space detection method based on network model and multi-source data fusion of the present invention, step S2 involves preprocessing and enhancing the acquired image information, including: using a denoising algorithm to denoise the acquired image information, and using an image restoration algorithm to reconstruct and restore the pixels of the denoised image information.

[0013] Optionally, in the parking space detection method based on network model and multi-source data fusion of the present invention, in step S2, a non-local mean denoising algorithm is used to denoise the image information, and an image inpainting algorithm based on partial differential equations is used to reconstruct and repair the missing areas of the denoised image information.

[0014] Optionally, in the parking space detection method based on network model and multi-source data fusion of the present invention, in step S3, a Bayesian classifier is used to classify the pixels of the preprocessed and enhanced image information, calculate the posterior probability of each pixel belonging to the state of having a car or not having a car, and identify the parking space status.

[0015] Optionally, in the parking space detection method based on network model and multi-source data fusion of the present invention, step S3, updating the undetermined state transmission probability in the parking space combination state network model includes: iteratively calculating the classification result obtained by the Bayesian classifier and the undetermined state transmission probability matrix, and dynamically updating the undetermined state transmission probability in the undetermined state transmission probability matrix.

[0016] Optionally, in the parking space detection method based on network model and multi-source data fusion of the present invention, step S4 optimizes the updated parking space combination state network model according to the maximum posterior probability, including: finding the model parameters of the parking space combination state network model that maximizes the posterior probability based on the maximum posterior probability criterion.

[0017] Optionally, in the parking space detection method based on network model and multi-source data fusion of the present invention, in step S4, a distributed computing architecture is used to decompose the optimization task of the parking space combination state network model, and the decomposed task is distributed to multiple computing nodes for parallel processing. Each computing node independently completes local calculations and then coordinates through communication to achieve global parameter optimization.

[0018] Optionally, the parking space detection method based on network model and multi-source data fusion of the present invention includes step S5, which includes: Multi-source data, including real-time parking space status data, video surveillance data, and tourist flow data, are collected. A data fusion method based on tensor decomposition is used to extract common features of the multi-source data and eliminate redundancy and conflicts between data to obtain fused data. A long short-term memory network model is used to predict changes in parking space demand in future periods based on historical fusion data, and a convolutional neural network model is used to analyze real-time fusion data to detect parking space status information, abnormal events, and violations within the parking area. By establishing a message push mechanism, the detected parking space status information is sent to the car owner's mobile terminal in real time. At the same time, a data interface is established with the scenic area's intelligent navigation system to provide car owners with optimal parking route planning and real-time parking space guidance services. Among them, the optimal parking route planning adopts a dynamic path optimization algorithm based on real-time traffic conditions and parking space availability.

[0019] According to a second aspect of the present invention, a parking space detection system based on a network model and multi-source data fusion is provided. The system includes a detection server, which comprises: The model building module is used to build a network model based on the set of parking space state nodes, establish directed branches between nodes based on the probability matrix of undetermined states, and obtain a parking space combined state network model. The image acquisition module is used to acquire image information of each node in the parking space combination state network model, and to preprocess and enhance the acquired image information. The update module is used to identify parking space status by performing pixel classification on pre-processed and enhanced image information, and update the undetermined state transmission probability in the parking space combination state network model. The model optimization module is used to optimize the updated parking space combination state network model based on the maximum posterior probability, and to decompose and process the model optimization task in parallel. The detection and release module is used to analyze multi-source data using an optimized parking space combination status network model and release parking space status information in real time.

[0020] The present invention provides a parking space detection method and system based on network models and multi-source data fusion, which has the following beneficial technical effects: 1. Significantly improves detection accuracy by using advanced image processing and Bayesian classification to ensure accurate judgment of parking space status.

[0021] 2. It has real-time adaptive capability, dynamically adjusting parameters according to traffic flow to adapt to parking needs at different times.

[0022] 3. Significantly improve computational efficiency by adopting a distributed architecture for parallel optimization, enabling minute-level model updates.

[0023] 4. Enhance comprehensive decision-making capabilities by integrating multi-source data and using deep learning to predict demand and monitor anomalies.

[0024] 5. Improve the parking experience for tourists by reducing parking time and alleviating congestion through real-time information dissemination and intelligent navigation. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is an example architecture diagram of a parking space detection system based on a network model and multi-source data fusion according to an embodiment of the present invention; Figure 2 This is an example diagram of the architecture of a detection server according to an embodiment of the present invention; Figure 3 This is a flowchart illustrating the steps of a parking space detection method based on a network model and multi-source data fusion according to an embodiment of the present invention. Figure 4 This is a schematic diagram of the structure of the device provided by the present invention. Detailed Implementation

[0027] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0028] It should be noted that, in the absence of conflict, the following embodiments and features can be combined with each other; and, based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0029] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0030] Figure 1 This is an example architecture diagram of a parking space detection system based on a network model and multi-source data fusion according to an embodiment of the present invention, as shown below. Figure 1 As shown, the system may include a detection server 101, a communication network 102, and / or one or more detection clients 103. Figure 1 The example in the text is multiple detection clients 103.

[0031] The detection server 101 can be any suitable server for storing information, data, programs, and / or any other suitable type of content. In some embodiments, the detection server 101 can perform appropriate functions. For example, in some embodiments, the detection server 101 can be used to detect parking spaces based on a network model and multi-source data fusion. As an optional example, in some embodiments, the detection server 101 can be used to: construct a network model based on a set of parking space status nodes, establish directed branches between nodes based on the probability matrix of undetermined states, and obtain a parking space combination state network model; collect image information of each node in the parking space combination state network model, and preprocess and enhance the collected image information; identify parking space status by pixel classification of the preprocessed and enhanced image information, and update the undetermined state transmission probability in the parking space combination state network model; optimize the updated parking space combination state network model based on the maximum a posteriori probability, decompose and parallelize the model optimization task; analyze multi-source data using the optimized parking space combination state network model, and publish parking space status information in real time.

[0032] Figure 2 This is an example architecture diagram of a detection server according to an embodiment of the present invention, such as... Figure 2 As shown, in this embodiment, the detection server includes: The model building module is used to build a network model based on the set of parking space state nodes, establish directed branches between nodes based on the probability matrix of undetermined states, and obtain a parking space combined state network model. The image acquisition module is used to acquire image information of each node in the parking space combination state network model, and to preprocess and enhance the acquired image information. The update module is used to identify parking space status by performing pixel classification on pre-processed and enhanced image information, and update the undetermined state transmission probability in the parking space combination state network model. The model optimization module is used to optimize the updated parking space combination state network model based on the maximum posterior probability, and to decompose and process the model optimization task in parallel. The detection and release module is used to analyze multi-source data using an optimized parking space combination status network model and release parking space status information in real time.

[0033] As another example, in some embodiments, the detection server 101 may send a parking space detection method based on network model and multi-source data fusion to the detection client 103 for user use, according to the request of the detection client 103.

[0034] As an optional example, in some embodiments, the detection client 103 is used to provide a visual detection interface, which is used to receive a user's selection input operation for detecting parking spaces based on network models and multi-source data fusion, and to obtain from the detection server 101 the detection interface corresponding to the option selected by the selection input operation and display the detection interface in response to the selection input operation. The detection interface displays at least the information on detecting parking spaces based on network models and multi-source data fusion, as well as operation options for the information on detecting parking spaces based on network models and multi-source data fusion.

[0035] In some embodiments, communication network 102 can be any suitable combination of one or more wired and / or wireless networks. For example, communication network 102 can include any one or more of the following: the Internet, intranet, wide area network (WAN), local area network (LAN), wireless network, digital subscriber line (DSL) network, frame relay network, asynchronous transfer mode (ATM) network, virtual private network (VPN), and / or any other suitable communication network. Detection client 103 can connect to communication network 102 via one or more communication links (e.g., communication link 104), which can be linked to detection server 101 via one or more communication links (e.g., communication link 105). Communication links can be any communication link suitable for transmitting data between detection client 103 and detection server 101, such as network links, dial-up links, wireless links, hardwired links, any other suitable communication links, or any suitable combination of such links.

[0036] The detection client 103 may include any one or more clients that present an interface related to the detection of parking spaces based on network models and multi-source data fusion in an appropriate form for user use and operation. In some embodiments, the detection client 103 may include any suitable type of device. For example, in some embodiments, the detection client 103 may include a mobile device, tablet computer, laptop computer, desktop computer, and / or any other suitable type of client device.

[0037] Although the detection server 101 is illustrated as a single device, in some embodiments, any suitable number of devices may be used to perform the functions performed by the detection server 101. For example, in some embodiments, multiple devices may be used to implement the functions performed by the detection server 101. Alternatively, cloud services may be used to implement the functions of the detection server 101.

[0038] Based on the above system, this invention provides a parking space detection method based on network model and multi-source data fusion, which will be described in the following embodiments.

[0039] Figure 3 This is a flowchart illustrating the steps of a parking space detection method based on network model and multi-source data fusion according to an embodiment of the present invention. The parking space detection method based on network model and multi-source data fusion of this embodiment can be executed on a detection server. This parking space detection method based on network model and multi-source data fusion includes the following steps: Step S1: Construct a network model based on the set of parking space state nodes, establish directed branches between nodes based on the probability matrix of undetermined states, and obtain the parking space combination state network model.

[0040] In this embodiment, the parking space status node set includes the parking space's car-occupied and car-free status nodes. The elements of the undetermined state transmission probability matrix are used to represent the undetermined state transmission probability of parking space status changes. The undetermined state transmission probability is set as the directed branch between the two nodes before and after the state change.

[0041] For example, in this embodiment, let... Represents the set of parking space status nodes, where This indicates that the vehicle is in use. This represents a car-free state. The fixed-state propagation probability matrix of a directed branch. medium elements Indicates from state to state The probability of propagation of undetermined states, .

[0042] Step S2: Collect image information of each node in the parking space combination state network model, and perform preprocessing and enhancement processing on the collected image information.

[0043] This embodiment first acquires the image information of each node in the parking space combination state network model in real time through image acquisition equipment deployed in the parking area. After acquiring the image information, a denoising algorithm is used to denoise the acquired image information to remove random noise in the image information and ensure image clarity. Then, an image restoration algorithm is used to reconstruct and restore the pixels of the denoised image information.

[0044] As an optional example, this embodiment uses a non-local means denoising algorithm to denoise the image information. For example, for pixels in the image... Its denoised pixel values Calculated by the following formula:

[0045] in, It revolves around pixels neighborhood window, It is a pixel grayscale value, weight Defined as:

[0046] in, and They are respectively based on and The neighborhood block centered on, It has standard deviation The weighted Euclidean distance, It is a parameter that controls the smoothness.

[0047] As an optional example, this embodiment uses an image inpainting algorithm based on partial differential equations to reconstruct and repair pixel-based missing areas of image information after denoising (such as local missing or blurred parts of the image caused by vegetation obstruction or building shadows in scenic areas).

[0048] For example, this embodiment uses the Criminisi algorithm based on partial differential equations for the region to be repaired. pixels within Its repair value Solve iteratively using the following formula:

[0049] in, t For time ， It is a diffusion tensor that is dynamically adjusted based on the structural information of the image to ensure accurate reconstruction of texture and structure when repairing image defects caused by vegetation occlusion and building shadows.

[0050] Step S3: By performing pixel classification on the preprocessed and enhanced image information, identify the parking space status and update the undetermined state transmission probability in the parking space combination state network model.

[0051] In this embodiment, a Bayesian classifier is used to classify the pixels of the pre-processed and enhanced image information, calculate the posterior probability of each pixel belonging to the state of having a car or not, and identify the parking space status.

[0052] For example, in this embodiment, let... It is the pixel feature vector of the parking space. This refers to the parking space status category (car present or absent). According to Bayes' theorem, the pixel belongs to the category. posterior probability for:

[0053] in, It is the likelihood probability, which can be obtained by learning from a large amount of labeled data. It is the prior probability.

[0054] In this embodiment, the classification results obtained by the Bayesian classifier and the undetermined state transit probability matrix are iteratively calculated, and the undetermined state transit probability in the undetermined state transit probability matrix is ​​dynamically updated.

[0055] For example, in this embodiment, given the pixel classification result, the probability obtained using a Bayesian classifier is combined with the previously defined undetermined state propagation probability matrix. The propagation probability of the undetermined state is updated through iterative calculation. Let... Let be the iteration number, at the th iteration. In the next iteration, the state propagation probability The updated formula is:

[0056] in, It's the learning rate. It is the sample size. It is the first The sample at the th Parking space status at the next iteration It is the Kronecker function.

[0057] In this embodiment, the following real-time dynamic optimization model can also be used: Let It is over time Variable traffic-related parameters (such as vehicle entry and exit frequencies at different times), key parameters in the network model (Such as state transit probability weights) are adjusted according to the following formula:

[0058] in, It's the step length. It is a loss function constructed based on the difference between the current model prediction results and the actual parking space status, and the parameters are continuously optimized using the gradient descent method. To accommodate the parking needs of the scenic area at different times.

[0059] In this embodiment, the real-time dynamic optimization model is started simultaneously to continuously collect vehicle traffic data in the scenic area. Based on the changes in the frequency and number of vehicles entering and leaving the parking lot at different times, such as peak and off-peak seasons, weekdays and weekends, and day and night, the model automatically and timely adjusts key parameters in the network model, such as the weight of the state transit probability, so that the model can always adapt to the complex and ever-changing parking needs and vehicle flow in the scenic area, and ensure that the parking space status can be accurately detected at all times.

[0060] Step S4: Optimize the updated parking space combination state network model based on the maximum posterior probability, and decompose and process the model optimization task in parallel.

[0061] In this embodiment, the model parameters of the parking space combination state network model that maximizes the posterior probability are found based on the maximum a posteriori probability criterion.

[0062] For example, this embodiment optimizes the parking space combination state network model based on the maximum posterior probability in the following manner: set up It is a set of model parameters. These are observational data (i.e., parking space status data). Based on the maximum a posteriori probability criterion, the optimization objective is to find the data that maximizes the posterior probability. The largest parameter value. By Bayes' theorem, By maximizing the log-posterior probability To solve this, we need to find the answer: ,in It is the likelihood function. It is the prior distribution of the parameters.

[0063] It should be noted that this embodiment uses a distributed computing architecture to decompose the optimization task of the parking space combination state network model, and distributes the decomposed task to multiple computing nodes for parallel processing. Each computing node independently completes local calculations and then coordinates through communication to achieve global parameter optimization.

[0064] For example, this embodiment decomposes the model optimization task into... Each sub-task was assigned to Parallel processing is performed on the nth computing node. Let the nth node be... The objective function corresponding to the subtask processed by each computing node is: Through a distributed optimization algorithm, each node iteratively updates its parameters. : ,in, It is the first The learning rate of each node. Each node exchanges information through network communication, and the resulting data is fused to obtain the globally optimal model parameters. This greatly improves optimization efficiency, enabling the model to complete complex calculations in a short time and quickly and accurately reflect the real-time status of each parking space, fully meeting the requirements for timely parking space detection when vehicles enter and exit the scenic area quickly.

[0065] Step S5: Analyze the multi-source data using the optimized parking space combination status network model and release parking space status information in real time.

[0066] Multi-source data, including real-time parking space status data, video surveillance data, and tourist flow data, are collected. A data fusion method based on tensor decomposition is used to extract common features of the multi-source data and eliminate redundancy and conflicts between data to obtain fused data. A long short-term memory network model is used to predict changes in parking space demand in future periods based on historical fusion data, and a convolutional neural network model is used to analyze real-time fusion data to detect parking space status information, abnormal events, and violations within the parking area. By establishing a message push mechanism, the detected parking space status information is sent to the car owner's mobile terminal in real time. At the same time, a data interface is established with the scenic area's intelligent navigation system to provide car owners with optimal parking route planning and real-time parking space guidance services. Among them, the optimal parking route planning adopts a dynamic path optimization algorithm based on real-time traffic conditions and parking space availability.

[0067] In practical applications, the method of this embodiment can utilize mobile terminals, such as mobile apps or SMS push notifications, to promptly transmit the latest parking space availability information to car owners, facilitating their advance planning of parking routes. Furthermore, by closely integrating with the scenic area's intelligent navigation system, when tourists use the system, it plans the most convenient and efficient parking route based on the driver's current location and the real-time status of surrounding parking spaces, guiding drivers to quickly find available parking spaces, significantly improving the tourist parking experience and effectively alleviating traffic congestion in the scenic area.

[0068] For example, in this embodiment, when information is published via a mobile terminal, parking space information is pushed in real time through a message queue system (such as Kafka). Let the message queue... When a new parking space status is updated, a message containing information such as the parking space ID, status, and location will be sent. Joining the queue, i.e. Mobile devices can subscribe to this queue to obtain the latest parking space information in real time, enabling timely delivery of information to car owners.

[0069] Integration with the scenic area's intelligent navigation system: Assume the driver's current location coordinates are... The surrounding parking spaces are grouped as Each parking space The coordinates are The A* algorithm is used to plan the shortest path, and its evaluation function is... for:

[0070] in, From current location to parking space The actual cost (such as distance). From parking space Estimated cost to reach the target point (such as the entrance to a scenic area). Through continuous searching and updating, the optimal parking route is found to guide drivers to quickly find available parking spaces.

[0071] In practical applications, this embodiment deeply integrates parking space detection data with multi-source data such as surveillance camera data and tourist flow monitoring data within the scenic area.

[0072] For example, in this embodiment, let the parking space detection data be... The data from the surveillance cameras in the scenic area is Tourist flow monitoring data is A multi-source data fusion method based on tensor decomposition is adopted to represent multi-source data as tensors. The core tensor is obtained through tensor decomposition. sum factor matrix ,Right now: The merged data retains the characteristic information of each data source, providing a more comprehensive data foundation for subsequent analysis. Advanced machine learning algorithms are used to perform intelligent analysis on the merged comprehensive data.

[0073] Parking demand forecasting: Parking demand forecasting is performed using a Long Short-Term Memory (LSTM) network model. Let the input data sequence be... ,in The output of the LSTM model includes fused visitor flow, time, and other feature information. for Predicting parking demand at any given time. The model is trained by minimizing the mean squared error (MSE) between the predicted and actual values:

[0074] The model parameters are continuously adjusted through the backpropagation algorithm to achieve accurate prediction of parking demand.

[0075] Anomaly Detection: This function analyzes surveillance video data using a Convolutional Neural Network (CNN) to detect anomalies around parking spaces. Let the input video frame image be... After a series of convolutional, pooling, and fully connected layers, the prediction result is obtained. It determines whether there are any anomalies (such as vehicles parked illegally for extended periods or abnormal gatherings of people). This is done using the cross-entropy loss function. Training the model:

[0076] in, It is the sample size. It is a real-world label that, through continuous optimization of model parameters, enables real-time and accurate monitoring and early warning of abnormal situations around parking spaces, providing comprehensive decision support for scenic area management departments.

[0077] This embodiment analyzes the correlation between visitor flow and parking space usage to predict parking demand at different times in scenic areas, helping scenic area management departments to prepare in advance, such as rationally allocating staff to guide parking and opening backup parking lots when necessary. Simultaneously, by combining surveillance video data, it monitors abnormal situations around parking spaces in real time, such as vehicles parking illegally for extended periods or people gathering abnormally in parking lots, and issues timely warnings, providing comprehensive and powerful decision-making support for scenic area management departments to maintain parking order and public safety within the scenic area.

[0078] In practical applications, the parking space detection method and system based on network model and multi-source data fusion in this embodiment have the following beneficial technical effects: 1. Improve the accuracy and reliability of parking space status detection. In the early stages of model construction, advanced image preprocessing and enhancement techniques were employed, including nonlocal mean denoising algorithms and the Criminisi image inpainting algorithm based on partial differential equations. These methods effectively eliminated random noise introduced during image acquisition and accurately repaired areas of missing or blurred image information caused by vegetation obstruction, building shadows, etc., ensuring high-quality and complete image data input to subsequent classification modules. Combined with the Bayesian classifier's fine-grained classification capabilities for pixel features and a dynamic update mechanism based on iteratively calculated state propagation probabilities, the system can accurately determine whether a parking space is "occupied" or "unoccupied."

[0079] 2. Real-time adaptive capability. By introducing a real-time dynamic optimization model, the system can continuously monitor and analyze changes in vehicle traffic flow within the scenic area, such as differences in vehicle entry and exit frequencies during peak and off-peak seasons, weekdays and weekends, and daytime and nighttime. Utilizing the gradient descent optimization algorithm, key model parameters (such as the weights of state transit probabilities) are automatically and promptly adjusted, enabling the entire system to flexibly adapt to the complex and ever-changing parking demands and vehicle flow patterns within the scenic area. This feature ensures that the system maintains high detection accuracy and response speed even during high-traffic periods, significantly improving the timeliness and applicability of parking management.

[0080] 3. Improved computational efficiency. A deep optimization of the parking space combination state network model is performed based on maximum a posteriori probability theory. An innovative distributed computing architecture is adopted, decomposing the complex model optimization task into multiple computing nodes for parallel processing. Each node independently updates its parameters through a distributed optimization algorithm, and finally, through collaborative communication, merges the results to form the global optimal solution. This parallel processing mechanism significantly reduces the time required for model optimization from several hours using traditional methods to minutes, thus enabling rapid and accurate reflection of the real-time status of each parking space. This effectively meets the stringent requirements for timely parking space detection in scenic areas with rapid vehicle entry and exit.

[0081] 4. Enhance the system's comprehensive service and decision support capabilities. It not only integrates parking space detection data but also incorporates information from multiple data sources, such as surveillance cameras and visitor flow monitoring equipment. A tensor decomposition-based fusion method effectively extracts and integrates feature information from multiple data sources. Based on this, a Long Short-Term Memory (LSTM) network model is used to accurately predict parking demand, and a Convolutional Neural Network (CNN) model is used to analyze surveillance video in real time to monitor abnormal situations around parking spaces (such as illegal parking and abnormal crowd gatherings). This provides scenic area management departments with comprehensive and in-depth decision-making support, enabling them to take proactive measures such as personnel allocation and activating backup parking lots, thereby effectively improving the overall operational efficiency and public safety level of the scenic area.

[0082] 5. An efficient real-time information dissemination mechanism significantly improves the parking experience for tourists and the traffic conditions within the scenic area. Processed parking space status information is pushed to drivers' mobile applications in real time via a message queue system and is deeply integrated with the scenic area's intelligent navigation system, enabling dynamic planning of optimal parking routes for tourists. This not only reduces the time tourists spend searching for parking spaces and alleviates traffic congestion caused by blindly detouring, but also further enhances the service quality and modernization level of the scenic area's management.

[0083] like Figure 4As shown, the present invention also provides a device including a processor 210, a communication interface 220, a memory 230 for storing a processor-executable computer program, and a communication bus 240. The processor 210, communication interface 220, and memory 230 communicate with each other via the communication bus 240. The processor 210 executes the executable computer program to implement the aforementioned parking space detection method based on a network model and multi-source data fusion.

[0084] The computer program in memory 230, when implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0085] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected based on actual needs to achieve the purpose of this embodiment. Those skilled in the art can understand and implement this without any creative effort.

[0086] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0087] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A parking space detection method based on network model and multi-source data fusion, characterized in that, The method comprises: Step S1: constructing a network model according to a parking space state node set, establishing a directed branch between nodes according to a pending state transition probability matrix, and obtaining a parking space combination state network model; Step S2: collecting image information of each node in the parking space combination state network model, and pre-processing and enhancing the collected image information; Step S3: identifying the parking space state by classifying the pixels of the pre-processed and enhanced image information, and updating the pending state transition probability in the parking space combination state network model; Step S4: optimizing the updated parking space combination state network model according to the maximum a posteriori probability, and decomposing and parallel processing the model optimization task; Step S5: analyzing the multi-source data by using the optimized parking space combination state network model, and publishing the parking space state information in real time.

2. The parking space detection method based on network model and multi-source data fusion according to claim 1, characterized in that, In step S1, the parking space state node set includes the occupied state and unoccupied state nodes of each parking space, and the elements of the pending state transition probability matrix are used to represent the pending state transition probability of the parking space state change, which is set as the directed branch between the two nodes before and after the state change. 3.The parking space detection method based on network model and multi-source data fusion according to claim 1, characterized in that, In step S2, the pre-processing and enhancement of the collected image information include: using a denoising algorithm to denoise the collected image information, and using an image repair algorithm to reconstruct and repair the pixels of the denoised image information.

4. The parking space detection method based on network model and multi-source data fusion according to claim 3, characterized in that, In step S2, a non-local mean denoising algorithm is used to denoise the image information, and an image repair algorithm based on partial differential equations is used to reconstruct and repair the pixels of the damaged area of the denoised image information.

5. The parking space detection method based on network model and multi-source data fusion according to claim 1, characterized in that, In step S3, the pixels of the pre-processed and enhanced image information are classified by using a Bayesian classifier, the posterior probability of each pixel belonging to the occupied or unoccupied state is calculated, and the parking space state is identified.

6. The parking space detection method based on network model and multi-source data fusion according to claim 1, characterized in that, In step S3, the pending state transition probability in the parking space combination state network model is updated, including: iteratively calculating the classification results obtained by the Bayesian classifier and the pending state transition probability matrix, and dynamically updating the pending state transition probability in the pending state transition probability matrix.

7. The parking space detection method based on network model and multi-source data fusion according to claim 1, characterized in that, In step S4, the updated parking space combination state network model is optimized according to the maximum a posteriori probability, including: based on the maximum a posteriori probability criterion, finding the model parameters of the parking space combination state network model that maximize the posterior probability. 8.The parking space detection method based on network model and multi-source data fusion according to claim 1, characterized in that, In step S4, the optimization task of the parking space combination state network model is decomposed using a distributed computing architecture, the decomposed tasks are distributed to multiple computing nodes for parallel processing, and each computing node independently completes local calculation and realizes global parameter optimization through communication coordination. 9.The parking space detection method based on network model and multi-source data fusion according to claim 1, characterized in that, Step S5 comprises: Collecting multi-source data including real-time parking space state data, video monitoring data and tourist flow data, using a data fusion method based on tensor decomposition to extract common features of the multi-source data and eliminate redundancy and conflicts between the data, and obtaining fused data; Using a long short-term memory network model to predict the parking space demand change in the future period according to the historical fused data, and using a convolutional neural network model to analyze the real-time fused data to detect the parking space state information, abnormal events and illegal behaviors in the parking area; By establishing a message pushing mechanism, the detected parking space state information is sent to the mobile terminal of the vehicle owner in real time, and a data interface is established with the intelligent navigation system of the scenic spot to provide optimal parking route planning and real-time parking guidance services for the vehicle owner, wherein the optimal parking route planning adopts a dynamic path optimization algorithm based on real-time traffic conditions and parking availability. 10.A parking space detection system based on network model and multi-source data fusion, characterized in that, The system comprises a detection server, wherein the detection server comprises: a model construction module, configured to construct a network model according to a parking space state node set, establish a directed branch between nodes according to a pending state transition probability matrix, and obtain a parking space combination state network model; an image acquisition module, configured to acquire image information of each node in the parking space combination state network model, and perform preprocessing and enhancement processing on the acquired image information; an update module, configured to identify the parking space state by pixel classification on the preprocessed and enhanced image information, and update the pending state transition probability in the parking space combination state network model; a model optimization module, configured to optimize the updated parking space combination state network model according to maximum a posteriori probability, decompose and parallelly process the model optimization task; and a detection and release module, configured to analyze multi-source data by using the optimized parking space combination state network model, and release the parking space state information in real time.

Images