Fuel truck occupation identification method based on intelligent parking system

Abstract

The invention discloses an intelligent parking system-based fuel tank vehicle occupation identification method, which comprises the following steps of: acquiring an occupation position area of vehicle parking, and shooting a first image of a vehicle entering a parking area when an occupation sensor signal is triggered; feature extraction is carried out on the vehicle in the first image, after a vehicle type result is obtained, a proper parking space type is matched from a pre-established parking space distribution rule, real-time path data required by parking navigation is generated, and a driver is guided to park; shooting a second image after the vehicle is parked in the distributed parking space; and analyzing the consistency of the second image and the first image by adopting an image comparison algorithm, judging whether the vehicle is correctly parked in the allocated parking space or not, obtaining a final confirmation result of the occupying state, updating an occupying state database of the parking lot, and pushing the occupying state database to other users through a mobile application to obtain available parking space information. According to the invention, intelligent identification, accurate guidance and automatic management of the fuel vehicle are realized, the operation efficiency of the parking lot is improved, and the user experience is optimized.

Description

Technical Field

[0001] The present invention belongs to the technical field of parking management, and particularly relates to a method for identifying fuel vehicle occupancy based on an intelligent parking system. Background Art

[0002] With the rapid popularization of new energy vehicles, parking lots are facing a complex technical challenge: how to accurately identify fuel vehicles and new energy vehicles among limited parking resources, achieve intelligent parking space allocation management, ensure that charging pile spaces are preferentially provided for new energy vehicles, and at the same time prevent fuel vehicles from mistakenly occupying these dedicated spaces. The core of this problem is that the parking lot needs to quickly determine the vehicle type at the moment of vehicle entry and allocate a suitable parking space for it, especially reserving the spaces with charging piles for new energy vehicles. However, it is difficult for traditional occupancy sensors to accurately distinguish vehicle types, which may lead to fuel vehicles mistakenly occupying new energy dedicated spaces and reducing the utilization efficiency of parking resources. At the same time, how to guide the vehicle to the correct parking space in real time during the driving process and how to ensure that the vehicle finally parks in the designated position are key issues that need to be solved. Therefore, how to quickly and accurately judge the vehicle type in a complex and changeable parking lot environment, make intelligent parking space allocation decisions based on this, maximize the utilization efficiency of parking resources, especially ensure that new energy vehicles can conveniently use charging facilities, and then build an efficient, accurate, and real-time intelligent parking management system is a major technical challenge faced by current parking lots. Summary of the Invention

[0003] To solve the above technical problems, the present invention provides a method for identifying fuel vehicle occupancy based on an intelligent parking system, including:

[0004] Obtaining the occupancy position area where the vehicle parks through occupancy sensors installed on the ground, and taking a first image of the vehicle entering the parking area when the signal of the occupancy sensor is triggered;

[0005] Using a deep learning algorithm to extract features of the vehicle in the first image and analyzing to obtain the vehicle type result;

[0006] According to the vehicle type result, matching a suitable parking space category from the pre-established parking space allocation rules, generating real-time path data required for parking navigation, and guiding the driver to park;

[0007] After the vehicle parks in the allocated parking space, taking a second image of the parking space;

[0008] Using an image comparison algorithm to analyze the consistency between the second image and the first image, judging whether the vehicle correctly parks in the allocated parking space, and obtaining the final confirmation result of the occupancy status;

[0009] According to the final confirmation result, update the occupancy status database of the parking lot and push it to other users through the mobile application to obtain information on available parking spaces.

[0010] Preferably, the process of obtaining the occupancy position area where the vehicle parks through a ground-mounted occupancy sensor includes:

[0011] When the ground-mounted occupancy sensor senses the vehicle weight or interrupts the infrared beam, an initial electronic signal containing position information is generated, and the signal is transmitted to the server through the wireless network. After receiving the signal, the server uses a preset threshold to judge the signal strength and determine the vehicle occupancy status;

[0012] If the signal strength exceeds the threshold, analyze the position information of the parking area through the clustering algorithm to obtain the distribution of the occupied areas;

[0013] According to the distribution of the occupied areas, obtain historical signal data and judge the occupancy time period;

[0014] For the occupancy time period, use the decision tree algorithm to predict the occupancy trend and obtain the prediction result;

[0015] Adjust the sensing frequency of the occupancy sensor through the prediction result to determine the optimized trigger interval.

[0016] Preferably, the process of taking the first image of the vehicle entering the parking area when the occupancy sensor signal is triggered includes:

[0017] Based on the high-definition camera, when the occupancy sensor detects that the vehicle enters the parking lot, obtain the first image through signal triggering;

[0018] Extract the original image data containing the vehicle body structure and license plate number from the first image through a preset image processing algorithm;

[0019] Use the optical character recognition algorithm to analyze the license plate number in the original image data to obtain the license plate text data;

[0020] If the license plate text data matches the pre-established database, determine the vehicle identity information through comparison;

[0021] According to the vehicle identity information, obtain the corresponding parking space occupancy record from the database and judge the occupancy status;

[0022] Transmit the judgment result to the terminal device through the server to update the parking space status data.

[0023] Preferably, the process of using the deep learning algorithm to extract the features of the vehicle in the first image and analyze to obtain the vehicle type result includes:

[0024] Preprocess the first image using a preset image processing tool, and then analyze the vehicle features in the image using a deep learning algorithm to extract the positions of the exhaust pipes and the features of the fuel inlets, obtaining a feature set;

[0025] According to the positions of the exhaust pipes and the features of the fuel inlets in the feature set, use a machine learning classification algorithm to determine the vehicle type, obtaining a type identifier;

[0026] If the type identifier matches the preset vehicle type library, query the database to obtain associated information and get extended data;

[0027] According to the extended data and the type identifier, use rule judgment to determine the vehicle type result.

[0028] Preferably, the process of matching a suitable parking space category from the pre-established parking space allocation rules according to the vehicle type result includes:

[0029] Extract the vehicle type features according to the vehicle type result, and use a matching algorithm to search for the corresponding parking space category in the pre-established allocation rules;

[0030] Generate an allocation instruction containing the parking space number according to the matching process, and transmit the instruction data to the cloud computing platform;

[0031] The cloud computing platform receives the allocation instruction and determines the allocation scheme according to the load balancing algorithm;

[0032] Obtain the parking space number in the allocation scheme and update the parking space status in the server database through the instruction;

[0033] For the updated parking space status, judge that if the occupancy rate exceeds the preset threshold, then regenerate an adjustment scheme through the cloud computing platform;

[0034] Obtain the final parking space allocation result from the adjustment scheme and push it to the terminal device through the server.

[0035] Preferably, the process of generating real-time path data required for parking navigation to guide the driver to park includes:

[0036] Obtain a transmission instruction through a wireless network and determine that the instruction content contains guiding information;

[0037] If the instruction contains arrow indications and color coding, display the corresponding information through an LED display screen to obtain real-time guiding data;

[0038] According to the real-time guiding data, obtain the path data in the parking lot and judge whether the path is available;

[0039] If the path is available, update the display information through the LED display screen to determine the real-time path required for the driver's navigation;

[0040] Collect path data in real time, determine whether the path data has changed, and obtain the change trend;

[0041] According to the change trend, adjust the transmission instruction through the wireless network to determine the updated guidance information;

[0042] Display the updated arrow indication and color coding through the LED display screen, and obtain the adjusted path data.

[0043] Preferably, after the vehicle is parked in the allocated parking space, the process of taking the second image for the parking space includes:

[0044] Take the second image of the allocated parking space, extract the license plate number and parking timestamp through image processing technology to obtain the occupancy evidence data;

[0045] Transmit the occupancy evidence data to the server through the data channel, and use a compression algorithm to reduce the transmission load to obtain a transmission completion confirmation;

[0046] The server stores the received occupancy evidence data and determines the storage location through database indexing technology;

[0047] If the license plate number matches the preset database, judge the vehicle identity through a comparison algorithm to obtain the identity verification result;

[0048] According to the identity verification result, obtain the associated data between the parking timestamp and the vehicle identity, and determine the parking duration through time series analysis;

[0049] According to the parking duration and the data of the allocated parking space, judge the occupancy status using rules to obtain the parking space usage record;

[0050] Update the occupancy status through the parking space usage record and the data stored in the server to obtain the latest parking space allocation information.

[0051] Preferably, the process of analyzing the consistency between the second image and the first image using an image comparison algorithm to determine whether the vehicle is correctly parked in the allocated parking space and obtaining the final confirmation result of the occupancy status includes:

[0052] Denoise the second image through a preset image preprocessing method to obtain the processed second image;

[0053] Extract the feature points of the first image and the processed second image, and correspondingly obtain the feature descriptors of the first image and the processed second image;

[0054] Use an image comparison algorithm to analyze the feature descriptors of the processed second image and the first image to obtain a consistency score;

[0055] If the consistency score exceeds the preset threshold, it is determined that the vehicle has been correctly parked in the allocated parking space, and the occupancy status is determined to be occupied;

[0056] By analyzing the judgment result, obtain the deviation value of the vehicle parking position, and obtain the position adjustment suggestion data;

[0057] According to the position adjustment suggestion data, determine whether the vehicle parking completely conforms to the boundary of the allocated parking space, and obtain the final confirmation result;

[0058] Use the recording module to store the associated data of the final confirmation result and the second image, and determine the persistent information of the occupancy status.

[0059] Preferably, according to the final confirmation result, the process of updating the occupancy status database of the parking lot and pushing it to other users through the mobile application to obtain available parking space information includes:

[0060] Collect the occupancy situation of the parking spaces in the parking lot, judge the occupancy status and generate the initial occupancy record;

[0061] Compare the initial occupancy record with the preset rules, and if they match, determine it as the final confirmation result;

[0062] Extract the parking space number and vehicle information from the final confirmation result, and update the occupancy status in the parking lot database to the latest record;

[0063] Use database query technology to obtain the updated available parking space information;

[0064] Through the mobile application interface, convert the available parking space information into push data, and transmit the push data to the user-side mobile application;

[0065] According to the feedback from the user side, judge whether the push information is successful. If not, re-obtain the available parking space information and push it.

[0066] Preferably, the method further includes:

[0067] The server extracts the vehicle parking timestamp and license plate number from the occupancy status database, calculates the parking duration through the timestamp, and obtains the charging basic data;

[0068] Process the charging basic data through the preset charging rules, generate the charging data including the parking duration, and determine the deducted amount;

[0069] Obtain the user account binding information from the payment system, transmit the charging data to the corresponding account through the interface, and judge whether the account balance is sufficient;

[0070] If the account balance is sufficient, perform the deduction operation through the payment system interface to complete the automatic deduction process;

[0071] If the account balance is insufficient, a prompt message is generated and transmitted to the device associated with the user account through the interface to obtain the user's payment supplement status;

[0072] Update the account balance according to the payment supplement status, and perform the deduction operation again through the payment system interface to obtain a record of successful deduction;

[0073] For the record of successful deduction, update the occupancy status from the database, determine whether the vehicle has left the parking space, and determine the end of the business process.

[0074] Compared with the prior art, the present invention has the following advantages and technical effects:

[0075] The present invention realizes the automatic identification, classification and guidance of fuel vehicles entering the parking lot through the occupancy sensors and high-definition cameras installed on the ground. First, the vehicle information is collected by the occupancy sensors and cameras, and then the vehicle characteristics are analyzed by using deep learning algorithms to determine the vehicle type. Based on the preset allocation rules, the system matches a suitable parking space for the vehicle and provides real-time navigation for the driver through the LED display screen. After the vehicle is parked, the present invention takes pictures again for comparison, confirms the occupancy status, and updates the database. Finally, the system automatically calculates the parking duration and completes the deduction. This intelligent management method not only improves the operation efficiency of the parking lot, but also provides a convenient parking experience for users, and at the same time realizes the precise management of fuel vehicles and the optimal allocation of resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] The drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0077] Figure 1 It is a schematic flowchart of the method of the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0078] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will refer to the drawings and combine the embodiments to detail this application.

[0079] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0080] As Figure 1 shown, in this embodiment, a method for identifying the occupancy of fuel vehicles based on an intelligent parking system is provided, including:

[0081] Obtain the occupied position area where the vehicle parks through a ground-mounted occupancy sensor, and when the occupancy sensor signal is triggered, capture a first image of the vehicle entering the parking area;

[0082] Use a deep learning algorithm to extract features of the vehicle in the first image and analyze to obtain the vehicle type result;

[0083] According to the vehicle type result, match the suitable parking space category from the pre-established parking space allocation rules, generate the real-time path data required for parking navigation, and guide the driver to park;

[0084] After the vehicle parks in the allocated parking space, capture a second image of the parking space;

[0085] Use an image comparison algorithm to analyze the consistency between the second image and the first image, determine whether the vehicle is correctly parked in the allocated parking space, and obtain the final confirmation result of the occupancy status;

[0086] According to the final confirmation result, update the occupancy status database of the parking lot and push it to other users through a mobile application to obtain available parking space information.

[0087] Furthermore, the process of obtaining the occupied position area where the vehicle parks through a ground-mounted occupancy sensor includes:

[0088] When the ground-mounted occupancy sensor senses the vehicle weight or interrupts the infrared beam, generate an initial electronic signal containing position information, transmit the signal to the server through a wireless network, and after the server receives the signal, use a preset threshold to judge the signal strength and determine the vehicle occupancy status;

[0089] If the signal strength exceeds the threshold, analyze the position information of the parking area through a clustering algorithm to obtain the occupancy area distribution;

[0090] According to the occupancy area distribution, obtain historical signal data and judge the occupancy time period;

[0091] For the occupancy time period, use a decision tree algorithm to predict the occupancy trend and obtain the prediction result;

[0092] Adjust the sensing frequency of the occupancy sensor through the prediction result to determine the optimized trigger interval.

[0093] Exemplarily, a ground-mounted occupancy sensor senses the weight of a vehicle through a pressure sensor. When a vehicle parks on the sensor and the sensor detects that the pressure value exceeds a preset threshold (e.g., 500 kg), the sensor activates the infrared beam detection. If the infrared beam is interrupted for more than 3 seconds, the sensor determines that the vehicle is occupying the space. The microprocessor built into the sensor generates an initial electronic signal containing location information based on the sensor data. The location information is obtained through a GPS module with an accuracy of up to 1 meter. The signal is transmitted to the server via the wireless network LoRa or NB-IoT at a transmission frequency of once every 10 seconds to ensure real-time performance. After receiving the signal, the server analyzes the signal using the random forest algorithm and makes a preliminary judgment in combination with the historical occupancy record data of the past 24 hours. The judgment accuracy can reach 95%. If it is determined that the space is occupied, the server stores the information in the database and triggers subsequent business logics, such as notifying the management personnel or updating the parking space status. The entire process requires no manual intervention and realizes automated processing.

[0094] Further, the process of capturing a first image of a vehicle entering the parking area when the occupancy sensor signal is triggered includes:

[0095] Based on a high-definition camera, when the occupancy sensor detects that a vehicle enters the parking lot, a first image is obtained through signal triggering;

[0096] According to the first image, raw image data including the vehicle body structure and license plate number is extracted through a preset image processing algorithm;

[0097] An optical character recognition algorithm is used to parse the license plate number in the raw image data to obtain license plate text data;

[0098] If the license plate text data matches the pre-established database, the vehicle identity information is determined through comparison;

[0099] According to the vehicle identity information, the corresponding parking space occupancy record is obtained from the database to judge the occupancy status;

[0100] The judgment result is transmitted to the terminal device through the server to update the parking space status data.

[0101] Exemplarily, the high-definition camera uses a 2-million-pixel CMOS sensor to monitor the parking space area in real time at a rate of 30 frames per second. When the ultrasonic occupancy sensor detects that the distance value is lower than 5 meters and lasts for 500 milliseconds, the image acquisition module is triggered. The camera compresses the image through the H.264 encoding algorithm and transmits the raw data with a resolution of 1280×720 through 5GHz band Wi-Fi. The transmission bandwidth requirement is not less than 4Mbps to ensure that the image delay is less than 200 milliseconds. After receiving the data, the server first performs Gaussian filtering denoising with σ = 5 using the GaussianBlur function of OpenCV, and then uses the YOLOv5 model for vehicle detection. The mAP@5 of this model trained on the COCO dataset reaches 89. For the detected vehicle area, the ROI is extracted through a license plate localization algorithm based on CNN. In this embodiment, this algorithm uses 3 convolutional kernels (5×5, 3×3, 1×1) for feature extraction, and the localization accuracy reaches 92%. In the license plate recognition stage, the CRNN network structure is used, and the number of LSTM units is 128. After training on a local dataset containing 500,000 license plates, the character recognition accuracy is 97%. When the system matches the recognition result with the database, an improved Levenshtein distance algorithm (threshold set to 2) is used to handle possible recognition errors, and at the same time, the timestamp (accurate to milliseconds) and GPS coordinates (WGS84 coordinate system, accuracy ±5 meters) are recorded to form a complete parking event record. All processing processes are completed in a Docker container. The average time-consuming for a single request is 380 milliseconds. The server is configured with 4 cores and 8G of memory and can support 50 concurrent processes per second.

[0102] Further, the process of using a deep learning algorithm to extract features of the vehicle in the first image and analyze to obtain the vehicle type result includes:

[0103] Preprocess the first image using a preset image processing tool, and then use a deep learning algorithm to analyze the vehicle features in the image, extract the exhaust pipe position and fuel filler port features, and obtain a feature set;

[0104] According to the exhaust pipe position and fuel filler port features in the feature set, use a machine learning classification algorithm to judge the vehicle type and obtain a type identifier;

[0105] If the type identifier matches the preset vehicle type library, obtain the associated information through database query to get the extended data;

[0106] According to the extended data and the type identifier, use rule judgment to determine the vehicle type result.

[0107] Exemplarily, after the server receives the vehicle image uploaded by the client, it first preprocesses the image through the YOLOv5 object detection algorithm, uses an input resolution of 640×640 pixels, sets the confidence threshold to 8, and identifies the position and bounding box of the vehicle in the image. Subsequently, the ResNet50 convolutional neural network is used to extract features from the vehicle area, and a 1024-dimensional feature vector is output at the last fully connected layer. For exhaust pipe positioning, the U-Net segmentation network is used to perform pixel-level segmentation on the vehicle bottom area, sets the learning rate to 0.01, and reaches an IoU index of 95% after 50 rounds of training, accurately identifying the precise position coordinates of the exhaust pipe (x = 325, y = 580). For the fuel filler opening feature, the MobileNetV3 lightweight network is used for key point detection, and the center coordinates (x = 420, y = 210) of the fuel filler cap and a circular area with a diameter of 35 pixels are marked in the input image. The extracted features are input into a pre-trained XGBoost classifier, which contains 500 decision trees and has a maximum depth of 6. Through the training of 20,000 vehicle samples, it can distinguish between fuel vehicles and new energy vehicle types with an accuracy of 97%. Finally, the recognition result (such as "Automobile - Single Exhaust Pipe on the Left") is stored together with the vehicle VIN code and detection timestamp in the vehicle_info table of the MySQL database. The InnoDB engine is used and a hash index of the VIN code is established to optimize the query performance. The entire processing flow takes an average of 320 milliseconds on the NVIDIA T4 GPU, and the Redis cache is used to store the most recently processed 1000 records to improve the response speed of repeated queries.

[0108] Further, the process of matching a suitable parking space category from the pre-established parking space allocation rules according to the vehicle type result includes:

[0109] Extract the vehicle type features according to the vehicle type result, and use a matching algorithm to find the corresponding parking space category in the pre-established allocation rules;

[0110] Generate an allocation instruction containing the parking space number according to the matching process, and transmit the instruction data to the cloud computing platform;

[0111] The cloud computing platform receives the allocation instruction and determines the allocation plan according to the load balancing algorithm;

[0112] Obtain the parking space number in the allocation plan and update the parking space status in the server database through the instruction;

[0113] For the updated parking space status, judge that if the occupancy rate exceeds the preset threshold, then regenerate an adjustment plan through the cloud computing platform;

[0114] Obtain the final parking space allocation result from the adjustment plan and push it to the terminal device through the server.

[0115] Exemplarily, after the server receives a request from a user for a vehicle type of SUV (such as a vehicle length of 8 meters and a wheelbase of 8 meters), it first calls a pre - set parking space matching rule library. This library is constructed using a decision tree algorithm and contains three - layer judgment logic: The first layer filters out parking spaces in Area A (2 meters × 5 meters) and Area C (0 meters × 3 meters) based on the vehicle length (threshold of 5 meters); the second layer excludes Area C through the wheelbase (threshold of 7 meters); the third layer combines real - time occupancy data (8 standard parking spaces remaining in Area A) to confirm the final area. This embodiment uses an improved First - Fit Algorithm (FFA) for allocation, calculates the center - point coordinates of each parking space in the current Area A (such as A - 12: x = 36, y = 12), takes the user entrance position (x = 10, y = 5) as the starting point, calculates the optimal path (minimum Manhattan distance of 43 meters) through the Dijkstra algorithm, and finally selects Parking Space A - 12. The cloud computing platform, based on the principle of the Kubernetes scheduler, completes resource allocation within 200 milliseconds, generates a JSON instruction containing the parking space coordinates (36, 12), the navigation path (sequence of turning points [(15, 5), (25, 12), (36, 12)]) and the effective duration (30 minutes), pushes it to the user terminal via the MQTT protocol, and simultaneously updates the parking space status database (MySQL row locks ensure concurrency safety, version number +1). Throughout the process, the load balancer (Nginx weighted round - robin, weight coefficient 7) monitors the CPU utilization rate (threshold of 70%) of each regional server in real - time and dynamically adjusts the calculation node allocation.

[0116] Further, the process of generating real - time path data required for parking navigation and guiding the driver to park includes:

[0117] Obtain a transmission instruction through a wireless network and determine that the instruction content contains guiding information;

[0118] If the instruction contains arrow indications and color coding, display the corresponding information through an LED display screen to obtain real - time guiding data;

[0119] According to the real - time guiding data, obtain the path data in the parking lot and determine whether the path is available;

[0120] If the path is available, update the display information through the LED display screen to determine the real - time path required for the driver's navigation;

[0121] Collect path data in real - time, determine whether the path data has changed, and obtain the change trend;

[0122] According to the change trend, adjust the transmission instruction through the wireless network to determine the updated guiding information;

[0123] The updated arrow indication and color coding are displayed on the LED display screen to obtain the adjusted path data.

[0124] Exemplarily, the allocation instruction is transmitted to the LED display screen in the parking lot through the wireless network. First, it is necessary to use a wireless communication module based on the TCP / IP protocol to send the instruction generated by the central control system to the display screen in the form of data packets.

[0125] For example, the central control system transmits data through the Wi-Fi module in the 4GHz frequency band. The size of each data packet is 128 bytes, including arrow indication, color coding, and path information. After receiving the data packet, the display screen extracts the valid information through a parsing algorithm. For example, the CRC32 check algorithm is used to ensure data integrity. If the check passes, the process continues. Next, the display screen calls the preset display algorithm according to the parsed arrow indication and color coding, and renders the information onto the LED screen in real time.

[0126] For example, the arrow indication is drawn using the Bresenham algorithm, and the color coding is dynamically adjusted according to the RGB values (such as red is 255, 0, 0, and green is 0, 255, 0). At the same time, the system calculates the optimal path through a real-time path planning algorithm (such as the Dijkstra algorithm) and synchronously updates the path data with the display screen information.

[0127] For example, when it is detected that a certain path is congested, the system will recalculate the path and send the new arrow indication and color coding to the display screen to ensure that the driver can obtain the latest navigation information in a timely manner. The entire process is implemented through multi-threaded technology for parallel processing to ensure the real-time nature of data transmission and display. For example, a thread pool is used to manage data transmission and display tasks, and each thread processes an independent data packet to minimize latency to the greatest extent.

[0128] Further, after the vehicle parks in the allocated parking space, the process of taking the second image of the parking space includes:

[0129] Taking the second image of the allocated parking space, extracting the license plate number and parking timestamp through image processing technology to obtain occupancy evidence data;

[0130] Transmitting the occupancy evidence data to the server through the data channel, using a compression algorithm to reduce the transmission load, and obtaining a transmission completion confirmation;

[0131] The server stores the received occupancy evidence data and determines the storage location through database indexing technology;

[0132] If the license plate number matches the preset database, the vehicle identity is judged through a comparison algorithm to obtain an identity verification result;

[0133] According to the identity verification result, the associated data of parking timestamp and vehicle identity is obtained, and the parking duration is determined through time series analysis;

[0134] According to the parking time and the data of the allocated parking space, the occupancy status is judged by rules to obtain the parking space usage record;

[0135] Through the parking space usage records and the data stored on the server, the occupancy status is updated to obtain the latest parking space allocation information.

[0136] Exemplarily, a high-definition camera uses a 2-megapixel sensor and H.265 encoding to monitor the parking area in real time at a rate of 25 frames per second. When a vehicle is detected to be fully parked in the parking space, the image acquisition mechanism is triggered, and the vehicle is detected by the YOLOv5 algorithm, with a confidence threshold set to 85. The system automatically captures the image area containing the license plate and uses a CNN-based license plate recognition model for character segmentation. The model is trained on the basis of the LPRNet architecture and has an accuracy rate of 97% for domestic license plate recognition. During the recognition process, perspective transformation is used to correct license plate images with a tilt angle of more than 15 degrees, and a 7×7 sliding window is used for feature extraction for character segmentation. The timestamp is generated by a server synchronized with the NTP protocol, accurate to milliseconds, and encapsulated with the image data in JSON format, including fields such as {"plate":"京A12345","timestamp":"2023-08-15T14:25:3789"}. The data transmission adopts MQTT protocol, and the QoS level is set to 1. After AES-256 encryption, it is transmitted to the cloud with a bandwidth of 2MB per second. After receiving the data, the server first performs SHA-256 verification and then writes it into the parking_records table of the MySQL database. The table design contains fields such as plate (VARCHAR12), entry_time (DATETIME3), and image_path (TEXT), and establishes a B+ tree structure with plate and entry_time as the joint index. During the storage process, duplicate license plates will be judged based on the time window. If the same license plate appears repeatedly within 10 minutes, an abnormal alarm will be triggered. The image file is stored in JPEG format with a compression ratio of 85%. The resolution is maintained at 1920×1080. It is stored in a distributed file system by date and directory. A single file occupies an average of about 800KB of space. The system performs data backup every 5 minutes, adopts an incremental backup strategy, and only transmits data blocks with a change of more than 1MB each time.

[0137] Furthermore, the image comparison algorithm is used to analyze the consistency between the second image and the first image, and the process of determining whether the vehicle is correctly parked in the allocated parking space and obtaining the final confirmation result of the parking space occupancy status includes:

[0138] Denoise the second image through a preset image preprocessing method to obtain the processed second image;

[0139] Extract the feature points of the first image and the processed second image, and correspondingly obtain the feature descriptors of the first image and the processed second image;

[0140] Use an image comparison algorithm to analyze the feature descriptors of the processed second image and the first image to obtain a consistency score;

[0141] If the consistency score exceeds the preset threshold, it is determined that the vehicle has parked correctly in the allocated parking space, and the occupancy status is determined to be occupied;

[0142] Analyze the judgment result to obtain the deviation value of the vehicle parking position, and obtain the position adjustment suggestion data;

[0143] According to the position adjustment suggestion data, judge whether the vehicle parking completely conforms to the boundary of the allocated parking space to obtain the final confirmation result;

[0144] Use a recording module to store the associated data of the final confirmation result and the second image, and determine the persistent information of the occupancy status.

[0145] Exemplarily, after receiving the second image, the server first preprocesses the image, including grayscale conversion, denoising, and edge detection, to improve the accuracy of subsequent comparison. Then, use a feature point-based image comparison algorithm, such as SIFT (Scale-Invariant Feature Transform) or ORB (Oriented FAST and Rotated BRIEF), to extract the key feature points in the first image and the second image. Determine the similarity of the two images by calculating the Euclidean distance between the feature points.

[0146] For example, set the similarity threshold to 85. If the calculated similarity is greater than this threshold, the two images are considered to be consistent. Further, use image segmentation technology, such as a deep learning-based semantic segmentation model, to accurately divide the parking space area to ensure the accuracy of the analysis range. Determine whether the vehicle has parked in the allocated parking space by calculating the pixel change in the parking space area.

[0147] For example, if the pixel change rate in the parking space area is lower than 5%, it is considered that the vehicle has parked correctly. Finally, combine the analysis results of the similarity and the pixel change rate to generate the final confirmation result of the occupancy status, such as "correctly parked" or "not correctly parked", and store the result in the database for subsequent query and analysis.

[0148] Furthermore, the process of updating the occupancy status database of the parking lot according to the final confirmation result and pushing it to other users through a mobile application to obtain available parking space information includes:

[0149] Collect the occupancy status of parking spaces in the parking lot, determine the occupancy status, and generate initial occupancy records;

[0150] Compare the initial occupancy records with the preset rules. If they match, determine the result as the final confirmed result;

[0151] Extract the parking space numbers and vehicle information from the final confirmed result, and update the occupancy status in the parking lot database to the latest record;

[0152] Adopt database query technology to obtain the updated available parking space information;

[0153] Through the mobile application interface, convert the available parking space information into push data, and transmit the push data to the user-side mobile application;

[0154] According to the feedback from the user side, determine whether the push information is successful. If not, re-obtain the available parking space information and push it.

[0155] Exemplarily, first receive the real-time data of the occupancy status of parking spaces in the parking lot. For example, the parking space numbered A01 is occupied by a vehicle. The sensor detects that the vehicle length is 8 meters and the width is 8 meters, and confirms the vehicle type as a vehicle through an image recognition algorithm. The module uses a hash algorithm to associate the parking space number A01 with the vehicle information to generate a unique identifier. For example, "3a7bd3e2360a3d29eea436fcfb7e44c735d117c7d8bc7f8a3b9f8c1e2b3c4d5e" is generated through the SHA-256 algorithm. Then, write the updated data into the occupancy status database of the parking lot. The database uses MySQL for storage, and the table structure includes fields such as parking space number, vehicle type, and vehicle size. For example, the inserted record is (A01, vehicle, 8 meters, 8 meters). After the update is completed, the module sends the latest information to the user side through the mobile application push service. The push content uses the JSON format. For example, {"parking space number": "A01", "vehicle type": "vehicle", "available status": "unavailable"}. After receiving the push, the user side parses the JSON data and calls the map API to update the parking space status in real time on the map. For example, mark the A01 parking space as red to indicate that it is unavailable. At the same time, the system analyzes the parking space utilization rate based on historical data. For example, by calculating that the occupancy time of the A01 parking space in the past 24 hours is 18 hours, the utilization rate is 75%, and it is predicted that the parking space may still be unavailable within the next 2 hours, so as to provide more accurate available parking space information for users.

[0156] Furthermore, the method further includes:

[0157] The server extracts the vehicle parking timestamp and license plate number from the occupancy status database, calculates the parking duration through the timestamp, and obtains the basic billing data;

[0158] Process the billing base data according to the preset billing rules, generate billing data including parking duration, and determine the deduction amount;

[0159] Obtain the user account binding information from the payment system, transmit the billing data to the corresponding account through the interface, and judge whether the account balance is sufficient;

[0160] If the account balance is sufficient, execute the deduction operation through the payment system interface to complete the automatic deduction process;

[0161] If the account balance is insufficient, generate a prompt message, transmit it to the device associated with the user account through the interface, and obtain the user's payment supplement status;

[0162] Update the account balance according to the payment supplement status, execute the deduction operation again through the payment system interface, and obtain a record of successful deduction;

[0163] For the record of successful deduction, update the occupancy status from the database, judge whether the vehicle has left the parking space, and determine the end of the business process.

[0164] Exemplarily, the server extracts the vehicle parking timestamp and license plate number from the occupancy status database. First, through the SQL query statement:

[0165] `SELECT timestamp, plate_number FROM parking_records WHERE status = ’occupied'`

[0166] Obtain the parking start time and license plate information of the currently occupied vehicle.

[0167] For example, the query result shows that the vehicle with the license plate "Yue A12345" started parking at 10:00:00 on October 1, 2023. Then, the server calculates the parking duration based on the current time of 12:00:00 on October 1, 2023, using the formula `duration = current_timestamp - start_timestamp`, and gets a parking duration of 2 hours. Then, the server calculates the fee based on the parking duration and the billing rule (such as 10 yuan per hour), with the formula `fee = duration * rate`, and obtains a fee of 20 yuan. After generating the billing data, the server encapsulates the data into the JSON format `{"plate_number": "Yue A12345", "fee": 20}` through the payment system interface, and calls the API interface `POST / payment / charge` of the payment system to complete the automatic deduction process. After receiving the request, the payment system deducts 20 yuan from the user's account and returns a successful deduction response `{"status": "success", "message": "Deduction successful"}`. The entire process is implemented through automation technology to ensure data accuracy and processing efficiency.

[0168] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An identification method for gasoline vehicle occupancy based on an intelligent parking system, characterized in that, Including: Obtain the occupied position area where the vehicle parks through a ground-mounted occupancy sensor, and when the signal of the occupancy sensor is triggered, capture a first image of the vehicle entering the parking area; Use a deep learning algorithm to extract features of the vehicle in the first image and analyze to obtain the vehicle type result; According to the vehicle type result, match a suitable parking space category from the pre-established parking space allocation rules, generate real-time path data required for parking navigation, and guide the driver to park; After the vehicle parks in the allocated parking space, capture a second image of the parking space; Use an image comparison algorithm to analyze the consistency between the second image and the first image, determine whether the vehicle is correctly parked in the allocated parking space, and obtain the final confirmation result of the occupancy status; According to the final confirmation result, update the occupancy status database of the parking lot and push it to other users through a mobile application to obtain available parking space information.

2. The method according to claim 1, wherein: The process of obtaining the occupied position area where the vehicle parks through a ground-mounted occupancy sensor includes: When the ground-mounted occupancy sensor senses the vehicle weight or interrupts the infrared beam, generate an initial electronic signal containing position information, transmit the signal to the server through a wireless network, and after the server receives the signal, use a preset threshold to judge the signal strength and determine the vehicle occupancy status; If the signal strength exceeds the threshold, analyze the position information of the parking area through a clustering algorithm to obtain the occupancy area distribution; According to the occupancy area distribution, obtain historical signal data and judge the occupancy time period; For the occupancy time period, use a decision tree algorithm to predict the occupancy trend and obtain the prediction result; Adjust the sensing frequency of the occupancy sensor through the prediction result to determine the optimized trigger interval.

3. The method according to claim 1, wherein: The process of capturing a first image of the vehicle entering the parking area when the signal of the occupancy sensor is triggered includes: Based on a high-definition camera, when the occupancy sensor detects that the vehicle enters the parking lot, obtain the first image through signal triggering; Extract the original image data including the vehicle body structure and license plate number from the first image through a preset image processing algorithm; Use an optical character recognition algorithm to analyze the license plate number in the original image data to obtain the license plate text data; If the license plate text data matches the pre-established database, determine the vehicle identity information through comparison; According to the vehicle identity information, obtain the corresponding parking space occupancy record from the database and judge the occupancy status; Transmit the judgment result to the terminal device through the server and update the parking space status data.

4. The method according to claim 1, wherein: The process of using a deep learning algorithm to extract features of the vehicle in the first image and analyze to obtain the vehicle type result includes: Use a preset image processing tool to preprocess the first image, and then use a deep learning algorithm to analyze the vehicle features in the image, extract the exhaust pipe position and fuel filler port features to obtain a feature set; Based on the exhaust pipe position and fuel filler port characteristics in the feature set, use a machine learning classification algorithm to determine the vehicle type and obtain a type identifier; If the type identifier matches the preset vehicle type library, obtain associated information through database query to get extended data; Based on the extended data and type identifier, use rule judgment to determine the vehicle type result.

5. The method according to claim 1, wherein: The process of matching a suitable parking space category from the pre-established parking space allocation rules according to the vehicle type result includes: Extract vehicle type characteristics according to the vehicle type result, and use a matching algorithm to search for the corresponding parking space category in the pre-established allocation rules; Generate an allocation instruction containing the parking space number according to the matching process, and transmit the instruction data to the cloud computing platform; The cloud computing platform receives the allocation instruction and determines the allocation scheme according to the load balancing algorithm; Obtain the parking space number in the allocation scheme and update the parking space status in the server database through an instruction; For the updated parking space status, judge that if the occupancy rate exceeds the preset threshold, then regenerate an adjustment scheme through the cloud computing platform; Obtain the final parking space allocation result from the adjustment scheme and push it to the terminal device through the server.

6. The method according to claim 1, wherein: The process of generating real-time path data required for parking navigation to guide the driver to park includes: Obtain a transmission instruction through a wireless network and determine that the instruction content contains guiding information; If the instruction contains arrow indications and color coding, display the corresponding information through an LED display screen to obtain real-time guiding data; According to the real-time guiding data, obtain the path data in the parking lot and judge whether the path is available; If the path is available, update the display information through the LED display screen to determine the real-time path required for the driver's navigation; Collect path data in real time, judge whether the path data has changed, and obtain a change trend; According to the change trend, adjust the transmission instruction through the wireless network to determine the updated guiding information; Display the updated arrow indications and color coding through the LED display screen to obtain the adjusted path data.

7. The method according to claim 1, wherein: The process of taking a second image of the parking space after the vehicle parks in the allocated parking space includes: Take a second image of the allocated parking space, and extract the license plate number and parking timestamp through image processing technology to obtain occupancy evidence data; Transmit the occupancy evidence data to the server through a data channel, and use a compression algorithm to reduce the transmission load to obtain a transmission completion confirmation; The server stores the received occupancy evidence data and determines the storage location through database indexing technology; If the license plate number matches the preset database, judge the vehicle identity through a comparison algorithm to obtain an identity verification result; According to the identity verification result, obtain the associated data of the parking timestamp and vehicle identity, and determine the parking duration through time series analysis; According to the parking duration and the data of the allocated parking space, use rule judgment to determine the occupancy status to obtain a parking space usage record; Update the occupancy status through the parking space usage record and the data stored in the server to obtain the latest parking space allocation information.

8. The method according to claim 1, wherein the process of analyzing the consistency between the second image and the first image by using an image comparison algorithm, determining whether the vehicle is correctly parked in the assigned parking space, and obtaining the final confirmation result of the occupancy status includes: denoising the second image by a preset image preprocessing method to obtain a processed second image; extracting feature points of the first image and the processed second image, and correspondingly obtaining feature descriptors of the first image and the processed second image; analyzing the feature descriptors of the processed second image and the first image by using an image comparison algorithm to obtain a consistency score; if the consistency score exceeds a preset threshold, it is determined that the vehicle is correctly parked in the assigned parking space, and the occupancy status is determined to be occupied; obtaining a deviation value of the vehicle parking position by analyzing the judgment result to obtain position adjustment suggestion data; judging whether the vehicle parking completely conforms to the boundary of the assigned parking space according to the position adjustment suggestion data to obtain the final confirmation result; using a recording module to store the associated data of the final confirmation result and the second image to determine the persistent information of the occupancy status.

9. The method according to claim 1, wherein the process of updating the occupancy status database of the parking lot according to the final confirmation result and pushing it to other users through a mobile application to obtain available parking space information includes: collecting the occupancy situation of parking spaces in the parking lot, judging the occupancy status and generating an initial occupancy record; comparing the initial occupancy record with a preset rule, and if it matches, determining it as the final confirmation result; extracting the parking space number and vehicle information from the final confirmation result, and updating the occupancy status in the parking lot database to the latest record; using database query technology to obtain the updated available parking space information; converting the available parking space information into push data through a mobile application interface, and transmitting the push data to the user-side mobile application; judging whether the push information is successful according to the user-side feedback, and if not, re-obtaining the available parking space information and pushing it.

10. The method according to claim 1, wherein the method further includes: the server extracts the vehicle parking timestamp and license plate number from the occupancy status database, calculates the parking duration through the timestamp to obtain the billing basic data; processing the billing basic data through a preset billing rule to generate billing data including the parking duration and determining the deduction amount; obtaining the user account binding information from the payment system, transmitting the billing data to the corresponding account through an interface, and judging whether the account balance is sufficient; if the account balance is sufficient, performing a deduction operation through the payment system interface to complete the automatic deduction process; if the account balance is insufficient, generating a prompt message, transmitting it to the device associated with the user account through an interface, and obtaining the user's payment supplement status; updating the account balance according to the payment supplement status, and performing a deduction operation again through the payment system interface to obtain a record of successful deduction; for the record of successful deduction, updating the occupancy status in the database, judging whether the vehicle has left the parking space, and determining the end of the business process.

Images