A method for managing new energy parking spaces and oil vehicle occupying spaces based on AI and flatbed parking lock

Abstract

This application discloses a method for managing the occupancy of new energy vehicle parking spaces by gasoline vehicles based on AI and a flat-panel parking lock, including the following steps: S1: vehicle entry and initial locking; S2: visual recognition of charging behavior; S3: multimodal state determination; S4: anomaly analysis and intelligent matching; S5: unlocking control. Belonging to the field of intelligent transportation technology, this structure can solve the problem of cross-pile control by visually recognizing the connection status of the charging gun, and achieve automatic correction when anomalies are identified through vehicle feature vector matching.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation and charging pile management technology. Background Technology

[0002] With the increasing popularity of new energy vehicles, the number of public charging stations is growing daily. However, in actual operation, there are many pain points in the management of charging spaces: 1. The serious problem of gasoline vehicles occupying charging spots and occupying them for extended periods after a full charge results in low utilization of charging resources.

[0003] Existing solutions typically involve installing smart locks in charging parking spaces, using geomagnetic or infrared sensors to detect vehicle entry.

[0004] However, this type of sensor cannot distinguish whether the vehicle has actually been charging, which can easily lead to false locking or missed locking.

[0005] 2. The existing charging pile ground lock control logic is mostly a "one-pile-per-position" binding mode, that is, the ground lock only receives the signal of the corresponding numbered charging pile.

[0006] In real-world scenarios, users often disconnect the charging cable from an adjacent charging station to charge their devices because the original charging station's cable is too short or the original charging head is damaged.

[0007] This cross-charging behavior can cause the system's judgment logic to fail. For example, if a vehicle is parked in parking space A and uses the charging cable of charging station B, the parking lock in parking space A cannot receive the charging signal from charging station B, resulting in the vehicle being mistakenly locked or unable to complete the transaction normally.

[0008] 3. Existing license plate recognition technology is affected by factors such as ambient light and dirt, resulting in a certain failure rate.

[0009] Existing processing methods typically rely on manual intervention, with control room personnel retrieving monitoring data, manually verifying information, and manually releasing it. This approach is inefficient, has high labor costs, and provides a poor user experience.

[0010] Therefore, there is an urgent need to develop a method based on AI and tablet parking locks to manage the parking spaces occupied by fuel vehicles in new energy vehicle parking spaces to solve the problems in the existing technology. Summary of the Invention

[0011] The purpose of this invention is to provide a method for managing the occupation of new energy vehicle parking spaces by fuel vehicles based on AI and a flatbed parking lock. This method can solve the problem of cross-pile control by visually recognizing the connection status of the parking gun and achieve automatic correction when abnormalities are identified by vehicle feature vector matching. The method is simple in structure and easy to use, thus solving the problems mentioned in the background art.

[0012] To achieve the above objectives, the present invention provides the following technical solution: A method for managing the occupation of new energy vehicle parking spaces by gasoline vehicles based on AI and a tablet parking lock includes the following steps: S1: Vehicle entry and initial locking: The vision acquisition device collects real-time images of the parking space, identifies the vehicle and license plate area in the parking space, and controls the smart lock to lock the vehicle. S2: Visual recognition of charging behavior: Using a deep learning model to identify the charging gun insertion status, track the extension path of the charging cable and identify the charging pile number, thereby determining the target charging pile ID; compare the parking space number with the target charging pile ID, and if the two are inconsistent, generate a cross-pile charging association event and switch the current vehicle's binding relationship to the target charging pile. S3: Multimodal state determination: Integrates visual recognition data and charging pile IoT data to determine the current state of the vehicle; S4: Anomaly Analysis and Intelligent Matching: When license plate recognition fails or the recognition result is abnormal, a vehicle feature fingerprint is constructed, and the similarity distance between the vehicle feature fingerprint and the pre-stored user vehicle feature database is calculated; based on the similarity matching result, an identity confirmation request is sent to the user terminal, and the license plate information is corrected based on the confirmation information returned by the user terminal. S5: Unlock Control: Controls the lock to perform unlocking or locking actions based on the status determination result.

[0013] By adopting the above technical solution, visual recognition technology not only completes the basic vehicle entry locking, but more importantly, it can identify cross-charging behavior and automatically switch the binding relationship. At the same time, it has the ability to intelligently match and correct when the license plate recognition is abnormal, ensuring a high degree of consistency and accuracy between the ground lock control logic and the actual charging status.

[0014] As a further aspect of the present invention: the vehicle entry and initial locking steps in S1 specifically include: S11: The monitoring image is processed using a target detection algorithm. When a vehicle is detected to have completely entered the parking space and is stationary, the license plate number is extracted as the vehicle identification number. S12: Send a lift command to the smart lock to raise the lock arm and lock the vehicle.

[0015] By adopting the above technical solution and using target detection algorithms to determine the vehicle status, the locking action is only performed when the vehicle is completely stationary, thus realizing automated control of vehicle identification and parking space occupancy.

[0016] As a further aspect of the present invention: the extension path of the tracking charging cable in S2 specifically includes: S21: Use a semantic segmentation network to extract cable-like pixels in the image and generate a binarized mask; S22: Extract the skeleton from the binary mask to obtain the centerline pixel path of the cable; S23: Detect the straight segment of the centerline and calculate the main extension direction of the cable; S24: Search the charging pile area along the main extension direction and identify the number character of the charging pile area.

[0017] By adopting the above technical solution, through image segmentation and skeleton extraction technology, the system can accurately identify the extension path of the charging cable and the connecting pile, thereby determining the actual charging pile ID and providing core data support for cross-pile charging determination.

[0018] As a further aspect of the present invention: in the multimodal state determination step of S3, the determined vehicle state specifically includes: A legal charging status means that the charging gun is visually detected to be inserted, the output power of the target charging station is greater than zero, and the payment order is in the open state. An illegal occupancy status refers to either visually detecting that the charging gun is not inserted, or having the charging gun inserted but the charging pile's output power remains at zero for a duration exceeding a preset first threshold, satisfying either one of these conditions. The "fully occupied" status indicates that the charging order has been completed and the vehicle has remained in the parking space for longer than the preset free buffer time.

[0019] By adopting the above technical solution and integrating visual perception data with charging pile IoT data, the system can accurately distinguish between three states: legal charging, illegal occupation, and fully charged occupation, providing a reliable decision-making basis for the subsequent control of the ground lock.

[0020] As a further aspect of the present invention: the construction of vehicle feature fingerprint in S4 specifically includes extracting vehicle model, color, decal location, interior decorations and year features; Specifically, calculating the similarity distance in S4 includes: calculating the geometric distance between the current vehicle feature vector and the registered vehicle feature vectors in the database.

[0021] By adopting the above technical solution, and by extracting multi-dimensional features of vehicles and calculating geometric distances, the difficult-to-compare identity recognition problem is transformed into a computable feature vector matching problem, laying the foundation for automatic identity verification in abnormal situations.

[0022] As a further aspect of the present invention: the step S4 of sending an identity verification request to the user terminal based on the similarity matching result specifically includes: S41: When the calculated geometric distance is less than the preset distance threshold, the match is considered successful; S42: Send an identity verification request to the user terminal with the highest matching degree; S43: Receive confirmation information returned by the user terminal and establish a temporary charging whitelist.

[0023] By adopting the above technical solution and setting thresholds and terminal interaction mechanisms, the system can automatically complete identity verification and whitelist establishment under human-machine collaboration, replacing the inefficient mode of traditional manual intervention.

[0024] As a further aspect of the present invention: the unlocking control step in S5 specifically includes: S51: For legitimate charging status, when the charging order settlement is detected and the vehicle's reversing lights are turned on and the vehicle begins to move, a lowering command is sent to the ground lock. S52: In case of illegal occupation or full occupation, control the ground lock to remain locked until either a signal from the user terminal to pay the occupation fee or a central control authorization signal is received.

[0025] By adopting the above technical solutions, legitimate users can leave smoothly, while physical locking is implemented to restrain illegal occupation of charging spaces, effectively maintaining the operational order of charging spaces.

[0026] As a further aspect of the present invention: in the vehicle entry and initial locking step of S1, before controlling the smart lock to lock the vehicle, the following steps are also included: A1: Detect the presence of foreign objects under the floor lock arm using a pressure sensor; A2: If an abnormal pressure value is detected, the locking action will be paused and an alarm will be issued.

[0027] By adopting the above technical solution, foreign object detection is performed before the locking action is executed, which effectively prevents the mechanical structure of the ground lock from damaging the vehicle or the equipment from malfunctioning due to foreign objects blocking it, thus improving the security of the system.

[0028] This invention also discloses a system for license plate anomaly analysis and matching based on visual AI, comprising: The initial locking module is used to perform the vehicle entry and initial locking steps described above; The behavior recognition module is used to perform the above-mentioned visual recognition steps for charging behavior; The state determination module is used to execute the multimodal state determination steps described above. The anomaly matching module is used to perform the anomaly analysis and intelligent matching steps described above. The unlock control module is used to perform the unlock control steps described above.

[0029] By adopting the above technical solutions and through the collaborative work of various functional modules, stable operation of the entire process from data acquisition and visual recognition to logic control is ensured.

[0030] The present invention also discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0031] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention identifies the actual charging pile in use by visually tracking the cable path and automatically switches the binding relationship, avoiding vehicle mis-locking or billing errors caused by misaligned charging guns and cables. It significantly improves the intelligence level of the system and the user experience, and solves the management blind spot problem caused by cross-charging. 2. This invention achieves automatic identity correction through vehicle feature fingerprint matching and interaction with user terminals, realizing unattended closed-loop management in abnormal situations, significantly reducing operating costs and improving processing efficiency, and solving the problem of manual intervention required for license plate recognition anomalies.

[0032] Other features and advantages of the present invention will be disclosed in detail in the following detailed description and accompanying drawings. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of the overall structure of a method for managing the occupation of new energy vehicle parking spaces by fuel vehicles based on AI and a flat-panel parking lock, according to an embodiment of the present invention. Detailed Implementation

[0034] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] Example 1 This invention provides a method for managing the occupation of new energy vehicle parking spaces by gasoline vehicles based on AI and a tablet parking lock. (See also...) Figure 1 As shown, it includes the following steps: Step S1: Vehicle entry and initial locking During the vehicle entry phase, the system executes high-precision perception and safety locking logic.

[0036] First, the visual acquisition device acquires parking space monitoring images in real time at a rate of 25 frames per second. The image data is input into a pre-trained target detection model, such as YOLOv5 or YOLOv8. The model extracts features from the images and outputs the coordinates of the vehicle bounding box and the corner points of the license plate.

[0037] The system performs geometric calculations on the vehicle bounding box and the polygonal region of the parking space line to calculate the overlap ratio (IoU) between the vehicle's bounding rectangle and the parking space region. When the overlap ratio of 10 consecutive frames is greater than 0.95 and the displacement of the center point of the vehicle bounding box is less than 2 pixels, it is determined that the vehicle has completely entered the parking space and is in a stationary state.

[0038] Secondly, the system performs perspective transformation correction on the license plate area to remove tilt distortion, and inputs it into the CRNN convolutional recurrent neural network recognition engine to extract the license plate number string as the vehicle identification.

[0039] Before sending the lift command to the smart lock, the system activates a security protection mechanism.

[0040] The integrated thin-film pressure sensor inside the ground lock monitors the pressure value under the stop arm in real time. The microcontroller reads the pressure analog signal through an analog-to-digital converter and converts it into a digital pressure value. The system compares this value with a preset safety threshold. If abnormal pressure fluctuations or a sustained pressure greater than zero are detected, indicating the presence of foreign objects such as small animals, stones, or low-lying parts of the vehicle chassis, the system immediately locks the ground lock motor, suspends the lifting action, and triggers an audible and visual alarm to alert on-site personnel to clear the object. If the pressure value is normal, the system sends a positive drive signal to the ground lock motor through the relay control circuit, causing the ground lock arm to mechanically rise to its vertical limit position. The system then receives a feedback signal from the Hall sensor to complete the vehicle locking process.

[0041] Step S2: Visual Recognition of Charging Behavior To address the pain point of misaligned charging cables in charging pile management, this step utilizes computer vision technology to achieve refined behavior recognition.

[0042] The system utilizes semantic segmentation networks such as DeepLabV3+ to perform pixel-level classification of surveillance images. The network output includes segmentation masks for charging guns, charging cables, charging piles, and vehicle categories.

[0043] To identify the charging gun insertion status, the system calculates the overlap area between the charging gun mask area and the vehicle charging port mask area. When the overlap area ratio exceeds a preset crossover ratio threshold such as 0.6, it is determined that the charging gun has been inserted into the vehicle charging port.

[0044] For cable tracing, the system first performs binarization on the cable mask. Then, it applies a skeleton extraction algorithm, such as the Zhang-Suen thinning algorithm, to refine the cable area into a centerline skeleton with a width of one pixel. Based on this, the system performs a Hough transform on the centerline, detects the straight line components of the line segments, filters out short noise line segments, and retains the main extension path. The system starts from the charging gun insertion point and traces backward along the centerline until it extends to the charging pile body area.

[0045] The system uses OCR (Optical Character Recognition) technology to identify the serial numbers on the charging pile panel.

[0046] The specific process is as follows: detect the text region in the image, perform binarization and noise reduction on the text region, extract text features using a convolutional neural network, and output the charging pile number string, i.e., the target charging pile ID, through sequence decoding.

[0047] The system compares the target charging pile ID with the binding number of the parking space where the vehicle is currently located. If the two do not match, for example, the vehicle is parked in parking space A and bound to charging pile A, but the cable tracking result shows that it is connected to charging pile B, a cross-charging association event is triggered. The cloud server updates the session binding relationship of the vehicle, establishes a connection between its billing logic and charging pile B, and disconnects the logical connection with charging pile A.

[0048] Step S3: Multimodal state determination This step integrates visual perception data with charging pile IoT data to build a multi-dimensional state determination model.

[0049] The system first acquires the switching signal of the relay in the power supply circuit of the charging pile. The relay is connected in series in the main power supply circuit of the charging pile. When the relay contacts are closed and output a high-level signal, the system determines that the charging pile is in the state of external power supply; when the relay contacts are open or output a low-level signal, the system determines that the charging pile is in the state of power outage. The system adopts the hardware signal priority principle and uses the relay signal as the basis for determining the charging status.

[0050] The system reads the output power, voltage, current and charging order status data of the target charging pile in a polling manner through the edge gateway. The visual perception data and IoT data are aligned with a unified timestamp.

[0051] For legitimate charging status, the relay signal is the primary criterion. When a signal is detected on the relay, it is directly determined to be legitimate charging. At the same time, visual recognition data is used as a supplement. If the visual recognition shows that the charging gun has been inserted, the validity of charging is double-confirmed.

[0052] To address unauthorized parking space occupancy, the system incorporates a time counter. If the system visually detects that the charging gun is not inserted and this condition persists for more than a preset first time threshold (e.g., 10 minutes), an unauthorized parking space determination is triggered. Alternatively, if the charging gun is inserted, but the charging pile's output power remains at zero for more than a second time threshold (e.g., 15 minutes), and the order has not been initiated, this is also considered unauthorized parking space occupancy. This logic effectively distinguishes between situations where a gasoline vehicle is occupying a parking space and situations where charging is malfunctioning / not started.

[0053] For a fully charged parking space occupied state, the system monitors the charging order status. When the order status changes to completed, the system starts a dwell timer. If the timer value exceeds the preset free buffer time, such as 30 minutes, and the system visually detects that the vehicle is still in the parking space, it is determined to be a fully charged parking space occupied state, triggering the overtime billing logic.

[0054] Step S4: Anomaly Analysis and Intelligent Matching This step resolves the issue of license plate recognition failure preventing order creation, achieving automated closed-loop processing.

[0055] When the confidence score output by the license plate recognition engine is lower than a preset threshold, such as 0.7, or when there is no matching record in the registered user database, the system enters the anomaly analysis process.

[0056] The system calls the vehicle re-identification model, which is built on the ResNet50 backbone network. The input is a cropped vehicle image, and a high-dimensional feature vector is extracted, such as a 2048-dimensional vector. This feature vector covers fine-grained features such as vehicle model, color, decals, and interior, forming a vehicle feature fingerprint.

[0057] The system calculates the Euclidean distance or cosine similarity between the feature vector and the feature vectors of registered users' vehicles in the database. To improve retrieval efficiency, the system pre-builds an index structure for the feature vectors, such as the Faiss index.

[0058] When the calculated geometric distance is less than a preset distance threshold, the system determines it as a potential match. The system generates a confirmation message containing the captured image of the current vehicle and pushes it to the mobile terminal APP of the user with the highest match degree through the API interface. The message prompts: "The system has detected a vehicle that may be yours charging. Please confirm the license plate number." If the user clicks the confirmation button on the terminal, the terminal returns a confirmation signal carrying the user ID. After receiving the signal, the system automatically corrects the license plate information, adds the vehicle to the temporary charging whitelist, and establishes a charging order association. If the user clicks denial or there is no response after a timeout, the system automatically transfers the call to a human customer service representative for manual intervention.

[0059] Step S5: Unlock Control Based on the judgment results of the previous steps, this step executes a differentiated ground lock control strategy.

[0060] The system executes real-time control logic centered on relay signals. During the charging process, as long as a signal output is detected from the relay, the system determines that the vehicle is being effectively charged. At this time, regardless of whether the visual recognition detects any slight movement of the vehicle, the system will forcibly control the ground lock to remain in the lowered state to ensure that it is not locked and to ensure the continuity of the charging process. If the relay does not output a signal, the system will switch to the visual AI judgment logic to decide whether to lock based on whether the vehicle occupies the space.

[0061] The system establishes a dual monitoring mechanism for legitimate charging states.

[0062] On one hand, the system monitors the status of charging pile orders and triggers the departure permission flag when the order ends; on the other hand, it continuously analyzes the video stream, uses a target detection model to identify the taillight area of ​​the vehicle, determines whether the red reversing light is on through color space conversion, or determines whether the vehicle has moved backward through optical flow; when the departure permission flag is valid and the system visually confirms that the vehicle is ready to leave, the system sends a lowering command to the ground lock.

[0063] In case of illegal occupancy or full occupancy, the system sends a forced locking command to the ground lock. After receiving the command, the ground lock will keep the stop arm raised even if it detects that the vehicle is trying to move, to prevent escape.

[0064] The system synchronously sends payment notifications to the user terminal. Only when it receives a payment success signal forwarded by the cloud server, or a remote manual unlocking command sent by the central control room, will the system send a lowering command to the ground lock, allowing the vehicle to leave.

[0065] Example 2 In this embodiment, it specifically includes a visual perception module, a data processing module, and a control execution module.

[0066] 1. Hardware Environment A high-definition industrial camera and auxiliary lighting source are installed diagonally above the charging parking space to ensure that the field of view covers the parking space line, vehicle charging port, charging pile body and gun line area.

[0067] The smart parking lock has a built-in wireless communication module, enabling data interaction with a cloud server. The charging station has network connectivity, allowing it to report power output, order status, and station ID in real time.

[0068] 2. Visual Recognition Model Training To implement the above method, a convolutional neural network model needs to be pre-trained.

[0069] This embodiment uses an improved YOLOv8 model as the base network, and the specific training process is as follows: (1) Data collection and labeling Video streams of charging parking spaces were collected under different lighting conditions, such as daytime, nighttime, and backlighting, and under different weather conditions, such as sunny days, rainy days, and foggy days. Image frames were extracted from the videos to construct the original dataset.

[0070] Manually label the areas containing vehicles, license plates, charging guns, charging cables, and charging pile numbers in the image.

[0071] The labeling categories include vehicle bounding boxes, license plate character sequences, charging gun insertion status labels, cable segmentation masks, and pile number text.

[0072] (2) Data preprocessing Random rotation, cropping, color jittering, and Gaussian noise addition are performed on the original images to expand the dataset and enhance the model's generalization ability.

[0073] The resolution of the processed images was uniformly adjusted to 640 pixels by 640 pixels.

[0074] The dataset is divided into training, validation, and test sets in a ratio of 7:2:1.

[0075] In one feasible embodiment, the training set includes 50,000 images, the validation set includes 15,000 images, and the test set includes 5,000 images.

[0076] (3) Model structure configuration The backbone network adopts the CSPDarknet53 structure to extract multi-scale features of the image.

[0077] The neck network uses a PANet structure to fuse feature maps from different levels.

[0078] The detection head has three branches, which are used to detect small targets, such as license plate characters; medium targets, such as charging guns; and large targets, such as vehicles and cables.

[0079] In the pile number recognition branch, a CRNN convolutional recurrent neural network structure is introduced to recognize text sequences.

[0080] (4) Model training and optimization The total number of training rounds is set to 300.

[0081] The initial learning rate was set to 0.01, and a cosine annealing strategy was used to dynamically adjust the learning rate in the later stages of training to fine-tune the model parameters to converge to the global optimum.

[0082] The optimizer used is AdamW, with a weight decay factor set to 0.0005 to prevent model overfitting.

[0083] The loss function consists of three parts.

[0084] Among them, the bounding box regression loss uses CIoU Loss to measure the degree of overlap between the predicted box and the ground truth box, the distance between the center points, and the aspect ratio.

[0085] Focal Loss is used for classification to address the imbalance between positive and negative samples.

[0086] The segmentation loss uses Dice Loss to optimize the segmentation accuracy of the cable profile.

[0087] During training, the mean accuracy (mAP) of the model on the validation set is calculated every 10 iterations.

[0088] When mAP stops improving for 5 consecutive rounds, the early stopping mechanism is triggered, and the current optimal model weight parameters are saved.

[0089] This includes cable path tracing and cross-pillar determination logic, comprising the following contents: When the model detects that the charging gun coincides with the vehicle's charging port, the cable tracking program is initiated.

[0090] First, a semantic segmentation network is used to extract cable-like pixels from the image and generate a binarized mask.

[0091] Secondly, the skeleton of the binarized mask is extracted to obtain the centerline pixel path of the cable.

[0092] Next, the main extension direction of the cable is calculated by using the Hough transform to detect the straight segment of the center line.

[0093] Finally, the system searches along the main extension direction for the presence of charging pile areas. If found, it further identifies the number characters of the charging pile area and queries the database to obtain the real-time status of the charging pile corresponding to that number. If the charging pile is outputting power, it determines that cross-pile charging has occurred. The system automatically binds the current vehicle's billing account to the billing system of the cross-pile charging pile and unbinds the original charging pile associated with the parking space.

[0094] This includes vehicle feature fingerprint matching logic, which includes the following: When the license plate recognition module returns a confidence level of less than 0.6 or an empty recognition result, feature fingerprint matching is triggered.

[0095] The feature extraction network uses a ResNet50 model pre-trained on the VehicleID dataset.

[0096] The vehicle image is input into the network, and the output vector of the penultimate fully connected layer is extracted. The vector has a dimension of 2048 and is used as the feature vector of the vehicle.

[0097] Calculate the Euclidean distance between this feature vector and the feature vectors of registered vehicles in the cloud database.

[0098] The formula is as follows: Where p represents the feature vector of the current vehicle, q represents the feature vector of a registered vehicle in the database, and n represents the vector dimension.

[0099] A distance threshold of 0.3 is set. If the calculated Euclidean distance is less than this threshold, the two vehicles are considered to be similar in height.

[0100] The system retrieves the historical charging records of similar users. If the user has a history of frequent charging recently, the probability of a successful match is higher.

[0101] The system generates a push message containing a thumbnail of the current vehicle and sends it to the matching user's mobile terminal.

[0102] The message content is as follows: "The system has detected that your vehicle is suspected to be charging. Please confirm the license plate number." After the user clicks "confirm," the terminal returns a confirmation signal.

[0103] After receiving the signal, the system automatically completes the license plate information and records the correction log for subsequent iterative training of the model.

[0104] This includes the anti-accidental activation mechanism for the ground lock control logic, which includes the following: After receiving the lifting command, the ground lock first uses a pressure sensor to detect whether there are any foreign objects under the stop arm.

[0105] If an abnormal pressure value is detected, the lifting operation will be paused and an alarm will be issued to prevent damage to the vehicle chassis or tires.

[0106] After charging is complete, the system will not immediately send a descent command, but will instead open a preset buffer observation window with a duration of 5 minutes.

[0107] Within the buffer window, the vision system continuously monitors the vehicle's status. If the vehicle does not move and charging is not restarted, the system sends a notification to the user terminal, such as: "Charging complete, please drive away as soon as possible." If the system detects that the vehicle's reverse lights are on, it determines that the vehicle is preparing to leave and immediately sends a lowering command.

[0108] If the vehicle does not move after the buffer window is exceeded, the system determines that the space is occupied and initiates the overtime billing logic. The parking lock remains locked until the user completes the payment of the overtime fee.

[0109] This invention provides a method for managing the occupation of new energy vehicle parking spaces by fuel vehicles based on AI and a flatbed parking lock. It can solve the problem of cross-pile control by visually recognizing the connection status of the charging gun and achieve automatic correction when abnormalities are identified by vehicle feature vector matching.

[0110] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0111] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for managing the occupation of new energy vehicle parking spaces by gasoline vehicles based on AI and a tablet parking lock, characterized in that, Includes the following steps: S1: Vehicle entry and initial locking: The vision acquisition device collects real-time images of the parking space, identifies the vehicle and license plate area in the parking space, and controls the smart lock to lock the vehicle. S2: Visual recognition of charging behavior: Using a deep learning model to identify the charging gun insertion status, track the extension path of the charging cable and identify the charging pile number, thereby determining the target charging pile ID. Compare the parking space number with the target charging pile ID. If they do not match, generate a cross-charging association event and switch the current vehicle's binding relationship to the target charging pile. S3: Multimodal state determination: Integrates visual recognition data and charging pile IoT data to determine the current state of the vehicle; S4: Anomaly Analysis and Intelligent Matching: When license plate recognition fails or the recognition result is abnormal, a vehicle feature fingerprint is constructed, and the similarity distance between the vehicle feature fingerprint and the pre-stored user vehicle feature database is calculated; based on the similarity matching result, an identity confirmation request is sent to the user terminal, and the license plate information is corrected based on the confirmation information returned by the user terminal. S5: Unlock Control: Controls the lock to perform unlocking or locking actions based on the status determination result.

2. The method for managing the occupation of new energy vehicle parking spaces by gasoline vehicles based on AI and a tablet parking lock, as described in claim 1, is characterized in that... The vehicle entry and initial locking steps in S1 specifically include: S11: The monitoring image is processed using a target detection algorithm. When a vehicle is detected to have completely entered the parking space and is stationary, the license plate number is extracted as the vehicle identification number. S12: Send a lift command to the smart lock to raise the lock arm and lock the vehicle.

3. The method for managing the occupation of new energy vehicle parking spaces by gasoline vehicles based on AI and a tablet parking lock, as described in claim 1, is characterized in that... The extension path of the tracking charging cable in S2 specifically includes: S21: Use a semantic segmentation network to extract cable-like pixels in the image and generate a binarized mask; S22: Extract the skeleton from the binary mask to obtain the centerline pixel path of the cable; S23: Detect the straight segment of the centerline and calculate the main extension direction of the cable; S24: Search the charging pile area along the main extension direction and identify the number character of the charging pile area.

4. The method for managing the occupation of new energy vehicle parking spaces by gasoline vehicles based on AI and a tablet parking lock, as described in claim 1, is characterized in that... In the multimodal state determination step of S3, the determined vehicle state specifically includes: A legal charging status means that the charging gun is visually detected to be inserted, the output power of the target charging station is greater than zero, and the payment order is in the open state. An illegal occupancy status refers to either visually detecting that the charging gun is not inserted, or having the charging gun inserted but the charging pile's output power remains at zero for a duration exceeding a preset first threshold, satisfying either one of these conditions. The parking space is in a "fully occupied" state, meaning that the charging order has been completed and the vehicle has remained in the parking space for longer than the preset free buffer time.

5. The method for managing the occupation of new energy vehicle parking spaces by fuel vehicles based on AI and a tablet parking lock, as described in claim 1, is characterized in that... The construction of vehicle feature fingerprints in S4 specifically includes extracting features such as vehicle model, color, decal location, interior decorations, and model year. Specifically, calculating the similarity distance in S4 includes: calculating the geometric distance between the current vehicle feature vector and the registered vehicle feature vectors in the database.

6. The method for managing the occupation of new energy vehicle parking spaces by fuel vehicles based on AI and a tablet parking lock, as described in claim 5, is characterized in that... The step S4, which involves sending an identity verification request to the user terminal based on the similarity matching result, specifically includes: S41: When the calculated geometric distance is less than the preset distance threshold, the match is considered successful; S42: Send an identity verification request to the user terminal with the highest matching degree; S43: Receive confirmation information returned by the user terminal and establish a temporary charging whitelist.

7. The method for managing the occupation of new energy vehicle parking spaces by fuel vehicles based on AI and a tablet parking lock, as described in claim 4, is characterized in that... The unlocking control steps in S5 specifically include: S51: For legitimate charging status, when the charging order settlement is detected and the vehicle's reversing lights are turned on and the vehicle begins to move, a lowering command is sent to the ground lock. S52: In case of illegal occupation or full occupation, control the ground lock to remain locked until either a signal from the user terminal to pay the occupation fee or a central control authorization signal is received.

8. The method for managing the occupation of new energy vehicle parking spaces by gasoline vehicles based on AI and a tablet parking lock according to claim 1, characterized in that, In the vehicle entry and initial locking step of S1, before controlling the smart lock to lock the vehicle, the following steps are also included: A1: Detect the presence of foreign objects under the floor lock arm using a pressure sensor; A2: If an abnormal pressure value is detected, the locking action will be paused and an alarm will be issued.

9. A system for license plate anomaly analysis and matching based on visual AI, characterized in that, include: An initial locking module is used to perform the vehicle entry and initial locking steps of S1 as described in any one of claims 1-8; A behavior recognition module is used to perform the charging behavior visual recognition step of S2 as described in any one of claims 1-8; The state determination module is used to perform the multimodal state determination step of S3 as described in any one of claims 1-8; An anomaly matching module is used to perform the anomaly analysis and intelligent matching steps of S4 as described in any one of claims 1-8; An unlocking control module is used to perform the unlocking control step of S5 as described in any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements a method for managing the occupation of new energy vehicle parking spaces by oil vehicles based on AI and a flat-panel parking lock, as described in any one of claims 1-8.

Images