Intelligent parking management method and system based on video recognition

Abstract

The application discloses a kind of wisdom parking management method and system based on video identification, it is related to video identification technical field, the method comprises: the monitoring video stream of target parking area is collected, vehicle feature data, license plate feature data and parking space feature data in monitoring video stream are extracted, vehicle driving trajectory is determined according to vehicle feature data, license plate feature data and parking space feature data, vehicle parking behavior is determined based on driving trajectory, carries out parking behavior analysis to vehicle parking behavior, identifies abnormal parking behavior and parking space occupancy state, when identifying abnormal parking behavior or parking space occupancy exception, parking management instruction is generated, and corresponding parking management operation is executed according to parking management instruction.By the above-mentioned mode, the problem of low parking efficiency caused by inaccurate vehicle-parking space correlation judgment and abnormal behavior detection lag is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of video recognition, in particular to a smart parking management method and system based on video recognition. BACKGROUND

[0002] Current parking management systems based on video recognition mostly rely on single data sources (such as license plate recognition or vehicle detection) for static matching, which is difficult to effectively handle dynamic parking behaviors in complex scenarios. For example, traditional methods usually analyze vehicle positions or parking space states through fixed time interval video frames, but cannot accurately correlate the time sequence changes of vehicle trajectories and parking space states.

[0003] The above content is only used to assist in understanding the technical solutions of the present application and does not represent the acknowledgement of the above content as prior art. SUMMARY

[0004] The main purpose of the present application is to provide a smart parking management method and system based on video recognition, aiming to solve the technical problem of low parking efficiency caused by inaccurate vehicle-parking space correlation judgment and lagging abnormal behavior detection in the prior art.

[0005] To achieve the above purpose, the present application provides a smart parking management method based on video recognition, the method comprising:

[0006] Collecting monitoring video streams of a target parking area, extracting vehicle feature data, license plate feature data and parking space feature data in the monitoring video streams;

[0007] Determining a vehicle driving trajectory according to the vehicle feature data, the license plate feature data and the parking space feature data, and determining a vehicle parking behavior based on the driving trajectory;

[0008] Performing parking behavior analysis on the vehicle parking behavior, identifying abnormal parking behavior and parking space occupation state;

[0009] When abnormal parking behavior or parking space occupation anomaly is identified, generating a parking management instruction, and performing corresponding parking management operation according to the parking management instruction.

[0010] In an embodiment, the step of determining a vehicle driving trajectory according to the vehicle feature data, the license plate feature data and the parking space feature data, and determining a vehicle parking behavior based on the driving trajectory comprises:

[0011] Performing spatio-temporal alignment processing on the vehicle feature data, the license plate feature data and the parking space feature data to obtain a vehicle trajectory sequence with unified time stamp, and determining a vehicle driving trajectory according to the vehicle trajectory sequence;

[0012] identify spatial coordinates and state change information of the parking space based on the parking space feature data, and generate a parking space state sequence according to the spatial coordinates of the parking space and the state change information;

[0013] calculate an association weight between a vehicle trajectory point and a parking space state according to the vehicle trajectory sequence and the parking space state sequence;

[0014] weight and fuse the vehicle driving trajectory according to the association weight to generate a vehicle parking behavior reflecting a vehicle-parking space dynamic binding relationship.

[0015] In an embodiment, the step of calculating an association weight between a vehicle trajectory point and a parking space state according to the vehicle trajectory sequence and the parking space state sequence comprises:

[0016] mapping the position coordinates of each trajectory point in the vehicle trajectory sequence to a high-dimensional feature space to obtain a trajectory feature vector;

[0017] encoding the coordinates and state identifiers of each state point in the parking space state sequence to obtain a parking space feature vector;

[0018] calculating the correlation scores between each trajectory feature vector and all parking space feature vectors based on a multi-head attention mechanism;

[0019] normalizing the correlation scores to obtain the association weight between each vehicle trajectory point and each parking space state.

[0020] In an embodiment, the step of performing parking behavior analysis on the vehicle parking behavior to identify abnormal parking behavior and parking space occupancy state comprises:

[0021] extracting the stay duration of the vehicle trajectory in the vehicle parking behavior and the parking space occupancy time sequence feature;

[0022] when the stay duration is greater than a stay threshold, identifying as abnormal parking behavior;

[0023] when detecting that the parking space occupancy time sequence feature is continuous occupancy of the same parking space in different time periods, identifying as parking space occupancy anomaly.

[0024] In an embodiment, the step of identifying as parking space occupancy anomaly when detecting that the parking space occupancy time sequence feature is continuous occupancy of the same parking space in different time periods comprises:

[0025] obtaining an occupancy state sequence of a target parking space within a continuous time window, and calculating the frequency of state jumps based on the occupancy state sequence;

[0026] When the frequency of the state jump is greater than a normal fluctuation range, it is determined that there is a frequent occupation abnormality;

[0027] When the time sequence of the occupation state corresponding to the same vehicle conflicts, it is determined that there is a repeated occupation abnormality;

[0028] According to the frequent occupation abnormality and the repeated occupation abnormality, a parking space occupation abnormality event report is generated.

[0029] In an embodiment, when an abnormal parking behavior or a parking space occupation abnormality is identified, a parking management instruction is generated, and a corresponding parking management operation is performed according to the parking management instruction.

[0030] According to the type and severity of the abnormal parking behavior, a hierarchical alarm instruction is generated;

[0031] If a parking space occupation abnormality is identified, a parking space state reset instruction or a manual verification instruction is generated;

[0032] The hierarchical alarm instruction, the parking space state reset instruction, or the manual verification instruction is encapsulated as a parking management instruction;

[0033] According to the parking management instruction, a corresponding parking management operation is performed.

[0034] In an embodiment, the step of collecting a monitoring video stream of a target parking area, extracting vehicle feature data, license plate feature data, and parking space feature data in the monitoring video stream includes:

[0035] The collected original monitoring video stream is subjected to frame synchronization processing, and the frame-synchronized video stream is subjected to image denoising preprocessing to obtain a preprocessed video stream;

[0036] From the video frames of the preprocessed video stream, a vehicle region, a license plate region, and a parking space region are respectively located and extracted;

[0037] Visual features of the vehicle region are extracted as vehicle feature data, character information of the license plate region is identified and encoded as license plate feature data, and pixel change features of the parking space region are analyzed as parking space feature data.

[0038] In an embodiment, the method further includes:

[0039] A parking space query request sent by a user through a mobile terminal is received, and a time interval in the parking space query request is parsed;

[0040] Based on a historical vehicle parking behavior model, a parking position of a target vehicle in the time interval is predicted to obtain a parking position prediction result;

[0041] The parking position prediction result is fused with a real-time parking space state to generate a parking space availability suggestion and returned to a user terminal.

[0042] In an embodiment, the method further comprises:

[0043] Monitoring a traffic density of each area in the parking lot, and when the traffic density is greater than a preset density threshold, predicting a traffic distribution of a future period based on a vehicle parking behavior model to obtain a traffic distribution prediction result;

[0044] When the traffic distribution prediction result indicates that congestion will occur in a certain area, issuing a flow guide information to a driver through a variable message sign or a mobile terminal.

[0045] In addition, to achieve the above object, the present application further provides a smart parking management system based on video recognition, which comprises:

[0046] A data acquisition module is configured to acquire a monitoring video stream of a target parking area, and extract vehicle feature data, license plate feature data and parking space feature data in the monitoring video stream;

[0047] A behavior extraction module is configured to determine a vehicle driving track according to the vehicle feature data, the license plate feature data and the parking space feature data, and determine a vehicle parking behavior based on the driving track;

[0048] A behavior recognition module is configured to perform parking behavior analysis on the vehicle parking behavior, and identify abnormal parking behavior and parking space occupation state;

[0049] A parking management module is configured to generate a parking management instruction when abnormal parking behavior or parking space occupation anomaly is identified, and perform a corresponding parking management operation according to the parking management instruction.

[0050] In addition, to achieve the above object, the present application further provides a smart parking management device based on video recognition, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the computer program is configured to implement the steps of the above smart parking management method based on video recognition.

[0051] In addition, to achieve the above object, the present application further provides a storage medium, which is a computer readable storage medium, and the storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the above smart parking management method based on video recognition.

[0052] In addition, to achieve the above object, the application further provides a computer program product, which comprises a computer program, and the computer program realizes the steps of the intelligent parking management method based on video recognition as above when executed by a processor.

[0053] The application provides an intelligent parking management method based on video recognition, which comprises the following steps: collecting a monitoring video stream of a target parking area, extracting vehicle feature data, license plate feature data and parking space feature data in the monitoring video stream, determining a vehicle driving track according to the vehicle feature data, the license plate feature data and the parking space feature data, determining a vehicle parking behavior based on the driving track, performing parking behavior analysis on the vehicle parking behavior, identifying an abnormal parking behavior and a parking space occupation state, generating a parking management instruction when the abnormal parking behavior or the parking space occupation abnormality is identified, and performing a corresponding parking management operation according to the parking management instruction. In this way, the problem of low parking efficiency caused by inaccurate vehicle-space correlation judgment and lagging abnormal behavior detection is solved. BRIEF DESCRIPTION OF DRAWINGS

[0054] The accompanying drawings, which are incorporated into and form part of the specification, illustrate embodiments consistent with the present application and, together with the specification, serve to explain the principles of the application.

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0056] Figure 1 Flowchart of the intelligent parking management method based on video recognition of the application embodiment one;

[0057] Figure 2 Parking space flow guide schematic diagram of the intelligent parking management method based on video recognition of the application one embodiment;

[0058] Figure 3 Module structure schematic diagram of the intelligent parking management system based on video recognition of the application embodiment;

[0059] Figure 4 Device structure schematic diagram of the hardware running environment involved in the intelligent parking management method based on video recognition in the application embodiment.

[0060] The implementation of the object of the application, functional characteristics and advantages will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0061] It should be understood that the specific embodiments described herein are merely intended to explain the technical solutions of the present application, and are not intended to limit the present application.

[0062] In order to better understand the technical solutions of the present application, the following will be described in detail in combination with the drawings of the specification and specific embodiments.

[0063] The main solution of the embodiments of the present application is: collecting a monitoring video stream of a target parking area, extracting vehicle feature data, license plate feature data and parking space feature data in the monitoring video stream;

[0064] According to the vehicle feature data, the license plate feature data and the parking space feature data, the vehicle driving track is determined, and the vehicle parking behavior is determined based on the driving track;

[0065] The parking behavior analysis is performed on the vehicle parking behavior, and the abnormal parking behavior and the parking space occupation state are identified;

[0066] When the abnormal parking behavior or the parking space occupation abnormality is identified, a parking management instruction is generated, and a corresponding parking management operation is performed according to the parking management instruction.

[0067] At present, the current parking management system based on video recognition mostly relies on single data source (such as only license plate recognition or vehicle detection) for static matching, which is difficult to effectively handle dynamic parking behavior in complex scenes. For example, the traditional method usually analyzes the vehicle position or the parking space state through the video frames with fixed time interval, but cannot accurately associate the time sequence change of the vehicle track and the parking space state.

[0068] The present application provides a solution, which collects a monitoring video stream of a target parking area, extracts vehicle feature data, license plate feature data and parking space feature data in the monitoring video stream, determines a vehicle driving track according to the vehicle feature data, the license plate feature data and the parking space feature data, determines a vehicle parking behavior based on the driving track, performs parking behavior analysis on the vehicle parking behavior, identifies abnormal parking behavior and parking space occupation state, generates a parking management instruction when the abnormal parking behavior or the parking space occupation abnormality is identified, and performs a corresponding parking management operation according to the parking management instruction. Through the above-mentioned manner, the problem of low parking efficiency caused by inaccurate vehicle-parking space correlation judgment and lagging abnormal behavior detection is solved.

[0069] It should be noted that the execution subject of the embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, or an electronic device capable of realizing the above functions, a video recognition-based intelligent parking management device, etc. The embodiment does not make specific limitations on this. The following takes the video recognition-based intelligent parking management device as an example to describe the embodiment and the following embodiments.

[0070] All actions of obtaining signals, information or data in this application are carried out in accordance with the corresponding data protection regulations and policies of the country where the device is located, and with the authorization given by the corresponding device owner.

[0071] The embodiment of the present application provides a video recognition-based intelligent parking management method, which refers to Figure 1 Figure 1 The flowchart of the first embodiment of the video recognition-based intelligent parking management method of the present application.

[0072] In the embodiment, the video recognition-based intelligent parking management method includes steps S10-S40:

[0073] Step S10, collect the monitoring video stream of the target parking area, and extract the vehicle feature data, license plate feature data and parking space feature data in the monitoring video stream.

[0074] It should be noted that the monitoring video stream refers to real-time video data continuously captured and transmitted by one or more cameras deployed at key points in the parking lot, such as entrances, lanes, above parking spaces, etc. Vehicle feature data refers to visual information used to uniquely identify and track a vehicle, not limited to license plates, but also including vehicle color, model, brand, sub-brand, year of manufacture, etc. appearance attributes. License plate feature data refers to the license plate number and license plate type, such as blue plate, green plate, yellow plate, etc. structured text information positioned and identified from the vehicle area. Parking space feature data refers to information describing the state and location of a parking space itself, including but not limited to the spatial coordinates of the parking space, i.e. the position in the video picture or world coordinate system, the bounding box, and the state indicating whether it is currently occupied, including idle, occupied, etc.

[0075] It can be understood that first, the original video stream is preprocessed, such as frame synchronization and image denoising. Then, a pre-trained deep learning model such as YOLO, SSD, Faster R-CNN, etc. target detection model is used to analyze each frame of image, automatically locate and segment the three key regions of "vehicle", "license plate" and "parking space". Then, for the identified vehicle area, the vehicle feature data is obtained, for the identified license plate area, the license plate feature data is obtained, and for the identified parking space area, the parking space feature data is obtained.​

[0076] In one feasible implementation, the step of acquiring surveillance video streams of the target parking area and extracting vehicle feature data, license plate feature data, and parking space feature data from the surveillance video streams includes:

[0077] The acquired raw surveillance video stream is subjected to frame synchronization processing, and the synchronized video stream is subjected to image denoising preprocessing to obtain a preprocessed video stream.

[0078] The vehicle area, license plate area, and parking space area are located and extracted from the video frames of the preprocessed video stream, respectively.

[0079] Visual features of the vehicle area are extracted as vehicle feature data, character information of the license plate area is identified and encoded as license plate feature data, and pixel change features of the parking space area are analyzed as parking space feature data.

[0080] In a specific implementation, the steps of collecting surveillance video streams of the target parking area and extracting vehicle feature data, license plate feature data, and parking space feature data from the surveillance video streams may include video stream preprocessing, key area localization and interception, and multimodal feature extraction.

[0081] Video stream preprocessing prepares high-quality image data for subsequent accurate identification. The first step is frame synchronization. When using multiple cameras for monitoring, it's crucial to ensure that frames captured by different cameras at the same time can be correlated. This is typically achieved through hardware triggering or software-level GPS timing or Network Time Protocol (NTP), assigning a uniform timestamp `t` to each frame to form a time-aligned video stream sequence. Surveillance videos often introduce noise due to variations in lighting and sensor noise. Therefore, Gaussian filtering is used for noise reduction. By weighted averaging of pixel neighborhoods, the image is smoothed, effectively suppressing Gaussian noise. The Gaussian filtering formula is:

[0082]

[0083] in, The original image is in position pixel values, It is the Gaussian kernel function at the offset The output weight is defined as follows: , It's the standard deviation, which controls the smoothness. It is the size of the kernel function. This operation essentially replaces the value of the center pixel with the weighted average of the surrounding pixels. The weights are determined by a Gaussian distribution, and the closer the pixel is to the center, the greater its contribution.

[0084] Key region localization and cropping are then performed, and a deep learning object detection model is used to automatically find the region of interest in the image. At this time, advanced object detection algorithms such as YOLO or Faster R-CNN can be used. The core task of these models is to output a bounding box for each detected object. A bounding box is usually represented by the coordinates of its center , width and height , so the bounding box can be represented as . The model performs inference on the video frame , outputs a set of bounding boxes, and then crops the corresponding region image from the original image according to these coordinates.

[0085] Then multi-modal feature extraction is performed, and a convolutional neural network (CNN) is used as a feature extractor to input a vehicle region image into a pre-trained CNN. Before the last fully connected layer, a high-dimensional feature vector is obtained. This feature vector contains deep visual information of the vehicle, including but not limited to color, vehicle type, brand, etc. The obtained feature vector is the vehicle feature data. When extracting license plate feature data, the license plate region image can be binarized and corrected for tilt. Then a character detection model is used to segment individual characters, and finally a character classification model is used to recognize each character. The recognition result is finally encoded into a machine-readable text string. The occupancy status of the parking space can be regarded as a binary classification problem (free / occupied). An effective method is to calculate the feature change inside the parking space region. First, when the parking lot is initially empty, the background image of the parking space region is obtained. For the current frame of the parking space region image, the difference between it and the background model can be calculated. A simple and effective method is to use image difference combined with feature statistics. It can be represented as:

[0086]

[0087]

[0088] wherein is the difference image, is the current frame of the parking space region image, is the background image of the parking space region, is the mean of the difference image, is the standard deviation, is the weighting coefficient, is the parking space feature scalar.

[0089] Step S20, determining a vehicle driving trajectory according to the vehicle feature data, the license plate feature data and the parking space feature data, and determining a vehicle parking behavior based on the driving trajectory;

[0090] It should be noted that the vehicle driving track does not refer to the continuous GPS path of the vehicle, but rather the sequential record of the position of a vehicle over time in the discrete space-time coordinate system covered by the monitoring video, usually expressed in image coordinates or coordinates mapped onto the parking lot plan. Each track point contains a timestamp, vehicle position, and corresponding vehicle and license plate features. Vehicle parking behavior describes the complete interaction process between the vehicle and the parking space, including approaching the target parking space, parking in place, vehicle engine off and stationary (parking), vehicle starting, and driving away from the parking space, and a series of continuous and time-related action states.

[0091] It can be understood that first, the vehicle detection boxes captured by different cameras at different times are connected into continuous tracks through a data association algorithm: precise license plate feature data is used for strong identity binding, and when the license plate information is missing, the similarity calculation of the vehicle visual feature vector is used for auxiliary matching, and Kalman filtering and other motion models are used to predict and correct the vehicle position, thereby forming a smooth space-time track. Subsequently, behavior analysis is performed based on the track, the vehicle's stay time is identified by detecting long-time spatial aggregation points in the track, and the stay points are matched with the preset parking space geographic information, thereby accurately determining the type of parking behavior (such as driving in, parking, and driving away), the specific parking space associated, and the start and end time of the event, and finally completing the conversion from the bottom-level perception data to the high-level semantic behavior, such as "the vehicle with license plate number Beijing A·XXXX parked in the 102nd parking space in area B at 14:05".

[0092] In a feasible implementation, the step of determining the vehicle driving track based on the vehicle feature data, the license plate feature data, and the parking space feature data, and determining the vehicle parking behavior based on the driving track comprises:

[0093] spatiotemporal alignment processing of the vehicle feature data, the license plate feature data, and the parking space feature data to obtain a vehicle track sequence with unified timestamps, and determining the vehicle driving track based on the vehicle track sequence;

[0094] based on the parking space feature data, identifying the spatial coordinates and state change information of the parking space, and generating a parking space state sequence based on the spatial coordinates of the parking space and the state change information;

[0095] based on the vehicle track sequence and the parking space state sequence, calculating the association weight between the vehicle track point and the parking space state;

[0096] based on the association weight, weighted fusion of the vehicle driving track to generate a vehicle parking behavior reflecting the dynamic binding relationship between the vehicle and the parking space.

[0097] In a specific implementation, all feature data is aligned to the same timeline by using a uniform timestamp stamped in the pre-processing stage. For data not at the same time, interpolation processing can be needed to generate a fixed frequency (e.g., one frame per second) trajectory sequence. Through camera calibration technology, the image coordinates (pixel coordinates) of each camera are uniformly mapped to the same world coordinate system, such as the two-dimensional planar map coordinates of the parking lot. This is usually achieved by calculating a homography matrix. In this way, the position of the vehicle can be expressed in uniform map coordinates, regardless of which camera the vehicle appears in. Then a structured vehicle trajectory sequence is output. Each point in the sequence can be represented as: trajectory point = [timestamp , map coordinate , vehicle feature vector , license plate number ].

[0098] Then, through manual labeling or automatic learning, a polygonal region is defined for each parking space in the unified world coordinate system (map). Based on the extracted parking space feature data, a classifier is used to determine the state of each parking space in the t-th frame in real time, and a state sequence is output for each parking space. Then the proximity of the vehicle trajectory point position to the polygonal region of parking space K is determined. The simplest method is to determine whether the point is inside the polygon (with a weight of 1 or 0). A more refined method can calculate the distance from the point to the parking space boundary, and the closer the distance, the higher the weight, to determine the spatial correlation degree. In determining the time correlation degree, the synchronization of the duration of the vehicle trajectory and the change in the state of the parking space is analyzed. For example, if the time point at which a parking space changes from "free" to "occupied" is highly consistent with the time point at which a vehicle enters the parking space area and stops moving, then their correlation weight is very high. Then the comprehensive weight is calculated, and each "vehicle-parking space" pairing is scanned on its timeline. When the correlation weight exceeds a preset confidence threshold, a vehicle parking behavior reflecting the dynamic binding relationship between the vehicle and the parking space is generated.

[0099] In a feasible implementation, the step of calculating the correlation weight between the vehicle trajectory point and the parking space state according to the vehicle trajectory sequence and the parking space state sequence comprises:

[0100] mapping the position coordinates of each trajectory point in the vehicle trajectory sequence to a high-dimensional feature space to obtain a trajectory feature vector;

[0101] encoding the coordinates and state identifiers of each state point in the parking space state sequence to obtain a parking space feature vector;

[0102] calculate a correlation score between each of the trajectory feature vectors and all of the parking space feature vectors based on a multi-head attention mechanism;

[0103] normalize the correlation scores to obtain an association weight between each vehicle trajectory point and each parking space state.

[0104] In a specific implementation, a trajectory point not only has a position coordinate, but also can include a timestamp t, a vehicle speed, a vehicle heading direction, etc. We concatenate these original features into a base vector, and then map the base vector to a high-dimensional space, which can be 256-dimensional or 512-dimensional, through a fully connected neural network or a position encoding layer. This network learns how to encode information such as "the vehicle is located at (x, y) point at t time" into a unique feature vector that contains context information. The state information of the parking space, including the center point or region coordinates and the current state identifier, is also encoded into a high-dimensional feature vector using a fully connected network, obtaining a parking space feature vector. Then, based on a multi-head attention mechanism, a correlation score between each of the trajectory feature vectors and all of the parking space feature vectors is calculated, and three trainable parameter matrices are multiplied with each trajectory feature vector and each parking space feature vector to generate corresponding Q, K, and V vectors. This process is processed in parallel. Then, the correlation score is determined by similarity. After obtaining the correlation score, the correlation score can be converted into a probability distribution, so that the sum of the weights of all parking spaces for each trajectory point is 1. The Softmax function is used to normalize the scores of all parking spaces corresponding to each trajectory point, and for a given trajectory point i, the confidence probability associated with the parking space j can be determined.

[0105] Step S30, performing parking behavior analysis on the vehicle parking behavior to identify abnormal parking behavior and parking space occupancy state;

[0106] It should be noted that abnormal parking behavior refers to vehicle parking operations that do not conform to preset rules or normal patterns. For example, vehicle parking timeout, illegal parking, abnormal parking state, etc. The parking space occupancy state refers to the use of the parking space at the current time or in a specific time period. It usually includes: idle, legally occupied, abnormally occupied, etc. It can be understood that a natural paragraph is briefly described from the perspective of technical implementation.

[0107] In a specific implementation, the vehicle parking behavior is matched with a preset business rule library in real time to mark simple abnormalities such as overtime parking. Meanwhile, for complex scenarios such as channel occupation and irregular parking, a pre-trained abnormality detection model is called to perform deep analysis on the related video clips or image sequences, comprehensively judge the legality of the occupation state, and finally output structured abnormality event alarms and accurate parking space state maps.

[0108] In a feasible implementation, the step of performing parking behavior analysis on the vehicle parking behavior to identify abnormal parking behavior and parking space occupation state includes:

[0109] extracting a parking duration of a vehicle trajectory in the vehicle parking behavior and a parking space occupation time sequence feature;

[0110] when the parking duration is greater than a parking duration threshold, identifying the abnormal parking behavior;

[0111] when the parking space occupation time sequence feature is continuous occupation of a same parking space in different time periods, identifying a parking space occupation abnormality.

[0112] In a specific implementation, two key features are quantitatively extracted from the generated “vehicle-parking space” dynamic binding data, one is the parking duration of a vehicle trajectory, that is, the length of a time window in which the vehicle trajectory point sequence is in a stationary state, and the other is the parking space occupation time sequence feature, which describes a discrete sequence of changes in a single parking space occupation / idle state over time. The determination logic depends on preset thresholds and pattern matching: when the parking duration of a certain vehicle behavior exceeds the system-set parking duration threshold, an “overtime parking” type abnormality alarm is triggered; meanwhile, the system continuously analyzes the occupation time sequence of each parking space, and if it is found that a parking space is immediately occupied again after being released, and this “continuous occupation” pattern lasts for a long time, it may identify a parking space occupation abnormality such as “zombie car” or “long-term occupation by non-vehicle objects”.

[0113] In a feasible implementation, the step of identifying a parking space occupation abnormality when the parking space occupation time sequence feature is continuous occupation of a same parking space in different time periods includes:

[0114] obtaining an occupation state sequence of a target parking space in a continuous time window, and calculating a state jump frequency based on the occupation state sequence;

[0115] when the state jump frequency is greater than a normal fluctuation range, determining that there is a frequent occupation abnormality;

[0116] when there is a time sequence conflict in the occupation state corresponding to a same vehicle, determining that there is a repeated occupation abnormality;

[0117] generating a parking space occupation abnormal event report according to the frequent occupation abnormality and the repeated occupation abnormality.

[0118] In a specific implementation, first, a time window is slid, the occupation state sequence (such as [occupied, idle, occupied, occupied…]) of the target parking space is continuously sampled, and the state jump frequency is quantified by calculating the number of times of switching between the states of “occupied” and “idle”; if this frequency is significantly higher than the normal fluctuation range threshold set according to historical data or experience, the frequent occupation abnormality caused by artificial brushing or system misrecognition may be detected. At the same time, the system performs cross-vehicle logical consistency verification: by comparing the vehicle ID, trajectory and time stamp of the parking space occupation, if it is found that the same vehicle is assigned to two parking spaces that cannot be reached at the same time within the same time period, or the trajectory shows that it has left but the parking space state has not been released, it is determined that there is a time sequence conflict, indicating a repeated occupation abnormality, which is usually caused by ID switching error of the tracking algorithm or state update delay. Finally, the system integrates the detection results of the two types of abnormalities, the associated parking space ID, vehicle ID and time stamp and other information, and automatically generates a structured parking space occupation abnormal event report, providing accurate disposal basis for the background management.

[0119] Step S40, when an abnormal parking behavior or a parking space occupation abnormality is identified, a parking management instruction is generated, and a corresponding parking management operation is performed according to the parking management instruction.

[0120] It should be noted that the parking management instruction refers to a digital command or operation request automatically generated by the system according to the preset strategy after detecting the abnormal event. The parking management operation refers to a specific response measure automatically executed by the system or assisted by the management personnel according to the parking management instruction.

[0121] It can be understood that when an abnormality is identified, the system will immediately call the rule engine to match the type and severity of the abnormal event with the predefined disposal plan, so as to generate accurate parking management instructions. Subsequently, the execution module of the system will trigger the corresponding operation through a series of integrated interfaces according to the instruction content: for example, calling the message service API to send a reminder SMS to the vehicle owner, or creating an inspection task to be processed in the internal control system and notifying the relevant management personnel. This automatic response mechanism greatly improves the management efficiency and realizes the rapid intervention and closed-loop processing of the abnormal parking behavior.

[0122] In a feasible implementation, the step of generating a parking management instruction when an abnormal parking behavior or a parking space occupation abnormality is identified, and performing a corresponding parking management operation according to the parking management instruction comprises:

[0123] generating a hierarchical alarm instruction according to the type and severity of the abnormal parking behavior;

[0124] If the abnormal parking space occupation is identified, a parking space state resetting instruction or a manual verification instruction is generated;

[0125] The hierarchical alarm instruction, the parking space state resetting instruction or the manual verification instruction is packaged as a parking management instruction;

[0126] According to the parking management instruction, a corresponding parking management operation is performed.

[0127] In a specific implementation, first, the identified abnormality is quantitatively evaluated, for example, according to the length of time of overtime parking or the number of repeated occupation, the alarm is divided into different levels such as "reminder", "warning", "serious", and a structured hierarchical alarm instruction is generated according to the above, the content of which includes target license plate, abnormality level and preset disposal strategy. For parking space occupation abnormality, the system will generate two types of instructions: the parking space state resetting instruction is suitable for logical errors that can be automatically repaired, and the system directly calls the database update interface to forcibly reset the abnormal occupation state to idle; and the manual verification instruction is for complex situations that cannot be automatically determined, such as suspected zombie cars, and the system will create a work order containing detailed information and push it to the management platform. Subsequently, these instructions are uniformly packaged into a standard format of parking management instruction, and distributed and executed by the dispatch center, and finally the specific parking management operation is triggered through the integrated interface, such as pushing the hierarchical alarm message to the owner APP, executing the state resetting SQL in the database, or generating the task item to be manually verified in the work order system, thereby forming a complete automatic disposal closed loop.

[0128] In a feasible implementation, the method further comprises:

[0129] Receiving a parking space query request sent by a user through a mobile terminal, and parsing a time interval in the parking space query request;

[0130] Based on a historical vehicle parking behavior model, predicting a parking position of a target vehicle in the time interval to obtain a parking position prediction result;

[0131] Fusing the parking position prediction result with a real-time parking space state, generating a parking space availability suggestion and returning it to the user terminal.

[0132] In a specific implementation, after receiving a parking space query request of a user terminal through an API gateway and parsing a target time interval, a historical vehicle parking behavior model is first queried to predict a region or parking space where the vehicle is most likely to stop in the target time period, forming a parking location prediction result. Subsequently, the system fuses and compares the prediction result with real-time parking space state data from a parking lot Internet of Things sensing device head, and if the predicted parking space is currently or in the predicted time period occupied, the system will recommend a nearby available parking space, finally generating a parking space availability suggestion that comprehensively considers personal habits and real-time availability, and returning to the user terminal through a service interface.

[0133] In a feasible implementation, the method further comprises:

[0134] monitoring a vehicle flow density of each region in the parking lot, and when the vehicle flow density is greater than a preset density threshold, predicting a vehicle flow distribution in a future time period based on a vehicle parking behavior model to obtain a vehicle flow distribution prediction result;

[0135] when the vehicle flow distribution prediction result indicates that congestion will occur in a certain region, issuing a flow guide information to a driver through a variable information sign or a mobile terminal.

[0136] In a specific implementation, refer to Figure 2 , Figure 2 for a parking space flow guide schematic diagram. The vehicle flow density of each region is calculated in real time through a sensor network distributed in the parking lot, and when the density of a certain region exceeds a preset density threshold, the system triggers an early warning. Further, the vehicle parking behavior model is used to predict the vehicle flow distribution in a future time period to predict possible congestion regions. Once the vehicle flow distribution prediction result confirms that congestion will occur in a certain region, the system drives the variable information sign of the region to update and display the flow guide information through a control interface, and simultaneously pushes personalized route suggestions to the mobile terminal APP of the vehicle that has entered or is about to enter the parking lot through a message service platform, so as to realize active allocation and guidance of vehicle flow and improve overall traffic efficiency.

[0137] The embodiment provides a smart parking management method based on video recognition, acquires a monitoring video stream of a target parking area, extracts vehicle feature data, license plate feature data and parking space feature data in the monitoring video stream, determines a vehicle driving track according to the vehicle feature data, the license plate feature data and the parking space feature data, determines a vehicle parking behavior based on the driving track, performs parking behavior analysis on the vehicle parking behavior, identifies abnormal parking behavior and a parking space occupation state, generates a parking management instruction when the abnormal parking behavior or the parking space occupation anomaly is identified, and performs a corresponding parking management operation according to the parking management instruction. In the above manner, the problem of low parking efficiency caused by inaccurate vehicle-parking space correlation judgment and lagging abnormal behavior detection is solved.

[0138] It should be noted that the above examples are only for understanding the present application and do not constitute a limitation on the wisdom parking management method based on video recognition of the present application. More forms of simple transformation based on this technical concept are within the protection scope of the present application.

[0139] The present application also provides a wisdom parking management system based on video recognition, please refer to Figure 3 The wisdom parking management system based on video recognition comprises:

[0140] A data acquisition module 10 is configured to acquire a monitoring video stream of a target parking area, extract vehicle feature data, license plate feature data and parking space feature data in the monitoring video stream;

[0141] A behavior extraction module 20 is configured to determine a vehicle driving trajectory according to the vehicle feature data, the license plate feature data and the parking space feature data, and determine a vehicle parking behavior based on the driving trajectory;

[0142] A behavior recognition module 30 is configured to perform parking behavior analysis on the vehicle parking behavior, and identify abnormal parking behavior and parking space occupation state;

[0143] A parking management module 40 is configured to generate a parking management instruction when abnormal parking behavior or parking space occupation anomaly is identified, and perform corresponding parking management operation according to the parking management instruction.

[0144] In a feasible implementation, the behavior extraction module 20 is further configured to perform spatio-temporal alignment processing on the vehicle feature data, the license plate feature data and the parking space feature data, obtain a vehicle trajectory sequence with unified time stamp, and determine a vehicle driving trajectory according to the vehicle trajectory sequence;

[0145] The parking space feature data is used to identify spatial coordinates and state change information of the parking space, and a parking space state sequence is generated according to the spatial coordinates of the parking space and the state change information;

[0146] The vehicle trajectory sequence and the parking space state sequence are used to calculate a correlation weight between a vehicle trajectory point and a parking space state;

[0147] The vehicle driving trajectory is weighted and fused according to the correlation weight, and a vehicle parking behavior reflecting a vehicle-parking space dynamic binding relationship is generated.

[0148] In a feasible implementation, the behavior extraction module 20 is further configured to map position coordinates of each trajectory point in the vehicle trajectory sequence to a high-dimensional feature space, and obtain a trajectory feature vector;

[0149] Encode the coordinates and state identifiers of each state point in the parking space state sequence to obtain a parking space feature vector;

[0150] Calculate the correlation scores between each trajectory feature vector and all parking space feature vectors based on the multi-head attention mechanism;

[0151] Normalize the correlation scores to obtain the association weights between each vehicle trajectory point and each parking space state.

[0152] In a feasible implementation, the behavior recognition module 30 is further configured to extract the parking duration and parking space occupancy time sequence features of the vehicle trajectory in the vehicle parking behavior;

[0153] When the parking duration is greater than a parking threshold, the abnormal parking behavior is recognized;

[0154] When the parking space occupancy time sequence feature is detected as continuous occupancy of the same parking space in different time periods, the parking occupancy abnormality is recognized.

[0155] In a feasible implementation, the behavior recognition module 30 is further configured to obtain an occupancy state sequence of a target parking space within a continuous time window, and calculate the frequency of state jumps based on the occupancy state sequence;

[0156] When the frequency of state jumps is greater than a normal fluctuation range, it is determined that there is a frequent occupancy abnormality;

[0157] When the occupancy state corresponding to the same vehicle appears a time sequence conflict, it is determined that there is a repeated occupancy abnormality;

[0158] Generate a parking occupancy abnormality event report according to the frequent occupancy abnormality and the repeated occupancy abnormality.

[0159] In a feasible implementation, the parking management module 40 is further configured to generate a hierarchical alarm instruction according to the type and severity of the abnormal parking behavior;

[0160] If a parking space occupancy abnormality is recognized, a parking space state reset instruction or a manual verification instruction is generated;

[0161] Encapsulate the hierarchical alarm instruction, the parking space state reset instruction, or the manual verification instruction as a parking management instruction;

[0162] According to the parking management instruction, perform a corresponding parking management operation.

[0163] In a feasible implementation, the data acquisition module 10 is further configured to perform frame synchronization processing on the collected original monitoring video stream, and perform image denoising preprocessing on the frame-synchronized video stream to obtain a preprocessed video stream;

[0164] locating and cropping a vehicle region, a license plate region and a parking space region from the video frames of the pre-processed video stream respectively;

[0165] extracting visual features of the vehicle region as vehicle feature data, recognizing and encoding character information of the license plate region as license plate feature data, and analyzing pixel variation features of the parking space region as parking space feature data.

[0166] In an implementation, the parking suggestion module 50 is further configured to receive a parking space query request sent by a user through a mobile terminal, parse a time interval in the parking space query request;

[0167] predict a parking position of a target vehicle in the time interval based on a historical vehicle parking behavior model, to obtain a parking position prediction result;

[0168] fuse the parking position prediction result with a real-time parking space state, generate a parking space availability suggestion, and return the parking space availability suggestion to the user terminal.

[0169] In an implementation, the vehicle flow guiding module 60 is further configured to monitor a vehicle flow density of each region in a parking lot, when the vehicle flow density is greater than a preset density threshold, predict a vehicle flow distribution of a future time period based on a vehicle parking behavior model, to obtain a vehicle flow distribution prediction result;

[0170] when the vehicle flow distribution prediction result indicates that congestion will occur in a certain region, issue a flow guiding information to a driver through a variable message sign or a mobile terminal.

[0171] The intelligent parking management system based on video recognition provided in the application adopts the intelligent parking management method based on video recognition in the above embodiments, and can solve the technical problem of low parking efficiency caused by inaccurate vehicle-parking space correlation judgment and lagging abnormal behavior detection. Compared with the prior art, the intelligent parking management system based on video recognition provided in the application has the same beneficial effects as the intelligent parking management method based on video recognition provided in the above embodiments, and other technical features in the intelligent parking management system based on video recognition are the same as the features disclosed in the above embodiments, which will not be repeated here.

[0172] The application provides an intelligent parking management device based on video recognition, which comprises at least one processor and a memory in communication connection with the at least one processor. The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the intelligent parking management method based on video recognition in the above embodiments.

[0173] Reference will be made to the following Figure 4This document illustrates a structural schematic diagram of a video recognition-based smart parking management device suitable for implementing embodiments of this application. The video recognition-based smart parking management device in this application embodiment may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 4 The video recognition-based smart parking management device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0174] like Figure 4 As shown, the video recognition-based smart parking management device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in ROM (Read Only Memory) 1002 or a program loaded from storage device 1003 into RAM (Random Access Memory) 1004. RAM 1004 also stores various programs and data required for the operation of the video recognition-based smart parking management device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via bus 1005. Input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, LCDs (Liquid Crystal Displays), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the video recognition-based smart parking management device to exchange data wirelessly or via wired communication with other devices. Although the figure shows a video recognition-based smart parking management device with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0175] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer readable medium, the computer program containing program codes for executing the method shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from a network through a communication device, or installed from the storage device 1003, or installed from the ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiments disclosed in the present application are executed.

[0176] The intelligent parking management device based on video recognition provided by the present application adopts the intelligent parking management method based on video recognition in the above-mentioned embodiments, and can solve the technical problem of intelligent parking management based on video recognition. Compared with the prior art, the intelligent parking management device based on video recognition provided by the present application has the same beneficial effects as the intelligent parking management method based on video recognition provided by the above-mentioned embodiments, and other technical features in the intelligent parking management device based on video recognition are the same as the features disclosed in the previous embodiment method, which will not be repeated here.

[0177] It should be understood that various parts of the present application can be realized by hardware, software, firmware or a combination thereof. In the description of the above-mentioned embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0178] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within 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.

[0179] The present application provides a computer readable storage medium having stored thereon computer readable program instructions (i.e. computer program) for executing the intelligent parking management method based on video recognition in the above-mentioned embodiments.

[0180] The computer readable storage medium provided in the application may be, for example, a U disk, but is not limited to an electric, magnetic, optical, electromagnetic, infrared, or semiconductor system, system, or device, or any combination thereof. More specific examples of the computer readable storage medium may include, but are not limited to, an electric connection with one or more conductive wires, a portable computer disk, a hard disk, a RAM (Random Access Memory), a ROM (Read Only Memory), an EPROM (Erasable Programmable Read Only Memory or flash memory), an optical fiber, a CD-ROM (CD-Read Only Memory), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the embodiment, the computer readable storage medium may be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, system, or device. The program code contained on the computer readable storage medium can be transmitted by any suitable medium, including but not limited to an electric wire, an optical cable, an RF (Radio Frequency), and the like, or any suitable combination thereof.

[0181] The computer readable storage medium described above may be contained in the intelligent parking management device based on video recognition, or may exist independently without being assembled into the intelligent parking management device based on video recognition.

[0182] The computer readable storage medium described above carries one or more programs, which, when executed by the intelligent parking management device based on video recognition, cause the intelligent parking management device based on video recognition to: collect a monitoring video stream of a target parking area, extract vehicle feature data, license plate feature data, and parking space feature data in the monitoring video stream;

[0183] Determine a vehicle driving trajectory according to the vehicle feature data, the license plate feature data, and the parking space feature data, and determine a vehicle parking behavior based on the driving trajectory;

[0184] Perform parking behavior analysis on the vehicle parking behavior, and identify abnormal parking behavior and parking space occupation state;

[0185] When an abnormal parking behavior or parking space occupation anomaly is identified, generate a parking management instruction, and perform a corresponding parking management operation according to the parking management instruction.

[0186] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0187] The flow diagrams and the block diagrams in the drawings are illustrations of architectures, functionalities, and operations of possible implementations of systems, methods, and computer program products according to various embodiments of the present application. In this regard, each block in the flow diagrams or block diagrams can represent a module, a segment, or a portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently or the blocks may

[0188] The modules involved in the embodiments of the present application can be implemented in the form of software or in the form of hardware. In some cases, the name of the module does not constitute a limitation on the module itself.

[0189] The readable storage medium provided by the present application is a computer readable storage medium, which stores computer readable program instructions (i.e. computer programs) for executing the above-mentioned video recognition-based intelligent parking management method, and can solve the technical problem of video recognition-based intelligent parking management. Compared with the prior art, the computer readable storage medium provided by the present application has the same beneficial effects as the video recognition-based intelligent parking management method provided by the above-mentioned embodiments, which will not be described here.

[0190] The application further provides a computer program product comprising a computer program which, when executed by a processor, implements the steps of the video recognition-based intelligent parking management method as described above.

[0191] The computer program product provided by the application can solve the technical problem of video recognition-based intelligent parking management. Compared with the prior art, the beneficial effects of the computer program product provided by the application are the same as those of the video recognition-based intelligent parking management method provided by the above-mentioned embodiments, and are not described here.

[0192] The above is only some embodiments of the application, and does not limit the patent scope of the application. Any equivalent structural transformation made by using the content of the specification and drawings, or direct / indirect application in other related technical fields under the technical concept of the application is included in the patent protection scope of the application.

Claims

1. A smart parking management method based on video recognition, characterized in that, The video recognition-based smart parking management method includes: Collect surveillance video streams of the target parking area and extract vehicle feature data, license plate feature data, and parking space feature data from the surveillance video streams; The vehicle's driving trajectory is determined based on the vehicle feature data, the license plate feature data, and the parking space feature data; and the vehicle's parking behavior is determined based on the driving trajectory. The parking behavior of the vehicles is analyzed to identify abnormal parking behavior and parking space occupancy status; When abnormal parking behavior or abnormal parking space occupancy is detected, a parking management instruction is generated, and corresponding parking management operations are executed according to the parking management instruction; The step of determining the vehicle's driving trajectory based on the vehicle feature data, the license plate feature data, and the parking space feature data, and determining the vehicle's parking behavior based on the driving trajectory, includes: The vehicle feature data, license plate feature data, and parking space feature data are spatiotemporally aligned to obtain a vehicle trajectory sequence with a unified timestamp, and the vehicle driving trajectory is determined based on the vehicle trajectory sequence. Based on the parking space feature data, the spatial coordinates and state change information of the parking space are identified, and a parking space state sequence is generated according to the spatial coordinates and state change information of the parking space. Based on the vehicle trajectory sequence and the parking space status sequence, the association weight between the vehicle trajectory points and the parking space status is calculated. The vehicle's driving trajectory is weighted and fused according to the association weight to generate vehicle parking behavior that reflects the dynamic binding relationship between the vehicle and the parking space; The step of calculating the association weight between vehicle trajectory points and parking space status based on the vehicle trajectory sequence and the parking space status sequence includes: The position coordinates of each trajectory point in the vehicle trajectory sequence are mapped to a high-dimensional feature space to obtain the trajectory feature vector; The coordinates and status identifiers of each state point in the parking space state sequence are encoded to obtain the parking space feature vector; The correlation score between each trajectory feature vector and all parking space feature vectors is calculated based on a multi-head attention mechanism. The correlation scores are normalized to obtain the correlation weight between each vehicle trajectory point and the status of each parking space. Specifically, this includes: determining the spatial correlation degree based on the distance between the vehicle trajectory point and the boundary of the parking space, and determining the temporal correlation degree based on the duration of the vehicle trajectory point and the change in the status of the parking space. A comprehensive weighting is calculated based on the temporal correlation and the spatial correlation.

2. The method as described in claim 1, characterized in that, The steps of analyzing the parking behavior of the vehicles to identify abnormal parking behavior and parking space occupancy status include: Extract the dwell time of the vehicle trajectory and the time sequence features of parking space occupancy in the vehicle parking behavior; When the dwell time exceeds the dwell threshold, it is identified as abnormal parking behavior; When the parking space occupancy time sequence is detected to be the continuous occupancy of the same parking space at different time periods, it is identified as an abnormal parking space occupancy.

3. The method as described in claim 2, characterized in that, The step of identifying an abnormal parking space occupancy when the parking space occupancy time sequence characteristic is detected as continuous occupancy of the same parking space at different time periods includes: Obtain the occupancy status sequence of the target parking space within a continuous time window, and calculate the frequency of state transitions based on the occupancy status sequence; When the frequency of the state transitions is greater than the normal fluctuation range, it is determined that there is a frequent occupancy anomaly. When a timing conflict occurs in the occupancy status of the same vehicle, it is determined that there is a duplicate occupancy anomaly. A parking space occupancy anomaly report is generated based on the frequent and repeated occupancy anomalies.

4. The method as described in claim 1, characterized in that, The steps of generating a parking management instruction and executing corresponding parking management operations based on the parking management instruction when abnormal parking behavior or abnormal parking space occupancy is detected include: Generate tiered alarm commands based on the type and severity of abnormal parking behavior; If an abnormal parking space occupancy is detected, a parking space status reset command or a manual verification command will be generated. The hierarchical alarm instructions, parking space status reset instructions, or manual verification instructions are encapsulated into parking management instructions; Execute the corresponding parking management operation according to the parking management instruction.

5. The method as described in claim 1, characterized in that, The steps of collecting surveillance video streams of the target parking area and extracting vehicle feature data, license plate feature data, and parking space feature data from the surveillance video stream include: The acquired raw surveillance video stream is subjected to frame synchronization processing, and the synchronized video stream is subjected to image denoising preprocessing to obtain a preprocessed video stream. The vehicle area, license plate area, and parking space area are located and extracted from the video frames of the preprocessed video stream, respectively. Visual features of the vehicle area are extracted as vehicle feature data, character information of the license plate area is identified and encoded as license plate feature data, and pixel change features of the parking space area are analyzed as parking space feature data.

6. The method as described in claim 1, characterized in that, The method further includes: Receive parking space query requests sent by users via mobile terminals, and parse the time interval in the parking space query requests; Based on the historical vehicle parking behavior model, the parking location of the target vehicle within the time interval is predicted, and the parking location prediction result is obtained. The predicted parking location is fused with the real-time parking space status to generate a parking space availability suggestion, which is then returned to the user terminal.

7. The method as described in claim 1, characterized in that, The method further includes: Monitor the traffic flow density in each area of ​​the parking lot. When the traffic flow density is greater than a preset density threshold, predict the traffic flow distribution for future periods based on a vehicle parking behavior model to obtain the traffic flow distribution prediction result. When the traffic flow distribution prediction results indicate that congestion will occur in a certain area, traffic diversion information is released to drivers via variable message signs or mobile terminals.

8. A smart parking management system based on video recognition, characterized in that, The video recognition-based smart parking management system includes: The data acquisition module is used to acquire the surveillance video stream of the target parking area and extract vehicle feature data, license plate feature data and parking space feature data from the surveillance video stream; The behavior extraction module is used to determine the vehicle's driving trajectory based on the vehicle feature data, the license plate feature data, and the parking space feature data, and to determine the vehicle's parking behavior based on the driving trajectory. The behavior recognition module is used to analyze the parking behavior of the vehicle and identify abnormal parking behavior and parking space occupancy status. The parking management module is used to generate parking management instructions when abnormal parking behavior or abnormal parking space occupancy is detected, and to execute corresponding parking management operations according to the parking management instructions. The step of determining the vehicle's driving trajectory based on the vehicle feature data, the license plate feature data, and the parking space feature data, and determining the vehicle's parking behavior based on the driving trajectory, includes: The vehicle feature data, license plate feature data, and parking space feature data are spatiotemporally aligned to obtain a vehicle trajectory sequence with a unified timestamp, and the vehicle driving trajectory is determined based on the vehicle trajectory sequence. Based on the parking space feature data, the spatial coordinates and state change information of the parking space are identified, and a parking space state sequence is generated according to the spatial coordinates and state change information of the parking space. Based on the vehicle trajectory sequence and the parking space status sequence, the association weight between the vehicle trajectory points and the parking space status is calculated. The vehicle's driving trajectory is weighted and fused according to the association weight to generate vehicle parking behavior that reflects the dynamic binding relationship between the vehicle and the parking space; The step of calculating the association weight between vehicle trajectory points and parking space status based on the vehicle trajectory sequence and the parking space status sequence includes: The position coordinates of each trajectory point in the vehicle trajectory sequence are mapped to a high-dimensional feature space to obtain the trajectory feature vector; The coordinates and status identifiers of each state point in the parking space state sequence are encoded to obtain the parking space feature vector; The correlation score between each trajectory feature vector and all parking space feature vectors is calculated based on a multi-head attention mechanism. The correlation scores are normalized to obtain the correlation weight between each vehicle trajectory point and the status of each parking space. Specifically, this includes: determining the spatial correlation degree based on the distance between the vehicle trajectory point and the boundary of the parking space, and determining the temporal correlation degree based on the duration of the vehicle trajectory point and the change in the status of the parking space. A comprehensive weighting is calculated based on the temporal correlation and the spatial correlation.

Images