Automobile data recorder illegal parking detection method and device, electronic equipment and storage medium

Abstract

The invention discloses an illegal parking detection method and device for an automobile data recorder, electronic equipment and a storage medium, and is used for solving the problem that real-time warning or initiative reporting of illegal parking behaviors cannot be realized due to lack of an automatic illegal parking monitoring mechanism in related technologies. The method comprises the steps that when it is judged that a vehicle is in a static state, video frame data around the vehicle are continuously recorded through an automobile data recorder; preprocessing the video frame data, and extracting key image features from the preprocessed video frame data; inputting the key image features into an illegal parking detection model based on machine vision pre-training for illegal parking detection and identification; and when the vehicle is detected to have the illegal parking behavior, automatically triggering illegal parking warning, and generating illegal parking warning information. Therefore, in combination with an automobile data recorder and a machine vision algorithm, an automatic illegal parking detection means is provided, the illegal parking behavior of the vehicle can be monitored in real time, active warning is performed when the illegal parking behavior occurs, and illegal parking events are reduced.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a method, device, electronic equipment and storage medium for detecting illegal parking with a driving recorder. Background Art

[0002] Vehicle Violation Parking refers to the act of parking a vehicle at a location or time period where parking is not allowed. This behavior not only affects traffic order, but may also cause safety hazards. In order to effectively manage and detect vehicle violation parking, the current solutions mainly focus on urban fixed monitoring systems and some advanced driver assistance systems (ADAS).

[0003] Among them, fixed surveillance cameras are usually installed at key intersections and sections of the city. Through video surveillance software, image analysis is performed to identify whether the vehicle is parked in a prohibited parking area. However, such systems are often limited by fixed installation locations and limited monitoring ranges. In complex and changeable urban environments (such as occlusion, light changes, etc.), they are easily affected by weather, light conditions and background interference, resulting in high false alarm and missed alarm rates and low accuracy in illegal parking identification. At the same time, fixed surveillance cameras are expensive to deploy and difficult to achieve large-scale coverage. Especially in small streets or newly formed illegal parking hotspots, there are many monitoring blind spots.

[0004] As for ADAS, although some high-end models are equipped with sensor-based parking assistance functions, which can remind drivers of the legality of parking locations, these systems are mainly used to assist drivers to park correctly, and do not have the ability to monitor illegal parking after the vehicle is turned off, nor can they provide evidence of illegal parking to traffic management departments.

[0005] Furthermore, although there are cases of using dashcams to record traffic incidents in the market, most applications focus on determining accident responsibility. This is achieved by having users manually mark the incident and then upload the video clip. This approach lacks an automated illegal parking monitoring mechanism and cannot provide real-time warnings or proactively report illegal parking behaviors. Therefore, this method has little effect on illegal parking management. Summary of the invention

[0006] The present invention provides a driving recorder illegal parking detection method, device, electronic device and storage medium, which are used to solve or partially solve the technical problem that the related technology lacks an automated illegal parking monitoring mechanism and cannot achieve real-time warning or active reporting of illegal parking behavior.

[0007] The present invention provides a method for detecting illegal parking of a driving recorder, the method comprising:

[0008] When it is determined that the vehicle is stationary, continuously recording video frame data around the vehicle through a driving recorder;

[0009] Preprocessing the video frame data, and extracting key image features from the preprocessed video frame data;

[0010] Inputting the key image features into a parking violation detection model based on machine vision pre-training to perform parking violation detection and identification;

[0011] When illegal parking is detected for the vehicle, an illegal parking warning is automatically triggered and illegal parking warning information is generated.

[0012] Optionally, the key image features include a vehicle outline, license plate information, and the surrounding environment of the vehicle; the step of inputting the key image features into a parking violation detection model based on machine vision pre-training to perform parking violation detection and recognition includes:

[0013] Inputting the vehicle outline, the license plate information and the surrounding environment of the vehicle into a parking violation detection model based on machine vision pre-training to perform vehicle positioning and identification, and determine the current parking area of ​​the vehicle;

[0014] In combination with the surrounding environment of the vehicle, the parking type of the current parking area is identified, and based on the parking type and the preset illegal parking conditions, it is determined whether the vehicle has committed illegal parking.

[0015] Optionally, the identifying the parking type of the current parking area in combination with the surrounding environment of the vehicle, and judging whether the vehicle has committed illegal parking based on the parking type and a preset illegal parking condition, includes:

[0016] Identify a no-parking sign from the environment around the vehicle; the no-parking sign includes a no-parking sign for all day, a temporary parking sign, a no-parking sign during peak hours, and a sign for leaving the vehicle and illegally parking;

[0017] If a no-parking sign can be further identified from the vehicle surroundings, determining the parking type of the current parking area according to the no-parking sign;

[0018] If no parking prohibition sign can be further identified from the surrounding environment of the vehicle, the parking type of the current parking area is retrieved from a database remotely connected to the driving recorder; the parking types include all-day no-parking sections, temporary parking sections, peak-hour no-parking sections, and sections where the vehicle is parked illegally;

[0019] If the parking type of the current parking area is a no-parking section all day, it is directly determined that the vehicle has parked illegally;

[0020] If the parking type of the current parking area is a temporary parking section, obtaining a first parking duration of the vehicle, and retrieving a first allowed parking duration of the temporary parking section from the database;

[0021] If the first parking time exceeds the first allowed parking time, it is determined that the vehicle has parked illegally;

[0022] If the first parking time does not exceed the first allowed parking time, the time difference between the first parking time and the first allowed parking time is calculated as the first allowed temporary parking time, and the first allowed temporary parking time is notified to the vehicle owner through information push or in-vehicle broadcasting;

[0023] If the parking type of the current parking area is a no-parking section during peak hours, determining whether the current time point is during peak hours;

[0024] If the current time point is during a peak period, it is determined that the vehicle has parked illegally;

[0025] If the current time point is not in the peak period, the second temporary parking duration corresponding to the nearest peak period is calculated with the current time point as the starting point, and the second temporary parking duration is notified to the vehicle owner through information push or in-vehicle broadcasting;

[0026] If the parking type of the current parking area is a road section where the vehicle leaves the vehicle illegally, judging whether the vehicle owner has left the vehicle according to the surrounding environment of the vehicle;

[0027] If the owner leaves the vehicle, it is determined that the vehicle has been illegally parked;

[0028] If the vehicle owner has not left the vehicle, the second parking time of the vehicle is obtained, and the second allowed parking time of the illegal parking section is retrieved from the database;

[0029] If the second parking time exceeds the second allowed parking time, it is determined that the vehicle has parked illegally;

[0030] If the second parking time does not exceed the second allowed parking time, the time difference between the second parking time and the second allowed parking time is calculated as the third allowed temporary parking time, and the third allowed temporary parking time is notified to the vehicle owner through information push or in-vehicle broadcast.

[0031] Optionally, the construction and training process of the illegal parking detection model includes:

[0032] Building a parking violation detection architecture based on target detection algorithms; the target detection algorithms include YOLO, Faster R-CNN, and SSD;

[0033] Collect a large amount of labeled historical image data or historical video frame data as an original illegal parking data set; the original illegal parking data set contains illegal parking and non-illegal parking scenes of various types of vehicles in different environments;

[0034] Performing vehicle position marking, vehicle type marking, illegal parking status marking, and key image feature extraction processing on the original illegal parking data set to obtain a target illegal parking data set;

[0035] Dividing the target illegal parking data set into an illegal parking training set, an illegal parking verification set, and an illegal parking test set according to a preset ratio;

[0036] The illegal parking detection architecture constructed by training the illegal parking training set is continuously adjusted through the back propagation algorithm to minimize the gap between the predicted result and the actual label;

[0037] Evaluate the model performance of the trained illegal parking detection architecture according to the illegal parking validation set, and further tune the model parameters based on the model performance evaluation results;

[0038] The model generalization ability of the illegal parking detection architecture after parameter tuning is tested according to the illegal parking test set, and the obtained illegal parking detection model is deployed on the driving recorder of the vehicle.

[0039] Optionally, the driving recorder illegal parking detection method further includes:

[0040] Real-time collection of illegal parking feedback information after illegal parking detection and identification;

[0041] When a new illegal parking mode is determined based on the illegal parking feedback information, and / or a new illegal parking area is added, and / or the illegal parking rules are updated, a small amount of new illegal parking samples are collected, and the illegal parking detection model is fine-tuned in real time based on the new illegal parking samples;

[0042] On the basis of real-time model fine-tuning of the illegal parking detection model, new training data sets are regularly collected, and the illegal parking detection model is retrained based on the new training data sets to adapt to environmental changes; the newly added data label types in the new training data sets include seasonal illegal parking patterns and new vehicle types.

[0043] Optionally, the driving recorder illegal parking detection method further includes:

[0044] When it is necessary to deploy illegal parking detection models on vehicles in different areas, the target illegal parking data set is used as a general data set to train, verify and test the constructed illegal parking detection architecture to obtain a general illegal parking detection model;

[0045] A small number of target illegal parking samples are collected from the target area where the target vehicle where the illegal parking detection model needs to be deployed is located, and the general illegal parking detection model is adaptively fine-tuned according to the small number of target illegal parking samples to obtain the target illegal parking detection model corresponding to the target area.

[0046] Optionally, the driving recorder has a built-in buzzer, the driving recorder is bound to the smart terminal of the corresponding owner of the vehicle, and is remotely connected to the relevant traffic management department; the driving recorder illegal parking detection method also includes:

[0047] The illegal parking warning information is voice broadcasted through the buzzer, and / or the illegal parking warning information is prompted on the user display interface of the vehicle, and / or the illegal parking warning information is directly pushed to the smart terminal to notify the owner of the vehicle;

[0048] Extract key illegal parking information from the illegal parking warning information, and report the key illegal parking information to the relevant traffic management department.

[0049] The present invention also provides a driving recorder illegal parking detection device, comprising:

[0050] A data acquisition module, used for continuously recording video frame data around the vehicle through a driving recorder when it is determined that the vehicle is in a stationary state;

[0051] A feature extraction module, used to preprocess the video frame data and extract key image features from the preprocessed video frame data;

[0052] A parking violation detection and recognition module, used for inputting the key image features into a parking violation detection model based on machine vision pre-training to perform parking violation detection and recognition;

[0053] The illegal parking warning information generation module is used to automatically trigger an illegal parking warning and generate illegal parking warning information when it is detected that the vehicle has illegally parked.

[0054] The present invention also provides an electronic device, the device comprising a processor and a memory:

[0055] The memory is used to store program code and transmit the program code to the processor;

[0056] The processor is used to execute the driving recorder illegal parking detection method as described in any one of the above items according to the instructions in the program code.

[0057] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store program code, and the program code is used to execute the driving recorder illegal parking detection method as described in any one of the above items.

[0058] It can be seen from the above technical solutions that the present invention has the following advantages:

[0059] A method for detecting illegal parking with a dash cam is provided. When it is determined that the vehicle is stationary, the dash cam continuously records the video frame data around the vehicle; the video frame data is preprocessed, and key image features are extracted from the preprocessed video frame data; the key image features are input into the illegal parking detection model based on machine vision pre-training for illegal parking detection and identification; when illegal parking behavior of the vehicle is detected, an illegal parking warning is automatically triggered, and illegal parking warning information is generated. Thus, by combining the dash cam with the machine vision algorithm and providing an automated illegal parking detection method, the illegal parking behavior of the vehicle can be monitored in real time, and active warnings can be issued when illegal parking behavior occurs, thereby reducing the occurrence of illegal parking incidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0061] Figure 1 A flowchart of a method for detecting illegal parking with a driving recorder;

[0062] Figure 2 It is a schematic diagram of the overall process of a driving recorder illegal parking detection method;

[0063] Figure 3 The present invention is a structural block diagram of a driving recorder illegal parking detection device. DETAILED DESCRIPTION

[0064] The embodiments of the present invention provide a method, device, electronic device and storage medium for detecting illegal parking with a driving recorder, which are used to solve or partially solve the technical problem that the related technology lacks an automated illegal parking monitoring mechanism and cannot realize real-time warning or active reporting of illegal parking behavior.

[0065] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] As an example, in order to effectively manage and detect illegal parking of vehicles, the current solutions are mainly concentrated in urban fixed monitoring systems and some advanced driver assistance systems (ADAS).

[0067] Urban fixed monitoring systems are often limited by fixed installation locations and limited monitoring range. In complex and changing urban environments (such as occlusion, light changes, etc.), they are easily affected by weather, light conditions and background interference, resulting in high false alarm and missed alarm rates and low illegal parking recognition accuracy. At the same time, fixed monitoring cameras are expensive to deploy and difficult to achieve large-scale coverage. Especially in small streets or newly formed illegal parking hotspots, there are many monitoring blind spots.

[0068] As for ADAS, although some high-end models are equipped with sensor-based parking assistance functions, which can remind drivers of the legality of parking locations, these systems are mainly used to assist drivers to park correctly, and do not have the ability to monitor illegal parking after the vehicle is turned off, nor can they provide evidence of illegal parking to traffic management departments.

[0069] Furthermore, although there are cases of using dashcams to record traffic incidents in the market, most applications focus on determining accident responsibility. This is achieved by having users manually mark the incident and then upload the video clip. This approach lacks an automated illegal parking monitoring mechanism and cannot provide real-time warnings or proactively report illegal parking behaviors. Therefore, this method has little effect on illegal parking management.

[0070] Through further analysis, the present invention finds that both the fixed monitoring method and the current use of driving recorders are biased towards post-event evidence collection rather than real-time monitoring and immediate feedback, making it difficult to effectively prevent illegal parking and timely divert traffic.

[0071] The root cause of the above problems is the lack of a set of illegal parking detection solutions that integrate intelligent analysis and active response mechanisms. Traditional methods rely too much on manual monitoring or single-function hardware, and do not fully integrate the latest machine learning, cloud computing and Internet of Things technologies to achieve low-cost, high-efficiency and wide-coverage illegal parking management. Developing an intelligent driving recorder solution that can automatically identify illegal parking scenes, conduct real-time analysis and judgment, and actively alert car owners or management departments has become the key to solving the above problems.

[0072] Therefore, one of the core invention points of the embodiment of the present invention is: to improve the efficiency of illegal parking supervision through intelligent means, reduce traffic congestion and safety hazards caused by illegal parking, and provide a driving recorder illegal parking detection method combined with machine vision technology. On the one hand, the video data recorded by the driving recorder is automatically analyzed by machine vision algorithm to determine in real time whether the vehicle is in an illegal parking state. On the other hand, when illegal parking is detected, an alert is sent to the owner through the cloud platform or directly to achieve early detection and intervention of illegal parking, thereby reducing the pressure of urban traffic management. Thus, through real-time monitoring and automatic alarm, the speed of discovering illegal parking is significantly improved, traffic congestion caused by illegal parking is reduced, and the smoothness of urban traffic is improved. At the same time, through model adaptive learning and optimization, the illegal parking detection model can be continuously optimized according to historical data to improve recognition accuracy. The scheme provided by the embodiment of the present invention can effectively improve the efficiency and effect of urban parking management through automated and intelligent technical means, and provide strong support for building an intelligent transportation system.

[0073] Reference Figure 1 , shows a flowchart of a method for detecting illegal parking with a driving recorder provided by an embodiment of the present invention, which may specifically include the following steps:

[0074] Step 101, when it is determined that the vehicle is in a stationary state, continuously recording video frame data around the vehicle through a driving recorder;

[0075] Before checking for illegal parking, you need to select a suitable dashcam. The dashcam you choose should have high resolution and wide dynamic range to clearly capture vehicle and environmental details.

[0076] More specifically, the driving recorder may mainly include a visual sensing module, a photosensitive sensing module, a data processing module, and a network connection module.

[0077] The visual sensor module (camera system) is a key component in the dashcam, responsible for capturing images or video data in front of or around the vehicle. By continuously recording video information around the vehicle, key visual elements such as the vehicle's position and surrounding signs and markings can be obtained. The highly sensitive camera can obtain clear image records under different lighting conditions.

[0078] The light-sensitive sensor module (photosensitive sensor) is used to detect changes in ambient light. It can help the dashcam automatically adjust shooting parameters (such as exposure time, ISO sensitivity (International Organization for Standardization Sensitivity, a standardized indicator used to measure the sensitivity of image sensors to light in photography and videography)) to adapt to different lighting conditions. This allows for usable video recording data to be obtained even at night or in low-light environments.

[0079] The data processing module can be regarded as the "brain" of the driving recorder. The data processing module is responsible for receiving data from the visual sensor module and the photosensitive sensor module and processing it. The conventional processing content of this module may mainly include image analysis, compression, and storage. And according to the preset algorithm, it is determined whether there is a potential safety risk (for example, the distance to the front vehicle is too close, the lane is deviated, etc.), and the corresponding safety reminder is triggered. In an embodiment of the present invention, based on the data processing module, the collected image data or video frame data is processed in real time in the driving recorder or by working in collaboration with the cloud server, and the illegal parking detection algorithm is used to quickly identify the illegal parking behavior of the vehicle.

[0080] In order to transmit data and receive updates in real time, the driving recorder should have a stable network connection capability. For the network connection module, the driving recorder can be connected to the owner's smart terminal (such as a mobile phone, tablet or other smart device) through a built-in WiFi module or Bluetooth technology. Realize data transmission and remote control. For example, the owner can view real-time recordings, download video files, adjust the settings of the driving recorder, etc. through the APP (Application, mobile application) installed on the mobile phone. The APP connects to the hotspot created by the driving recorder wirelessly or synchronizes data directly through the Internet cloud service. In order to protect sensitive data from illegal access, it is also necessary to establish a data encryption and transmission security mechanism based on the network connection module. Since data encryption and transmission security mechanisms are not the key contents to be protected in the embodiments of the present invention, this part of the content can be implemented with reference to the relevant technology, and will not be repeated here.

[0081] In addition, the dashcam also needs to integrate additional sensors such as GPS (Global Positioning System) modules and accelerometers. GPS is used to obtain accurate geographic location information, and accelerometers are used to assist in determining vehicle status.

[0082] Traditional illegal parking monitoring usually relies on fixed cameras with limited coverage and fixed locations. However, the illegal parking detection solution based on dashcams uses mobile vehicles as monitoring points, which can cover almost every corner of the city. This greatly increases the density and flexibility of the monitoring network. It is especially suitable for monitoring complex or temporary control areas where it is difficult to preset monitoring points. Compared with the large-scale deployment of static surveillance cameras, using the already popular dashcams as the front end of illegal parking detection can significantly reduce the cost of infrastructure construction and maintenance. In addition, this distributed monitoring mode reduces the need for expensive professional monitoring equipment.

[0083] Before detecting illegal parking of a vehicle based on a dashcam, the vehicle status needs to be identified first. Specifically, an algorithm can be developed to combine the data measured by the GPS module and the accelerometer to determine whether the vehicle is stationary (i.e. parked). When the vehicle is judged to be stationary, the dashcam can be used to continuously record the video frame data around the vehicle for subsequent illegal parking detection and identification.

[0084] It is understandable that, regardless of whether the vehicle is currently stationary, the driving recorder can continue to record video frame data around the vehicle. Since the actual application scenario of the technical solution provided by the embodiment of the present invention is vehicle illegal parking detection and identification, the video frame data selected for subsequent processing is the video frame data in the stationary state.

[0085] Step 102, preprocessing the video frame data, and extracting key image features from the preprocessed video frame data;

[0086] Among them, the preprocessing of images such as video frames can mainly include noise reduction, brightness adjustment, color balance, etc. Through image preprocessing, the image quality can be improved, thereby further improving the subsequent detection accuracy.

[0087] Using deep learning models, such as Convolutional Neural Network (CNN), the pre-processed video frame data is subjected to feature recognition and extraction to obtain key image features including vehicle outline, license plate information, and vehicle surroundings. Among them, key vehicle features, such as license plate information including license plate numbers, record detailed information on illegally parked vehicles and can provide a basis for law enforcement.

[0088] Step 103, inputting the key image features into a parking violation detection model based on machine vision pre-training to perform parking violation detection and recognition;

[0089] Before actually performing illegal parking detection and recognition, it is necessary to build and train an illegal parking detection model. In the embodiment of the present invention, based on deep learning image recognition technology, a set of algorithm models that can identify illegal parking signs, parking space lines, no-parking areas, and vehicle parking status are trained. By dynamically analyzing the video frames, it is determined whether the vehicle is parked in the allowed parking area and whether the parking time exceeds the prescribed limit.

[0090] Specifically, the construction and training process of the illegal parking detection model can be implemented by executing the following sub-steps S01 to S07:

[0091] Step S01: construct a parking violation detection architecture based on the target detection algorithm.

[0092] Select the appropriate parking violation detection architecture according to the task requirements. The parking violation detection architecture can further include the target detection architecture and the classification model. The target detection architecture can be constructed by target detection algorithms such as YOLO (You Only Look Once, a single-stage target detection algorithm), Faster R-CNN (Region-based Convolutional Neural Network, a faster regional convolutional neural network), SSD (Single Shot MultiBox Detector, single shot multi-box detector). The parking violation detection architecture is used to identify and locate vehicles. The classification model is used to determine whether the vehicle is illegally parked.

[0093] On the basis of the classification model, it is also necessary to define the judgment criteria for illegal parking according to local traffic regulations, that is, to build an illegal parking judgment mechanism, such as prohibited parking areas, parking time limits, etc.

[0094] It is understandable that when building models and setting up related mechanisms, legal and ethical reviews must be followed and privacy protection must be emphasized. In other words, it is necessary to ensure that the system design complies with local privacy laws and take relevant measures to protect personal privacy when necessary, such as blurring the faces of pedestrians in the background, etc.

[0095] Step S02: Collect a large amount of historical image data with labels, or historical video frame data, as the original illegal parking data set.

[0096] First, it is necessary to collect a large amount of historical images or historical video (frame) data with labels (such as vehicle type, different weather conditions, various illegal parking and legal parking scenes, etc.) to form the original illegal parking data set. The original illegal parking data set contains illegal parking and non-illegal parking scenes of various types of vehicles in different environments. Among them, the data source can mainly be existing driving recorder videos, urban surveillance videos, etc. It should be noted that all data sources used in the embodiments of the present invention are legal.

[0097] Step S03: Perform vehicle position marking, vehicle type marking, illegal parking status marking and key image feature extraction processing on the original illegal parking data set to obtain a target illegal parking data set.

[0098] Next, the collected data needs to be annotated in detail. This mainly involves marking the location of the vehicle, the type of vehicle, and whether it is illegally parked. To improve the accuracy of the annotation, this step requires human participation.

[0099] In addition, to improve image quality and facilitate algorithm analysis, the images in the dataset can also be preprocessed, such as scaling, cropping, contrast enhancement, etc.

[0100] Using deep learning models, such as convolutional neural networks (CNNs), we perform feature recognition and extraction on the images in the preprocessed dataset to obtain historical key image features including vehicle outlines, license plate information, and vehicle surroundings. The extracted historical key image features and the original illegal parking dataset are used as the target illegal parking dataset for model training.

[0101] Step S04: Divide the target illegal parking data set into an illegal parking training set, an illegal parking verification set and an illegal parking test set according to a preset ratio.

[0102] For example, the target illegal parking data set is divided into illegal parking training set, illegal parking verification set and illegal parking test set in a ratio of 8:1:1.

[0103] Step S05: The illegal parking detection architecture is constructed according to the illegal parking training set, and the network weights are continuously adjusted through the back propagation algorithm to minimize the gap between the predicted results and the actual labels.

[0104] Step S06: Evaluate the model performance of the trained illegal parking detection architecture according to the illegal parking validation set, and further tune the model parameters based on the model performance evaluation results.

[0105] For the trained parking violation detection architecture, evaluate the model performance on the parking violation validation set and adjust hyperparameters such as learning rate and batch size to optimize model accuracy.

[0106] Step S07: The model generalization capability of the illegal parking detection architecture after parameter tuning is tested based on the illegal parking test set, and the obtained illegal parking detection model is deployed on the vehicle's driving recorder.

[0107] The generalization ability of the illegal parking detection architecture after parameter tuning is tested on an independent illegal parking test set, so that the model can perform well on new data.

[0108] The trained illegal parking detection model, together with the hardware equipment and communication module, is integrated and deployed on the vehicle's driving recorder. Through the collaborative work of various modules, the video stream is analyzed in real time to identify illegal parking behaviors.

[0109] Before the system is officially put into use, you can also use simulated video streams in a laboratory environment to test the system performance, including recognition speed, accuracy, stability, etc.

[0110] The illegal parking detection model used in the embodiment of the present invention can not only detect and identify illegal parking of vehicles, but also perform adaptive learning and optimization based on the model itself. On the one hand, by collecting real-time feedback, the model can continue to learn online. For example, when a new illegal parking mode, a new illegal parking area, or a change in illegal parking rules is added, the model can be fine-tuned with a small number of new samples.

[0111] In a specific implementation, the feedback information of illegal parking after the detection and identification of illegal parking of vehicles can be collected in real time (such as a new illegal parking mode is added to the feedback area, or a new illegal parking area is added on the basis of the original illegal parking area in the area, or the illegal parking rules of the area are updated for some reason, etc.). When the new illegal parking mode and / or the new illegal parking area and / or the updated illegal parking rules are determined based on the illegal parking feedback information, a small amount of new illegal parking samples are collected, and the illegal parking detection model is fine-tuned in real time based on the new illegal parking samples.

[0112] On the other hand, over time, the model can receive new data regularly for retraining to adapt to changes in the environment, thereby achieving incremental learning, such as seasonal parking violations, new vehicle types, etc.

[0113] That is to say, on the basis of real-time model fine-tuning of the illegal parking detection model, new training data sets can be collected regularly, and the illegal parking detection model can be retrained based on the new training data sets to adapt to environmental changes. Among them, the newly added data label types in the new training data sets can include seasonal illegal parking patterns, new vehicle types, etc.

[0114] In another case, if it is necessary to deploy the illegal parking detection model provided by the technical solution of the present invention in different cities or different countries, the model can be pre-trained on a general data set first, and then fine-tuned on a small amount of data in a specific area to quickly adapt to the local specific conditions and realize the transfer learning of the model.

[0115] Therefore, when it is necessary to deploy illegal parking detection models on vehicles in different areas, the target illegal parking dataset can be used as a general dataset to train, verify and test the constructed illegal parking detection architecture to obtain a general illegal parking detection model. Collect a small number of target illegal parking samples from the target area where the target vehicle where the illegal parking detection model needs to be deployed is located, and adaptively fine-tune the general illegal parking detection model based on a small number of target illegal parking samples to obtain the target illegal parking detection model corresponding to the target area.

[0116] Therefore, by setting up continuous model optimization methods, the algorithm can be regularly updated and optimized based on user feedback and system performance monitoring to adapt to new environmental and regulatory changes.

[0117] Through the above-mentioned relevant settings for the illegal parking detection model, the machine vision-based driving recorder illegal parking detection algorithm provided by the embodiment of the present invention can not only efficiently identify various parking signs, markings, and no-parking areas, and accurately determine the parking status of the vehicle, but also realize the automation of the entire process from data to decision-making. At the same time, through a large number of labeled sample training, combined with an adaptive learning mechanism, the accuracy and robustness of the model can be continuously improved, and high-precision illegal parking behavior recognition can be achieved to adapt to the complex and changing urban environment. Even under complex lighting and occlusion conditions, a high accuracy rate can be maintained.

[0118] In combination with the previous discussion, the process of inputting key image features into the illegal parking detection model based on machine vision pre-training for illegal parking detection and identification can be specifically as follows: inputting the vehicle outline, license plate information and the vehicle's surrounding environment into the illegal parking detection model based on machine vision pre-training for vehicle positioning and identification, and determining the vehicle's current parking area; combining the vehicle's surrounding environment to identify the parking type of the current parking area, and based on the parking type and the preset illegal parking conditions, determine whether the vehicle has illegally parked.

[0119] In a specific implementation, the process of identifying the parking type of the current parking area in combination with the surrounding environment of the vehicle, and judging whether the vehicle has illegally parked based on the parking type and the preset illegal parking conditions may mainly include the following sub-steps S11 to S14:

[0120] Step S11: Identify a no-parking sign from the vehicle's surroundings.

[0121] Among them, no parking signs can mainly include all-day no parking signs, temporary parking signs, peak-hour no parking signs, and leaving the vehicle illegal parking signs.

[0122] The all-day no parking sign indicates that parking is prohibited in the parking area at any time. The temporary parking sign indicates that parking is allowed in the parking area within a certain time period. Parking beyond this time period is considered illegal parking. The peak-time no parking sign indicates that parking is prohibited in the parking area during peak hours (such as the morning peak from 7 a.m. to 9 a.m., the evening peak from 5 p.m. to 7 p.m., or other peak periods set according to the local traffic peak), but parking is allowed for a period of time during non-peak hours. The leaving vehicle illegal parking sign indicates that when the owner leaves the vehicle in the parking area and gets off, it is considered illegal parking. When the owner does not leave the vehicle, parking is allowed for a period of time. Among them, there are differences in the time periods allowed for parking in various areas, and technicians in this field can set or enter information based on actual conditions.

[0123] Step S12: If a no-parking sign can be further identified from the vehicle's surroundings, the parking type of the current parking area is determined according to the no-parking sign.

[0124] Corresponding to the no parking signs, the parking types may include all-day no parking sections, temporary parking sections, peak-hour no parking sections, and sections where vehicles are illegally parked.

[0125] Step S13: If no parking prohibition sign can be further identified from the surrounding environment of the vehicle, the parking type of the current parking area is retrieved from a database remotely connected to the driving recorder.

[0126] In one case, the currently extracted vehicle surroundings may not identify relevant signs, and in this case, information can be obtained in the form of data call. Specifically, the parking type of the current parking area can be called from a database remotely connected to the driving recorder.

[0127] Step S14: Based on the parking type and the preset illegal parking conditions, determine whether the vehicle has committed illegal parking.

[0128] Furthermore, the illegal parking conditions and judgment basis corresponding to different parking types are as follows:

[0129] 1. All-day parking prohibited sections

[0130] If the parking type of the current parking area is a no-parking section throughout the day, it can be directly determined that the vehicle has parked illegally.

[0131] (II) Temporary Stop Sections

[0132] If the parking type of the current parking area is a temporary parking section, the first parking duration of the vehicle (i.e., the parking duration from the start of parking to the current time point) can be obtained, and the first allowed parking duration of the temporary parking section (i.e., the duration allowed to park in the parking area under the temporary parking section type) can be retrieved from the database.

[0133] When the first stop time exceeds the first allowed stop time, it can be determined that the vehicle has parked illegally. For example, if the vehicle stops for 2 minutes and the allowed stop time for the temporary stop section is 1 minute, the vehicle is parked illegally.

[0134] When the first parking time does not exceed the first allowed parking time, the time difference between the first parking time and the first allowed parking time can be calculated as the first temporary parking allowed time, and the first temporary parking allowed time can be notified to the owner through information push or in-car broadcast. If the vehicle stops for 2 minutes, the allowed parking time of the temporary parking section is 3 minutes. At this time, the data processing module of the driving recorder can calculate that the vehicle can stop for a maximum of 1 minute, and notify the owner of the information that the vehicle can stop for a maximum of 1 minute through information push or in-car broadcast, so as to inform the owner to control the parking time and avoid illegal parking.

[0135] (III) No-parking sections during peak hours

[0136] If the stop type of the current stop area is a no-stop section during peak hours, it can be determined whether the current time point is during peak hours.

[0137] If the current time is during the peak period, the vehicle is considered to have illegally parked. For example, if the current time is 8 a.m., which is during the morning peak period (7 a.m. to 9 a.m.), the vehicle is considered to have illegally parked.

[0138] If the current time point is not during the peak period, the second temporary parking time corresponding to the nearest peak period is calculated from the current time point as the starting point, and the second temporary parking time is notified to the owner through information push or in-car broadcast. If the current time point is 4 pm, and it is not during the nearest evening peak period (5 pm to 7 pm), the data processing module of the dashcam can calculate that there is 1 hour left before entering the evening peak period, which means that the vehicle can park for a maximum of 1 hour, and the information that the vehicle can park for a maximum of 1 hour is notified to the owner through information push or in-car broadcast.

[0139] In the second case, still taking the current time point of 4 pm as an example, if the no-parking section during peak hours also has an allowed parking time during non-peak hours, such as 10 minutes, and it is not during the recent evening peak period, and the vehicle has been parked for 2 minutes, the data processing module of the driving recorder can calculate that the vehicle can still park for a maximum of 8 minutes, and notify the owner of the vehicle through information push or in-car broadcast.

[0140] (IV) Sections where vehicles are illegally parked

[0141] If the parking type of the current parking area is an illegal parking section where the vehicle owner has left the vehicle, it can be determined whether the vehicle owner has left the vehicle based on the vehicle's surrounding environment and target detection for people.

[0142] When it is detected that the owner has left the vehicle, it is directly judged that the vehicle has been illegally parked.

[0143] When it is detected that the vehicle owner has not left the vehicle, the second parking time of the vehicle is obtained (i.e., the parking time from the start of parking to the current time point), and the second allowed parking time of the illegal parking section is retrieved from the database (i.e., the parking time allowed in the parking area under the illegal parking section type).

[0144] When the second stop time exceeds the second allowed stop time, the vehicle is judged to have parked illegally. For example, if the vehicle stops for 4 minutes and the allowed stop time for the temporary stop section is 2 minutes, the vehicle is considered to have parked illegally.

[0145] When the second stop time does not exceed the second permitted stop time, the time difference between the second stop time and the second permitted stop time is calculated as the third permitted temporary stop time, and the third permitted temporary stop time is notified to the owner through information push or in-car broadcast. For example, if the vehicle stops for 2 minutes, and the permitted stop time of the temporary stop section is 5 minutes, the data processing module of the driving recorder can calculate that the vehicle can stop for a maximum of 3 minutes at present, and notify the owner of the vehicle through information push or in-car broadcast.

[0146] Therefore, by combining the actual parking type with the preset illegal parking conditions, a flexible illegal parking judgment mechanism can be constructed, which can not only identify and judge the different parking types of vehicles, but also give relevant prompts in time when illegal parking occurs. At the same time, it can also give advance warnings for impending illegal parking behaviors, which can largely avoid illegal parking behaviors.

[0147] At the same time, by providing personalized setting options, car owners can adjust parameters such as the sensitivity of illegal parking judgment and the allowed parking time according to their own needs, so as to balance the strictness of supervision and user experience.

[0148] Traditional illegal parking supervision relies on manual patrols or fixed-point monitoring, which is inefficient and prone to omissions. In this step, by automatically analyzing video data with the help of deep learning and advanced image recognition algorithms, illegal parking behaviors can be automatically and accurately identified in real time or near real time in complex environments (such as different light conditions, vehicle occlusions, and changing backgrounds). For example, it can accurately identify prohibited parking areas, fire lanes, and disabled parking spaces, reduce false alarms and missed alarms, reduce reliance on manual monitoring, and improve the accuracy and fairness of law enforcement.

[0149] Step 104, when it is detected that the vehicle has illegally parked, an illegal parking warning is automatically triggered and illegal parking warning information is generated.

[0150] Once illegal parking is detected, the system automatically triggers an alert and generates illegal parking warning information. In addition, the vehicle owner can be notified through the built-in buzzer of the dashcam, the vehicle user display interface information prompt, or through the mobile phone APP push notification. At the same time, key illegal parking information can also be selectively reported to the relevant traffic management department.

[0151] In the specific implementation, the dash cam has a built-in buzzer (such as a vehicle speaker), and the dash cam is bound to the smart terminal of the corresponding owner of the vehicle (such as a mobile phone, tablet and other smart devices), and is remotely connected to the relevant traffic management department. After the illegal parking warning information is generated, the illegal parking warning information can also be voice broadcasted through the buzzer (or a warning sound is directly issued), and / or the illegal parking warning information is prompted on the user display interface of the vehicle, and / or the illegal parking warning information is directly pushed to the smart terminal to notify the corresponding owner of the vehicle. Therefore, through the immediate illegal parking warning, it is helpful to improve the owner's awareness of compliance with parking rules, prompt the owner to consciously find a legal parking space, reduce unintentional violations, and improve the urban parking order.

[0152] Furthermore, key illegal parking information (such as vehicle license plate information, illegal parking location, etc.) can be extracted from the illegal parking warning information, and the key illegal parking information can be reported to the relevant traffic management department.

[0153] After receiving the notification, the car owner can quickly move the vehicle according to the prompt to avoid fines or other penalties. The illegal parking warning information can also be displayed on the vehicle's user display interface or the car owner's mobile phone interface in the form of illegal parking detection results (illegal parking evidence). At the same time, it is also possible to promote users to regulate their parking behavior by setting technical means such as allowing car owners to view illegal parking evidence, submit complaints online, or learn parking rules.

[0154] In addition, the massive amount of illegal parking evidence collected (including images, video clips, time and location, etc.) will be securely stored and integrated into the background management system for later query, review and statistical analysis. Big data analysis can also reveal illegal parking hotspots and patterns, providing a scientific basis for urban traffic planning and parking resource optimization. The illegal parking evidence recorded in the system background can be used as a reference for self-monitoring by car owners or law enforcement by traffic management departments.

[0155] The machine vision-based driving recorder parking violation detection method provided in the embodiment of the present invention is widely applicable to any scenario that requires efficient and accurate parking violation management, such as urban street management, residential area and parking lot entrances, special area monitoring, highway service areas and rest areas, fleet management and car rental services, and private car self-regulation. It helps to improve the level of urban traffic management and maintain public safety and order.

[0156] In an embodiment of the present invention, a method for detecting illegal parking using a dashcam combined with machine vision technology is provided. On the one hand, a machine vision algorithm is used to automatically analyze the video data recorded by the dashcam to determine in real time whether the vehicle is in an illegal parking state. On the other hand, when illegal parking is detected, a warning is sent through a cloud platform or directly to the owner to achieve early detection and intervention of illegal parking, thereby reducing the pressure on urban traffic management. Thus, through real-time monitoring and automatic alarm, the speed of detecting illegal parking can be significantly improved, traffic congestion caused by illegal parking can be reduced, and the smoothness of urban traffic can be improved. At the same time, through model adaptive learning and optimization, the illegal parking detection model can be continuously optimized according to historical data to improve recognition accuracy. By adopting the solution provided by the embodiment of the present invention, the efficiency and effectiveness of urban parking management can be effectively improved through automated and intelligent technical means, providing strong support for building an intelligent transportation system.

[0157] For better explanation, refer to Figure 2 , showing an overall flow chart of a driving recorder parking violation detection method provided by an embodiment of the present invention. It should be pointed out that this embodiment only briefly describes the general process of driving recorder parking violation detection, and the specific implementation process of each step can be understood by referring to the relevant content in the aforementioned embodiment, which will not be described here. It can be understood that the present invention does not limit this.

[0158] Step 201: constructing a parking violation detection framework based on the target detection algorithm, training, verifying and testing the parking violation detection framework based on the collected historical parking violation data set, and deploying the obtained parking violation detection model on the vehicle's driving recorder;

[0159] Step 202: When it is determined that the vehicle is stationary, the driving recorder continuously records video frame data around the vehicle, pre-processes the video frame data, and extracts the vehicle outline, license plate information, and the vehicle surrounding environment from the pre-processed video frame data;

[0160] Step 203: Input the vehicle outline, license plate information and the surrounding environment of the vehicle into the illegal parking detection model to perform vehicle positioning and identification, determine the current parking area of ​​the vehicle, then identify the parking type of the current parking area in combination with the surrounding environment of the vehicle, and based on the parking type and the preset illegal parking conditions, determine whether the vehicle has illegal parking behavior;

[0161] Step 204: When a vehicle is detected to have illegally parked, an illegal parking warning is automatically triggered, and illegal parking warning information is generated;

[0162] Step 205: the illegal parking warning information is voice broadcasted through a buzzer, and / or the illegal parking warning information is prompted on a user display interface of the vehicle, and / or the illegal parking warning information is directly pushed to a smart terminal to notify the corresponding vehicle owner, and key illegal parking information is extracted from the illegal parking warning information, and the key illegal parking information is reported to the relevant traffic management department;

[0163] Step 206: During the use of the model, the illegal parking detection model is fine-tuned in real time and trained again to achieve adaptive learning and optimization of the model.

[0164] Reference Figure 3 , shows a structural block diagram of a driving recorder parking violation detection device provided by an embodiment of the present invention, which may specifically include:

[0165] The data acquisition module 301 is used to continuously record the video frame data around the vehicle through the driving recorder when it is determined that the vehicle is in a stationary state;

[0166] A feature extraction module 302 is used to preprocess the video frame data and extract key image features from the preprocessed video frame data;

[0167] The illegal parking detection and identification module 303 is used to input the key image features into the illegal parking detection model based on machine vision pre-training to perform illegal parking detection and identification;

[0168] The illegal parking warning information generating module 304 is used to automatically trigger an illegal parking warning and generate illegal parking warning information when it is detected that the vehicle has an illegal parking behavior.

[0169] In an optional embodiment, the key image features include the vehicle outline, license plate information and the vehicle surroundings; the illegal parking detection and recognition module 303 includes:

[0170] A vehicle positioning and recognition module, used to input the vehicle outline, the license plate information and the surrounding environment of the vehicle into a parking violation detection model based on machine vision pre-training to perform vehicle positioning and recognition, and determine the current parking area of ​​the vehicle;

[0171] The illegal parking identification and judgment module is used to identify the parking type of the current parking area in combination with the vehicle's surrounding environment, and based on the parking type and preset illegal parking conditions, determine whether the vehicle has committed illegal parking.

[0172] In an optional embodiment, the illegal parking identification and judgment module includes:

[0173] A no-parking sign recognition module is used to recognize a no-parking sign from the surrounding environment of the vehicle; the no-parking sign includes a no-parking sign for all day, a temporary parking sign, a no-parking sign during peak hours, and a sign for leaving the vehicle and illegally parking;

[0174] a first parking type determination module, configured to determine the parking type of the current parking area according to the no parking sign when a no parking sign can be further identified from the surrounding environment of the vehicle;

[0175] The second parking type determination module is used to retrieve the parking type of the current parking area from a database remotely connected to the driving recorder when no further parking prohibition signs can be identified from the surrounding environment of the vehicle; the parking types include all-day no-parking sections, temporary parking sections, peak-hour no-parking sections, and sections for leaving the vehicle and illegally parking;

[0176] An all-day no-parking road section judgment module is used to directly judge that the vehicle has illegally parked when the parking type of the current parking area is an all-day no-parking road section;

[0177] The temporary stop section judgment module is specifically used for:

[0178] If the parking type of the current parking area is a temporary parking section, obtaining a first parking duration of the vehicle, and retrieving a first allowed parking duration of the temporary parking section from the database;

[0179] If the first parking time exceeds the first allowed parking time, it is determined that the vehicle has parked illegally;

[0180] If the first parking time does not exceed the first allowed parking time, the time difference between the first parking time and the first allowed parking time is calculated as the first allowed temporary parking time, and the first allowed temporary parking time is notified to the vehicle owner through information push or in-vehicle broadcasting;

[0181] The peak-hour no-parking section judgment module is specifically used for:

[0182] If the parking type of the current parking area is a no-parking section during peak hours, determining whether the current time point is during peak hours;

[0183] If the current time point is during a peak period, it is determined that the vehicle has parked illegally;

[0184] If the current time point is not in the peak period, the second temporary parking duration corresponding to the nearest peak period is calculated with the current time point as the starting point, and the second temporary parking duration is notified to the vehicle owner through information push or in-vehicle broadcasting;

[0185] The module for judging illegal parking sections is specifically used for:

[0186] If the parking type of the current parking area is a road section where the vehicle leaves the vehicle illegally, judging whether the vehicle owner has left the vehicle according to the surrounding environment of the vehicle;

[0187] If the owner leaves the vehicle, it is determined that the vehicle has been illegally parked;

[0188] If the vehicle owner has not left the vehicle, the second parking time of the vehicle is obtained, and the second allowed parking time of the illegal parking section is retrieved from the database;

[0189] If the second parking time exceeds the second allowed parking time, it is determined that the vehicle has parked illegally;

[0190] If the second parking time does not exceed the second allowed parking time, the time difference between the second parking time and the second allowed parking time is calculated as the third allowed temporary parking time, and the third allowed temporary parking time is notified to the vehicle owner through information push or in-vehicle broadcast.

[0191] In an optional embodiment, the driving recorder parking violation detection device further includes:

[0192] A parking violation detection architecture building module, used to build a parking violation detection architecture based on a target detection algorithm; the target detection algorithm includes YOLO, Faster R-CNN, and SSD;

[0193] The original illegal parking data set construction module is used to collect a large amount of historical image data with labels, or historical video frame data, as the original illegal parking data set; the original illegal parking data set contains illegal parking and non-illegal parking scenes of various types of vehicles in different environments;

[0194] A target illegal parking data set generation module is used to perform vehicle position marking, vehicle type marking, illegal parking status marking and key image feature extraction processing on the original illegal parking data set to obtain a target illegal parking data set;

[0195] A data set division module, used for dividing the target illegal parking data set into an illegal parking training set, an illegal parking verification set and an illegal parking test set according to a preset ratio;

[0196] A model training module is used to train the constructed illegal parking detection architecture according to the illegal parking training set, and continuously adjust the network weights through a back-propagation algorithm to minimize the gap between the predicted results and the actual labels;

[0197] A model verification module, used to evaluate the model performance of the trained illegal parking detection architecture according to the illegal parking verification set, and further tune the model parameters based on the model performance evaluation results;

[0198] The model testing module is used to test the model generalization ability of the illegal parking detection architecture after parameter tuning according to the illegal parking test set, and deploy the obtained illegal parking detection model on the driving recorder of the vehicle.

[0199] In an optional embodiment, the driving recorder parking violation detection device further includes:

[0200] The illegal parking feedback information acquisition module is used to collect illegal parking feedback information after the illegal parking detection and identification of the vehicle in real time;

[0201] A real-time model fine-tuning module, for collecting a small amount of new illegal parking samples when determining a new illegal parking mode and / or a new illegal parking area and / or updating illegal parking rules according to the illegal parking feedback information, and performing real-time model fine-tuning on the illegal parking detection model according to the new illegal parking samples;

[0202] The model periodic update module is used to regularly collect new training data sets on the basis of real-time model fine-tuning of the illegal parking detection model, and to perform secondary training on the illegal parking detection model based on the new training data sets to adapt to environmental changes; the newly added data label types in the new training data sets include seasonal illegal parking patterns and new vehicle types.

[0203] In an optional embodiment, the driving recorder parking violation detection device further includes:

[0204] A general illegal parking detection model training module is used to train, verify and test the constructed illegal parking detection architecture using the target illegal parking data set as a general data set when the illegal parking detection model needs to be deployed on vehicles in different areas, so as to obtain a general illegal parking detection model;

[0205] The adaptive fine-tuning module is used to collect a small number of target illegal parking samples in the target area where the target vehicle needs to deploy the illegal parking detection model, and adaptively fine-tune the general illegal parking detection model based on the small number of target illegal parking samples to obtain the target illegal parking detection model corresponding to the target area.

[0206] In an optional embodiment, the driving recorder has a built-in buzzer, the driving recorder is bound to the smart terminal of the corresponding owner of the vehicle, and is remotely connected to the relevant traffic management department; the driving recorder illegal parking detection device also includes:

[0207] a parking violation warning information notification module, configured to broadcast the parking violation warning information by voice through the buzzer, and / or to prompt the parking violation warning information on the user display interface of the vehicle, and / or to directly push the parking violation warning information to the smart terminal to notify the owner of the vehicle;

[0208] The key illegal parking information reporting module is used to extract the key illegal parking information from the illegal parking warning information and report the key illegal parking information to the relevant traffic management department.

[0209] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the aforementioned method embodiment.

[0210] It should be noted that, in order to enable those skilled in the art to better distinguish data of the same type but with different actual meanings, the embodiments of the present invention use terms such as first and second to distinguish and explain some technical features. The terms such as first and second are only used to distinguish data and have no other special meanings. It can be understood that the present invention is not limited to this.

[0211] An embodiment of the present invention further provides an electronic device, the device comprising a processor and a memory:

[0212] The memory is used to store the program code and transmit the program code to the processor;

[0213] The processor is used to execute the driving recorder illegal parking detection method of any embodiment of the present invention according to the instructions in the program code.

[0214] An embodiment of the present invention further provides a computer-readable storage medium, which is used to store program code, and the program code is used to execute the driving recorder illegal parking detection method of any embodiment of the present invention.

[0215] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0216] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0217] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0218] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0219] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0220] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A driving recorder illegal parking detection method, characterized in that: include: When it is determined that the vehicle is stationary, continuously recording video frame data around the vehicle through a driving recorder; Preprocessing the video frame data, and extracting key image features from the preprocessed video frame data; Inputting the key image features into a parking violation detection model based on machine vision pre-training to perform parking violation detection and identification; When illegal parking is detected for the vehicle, an illegal parking warning is automatically triggered and illegal parking warning information is generated.

2. The driving recorder parking violation detection method according to claim 1, characterized in that: The key image features include the vehicle outline, license plate information and the vehicle surrounding environment; the inputting the key image features into the illegal parking detection model based on machine vision pre-training to perform illegal parking detection and identification includes: Inputting the vehicle outline, the license plate information and the surrounding environment of the vehicle into a parking violation detection model based on machine vision pre-training to perform vehicle positioning and identification, and determine the current parking area of ​​the vehicle; In combination with the surrounding environment of the vehicle, the parking type of the current parking area is identified, and based on the parking type and the preset illegal parking conditions, it is determined whether the vehicle has committed illegal parking.

3. The driving recorder parking violation detection method according to claim 2, characterized in that: The identifying the parking type of the current parking area in combination with the surrounding environment of the vehicle, and judging whether the vehicle has illegal parking behavior based on the parking type and preset illegal parking conditions, includes: Identify a no-parking sign from the environment around the vehicle; the no-parking sign includes a no-parking sign for all day, a temporary parking sign, a no-parking sign during peak hours, and a sign for leaving the vehicle and illegally parking; If a no-parking sign can be further identified from the vehicle surroundings, determining the parking type of the current parking area according to the no-parking sign; If no parking prohibition sign can be further identified from the surrounding environment of the vehicle, the parking type of the current parking area is retrieved from a database remotely connected to the driving recorder; the parking types include all-day no-parking sections, temporary parking sections, peak-hour no-parking sections, and sections where the vehicle is parked illegally; If the parking type of the current parking area is a no-parking section all day, it is directly determined that the vehicle has parked illegally; If the parking type of the current parking area is a temporary parking section, obtaining a first parking duration of the vehicle, and retrieving a first allowed parking duration of the temporary parking section from the database; If the first parking time exceeds the first allowed parking time, it is determined that the vehicle has parked illegally; If the first parking time does not exceed the first allowed parking time, the time difference between the first parking time and the first allowed parking time is calculated as the first allowed temporary parking time, and the first allowed temporary parking time is notified to the vehicle owner through information push or in-vehicle broadcasting; If the parking type of the current parking area is a no-parking section during peak hours, determining whether the current time point is during peak hours; If the current time point is during a peak period, it is determined that the vehicle has parked illegally; If the current time point is not in the peak period, the second temporary parking duration corresponding to the nearest peak period is calculated with the current time point as the starting point, and the second temporary parking duration is notified to the vehicle owner through information push or in-vehicle broadcasting; If the parking type of the current parking area is a road section where the vehicle leaves the vehicle illegally, judging whether the vehicle owner has left the vehicle according to the surrounding environment of the vehicle; If the owner leaves the vehicle, it is determined that the vehicle has been illegally parked; If the vehicle owner has not left the vehicle, the second parking time of the vehicle is obtained, and the second allowed parking time of the illegal parking section is retrieved from the database; If the second parking time exceeds the second allowed parking time, it is determined that the vehicle has parked illegally; If the second parking time does not exceed the second allowed parking time, the time difference between the second parking time and the second allowed parking time is calculated as the third allowed temporary parking time, and the third allowed temporary parking time is notified to the vehicle owner through information push or in-vehicle broadcast.

4. The driving recorder parking violation detection method according to claim 2 or 3, characterized in that: The construction and training process of the illegal parking detection model includes: Building a parking violation detection architecture based on target detection algorithms; the target detection algorithms include YOLO, Faster R-CNN, and SSD; Collect a large amount of labeled historical image data or historical video frame data as an original illegal parking data set; the original illegal parking data set contains illegal parking and non-illegal parking scenes of various types of vehicles in different environments; Performing vehicle position marking, vehicle type marking, illegal parking status marking, and key image feature extraction processing on the original illegal parking data set to obtain a target illegal parking data set; Dividing the target illegal parking data set into an illegal parking training set, an illegal parking verification set, and an illegal parking test set according to a preset ratio; The illegal parking detection architecture constructed by training the illegal parking training set is continuously adjusted through the back propagation algorithm to minimize the gap between the predicted result and the actual label; Evaluate the model performance of the trained illegal parking detection architecture according to the illegal parking validation set, and further tune the model parameters based on the model performance evaluation results; The model generalization ability of the illegal parking detection architecture after parameter tuning is tested according to the illegal parking test set, and the obtained illegal parking detection model is deployed on the driving recorder of the vehicle.

5. The driving recorder parking violation detection method according to claim 4, characterized in that: Also includes: Real-time collection of illegal parking feedback information after illegal parking detection and identification; When a new illegal parking mode is determined based on the illegal parking feedback information, and / or a new illegal parking area is added, and / or the illegal parking rules are updated, a small amount of new illegal parking samples are collected, and the illegal parking detection model is fine-tuned in real time based on the new illegal parking samples; On the basis of real-time model fine-tuning of the illegal parking detection model, new training data sets are regularly collected, and the illegal parking detection model is retrained based on the new training data sets to adapt to environmental changes; The newly added data label types in the new training data set include seasonal illegal parking patterns and new vehicle types.

6. The driving recorder parking violation detection method according to claim 5, characterized in that: Also includes: When it is necessary to deploy illegal parking detection models on vehicles in different areas, the target illegal parking data set is used as a general data set to train, verify and test the constructed illegal parking detection architecture to obtain a general illegal parking detection model; A small number of target illegal parking samples are collected from the target area where the target vehicle where the illegal parking detection model needs to be deployed is located, and the general illegal parking detection model is adaptively fine-tuned according to the small number of target illegal parking samples to obtain the target illegal parking detection model corresponding to the target area.

7. The driving recorder parking violation detection method according to claim 1, characterized in that: The driving recorder has a built-in buzzer, and the driving recorder is bound to the smart terminal of the corresponding owner of the vehicle and is remotely connected to the relevant traffic management department; The driving recorder parking violation detection method further includes: The illegal parking warning information is voice broadcasted through the buzzer, and / or the illegal parking warning information is prompted on the user display interface of the vehicle, and / or the illegal parking warning information is directly pushed to the smart terminal to notify the owner of the vehicle; Extract key illegal parking information from the illegal parking warning information, and report the key illegal parking information to the relevant traffic management department.

8. A driving recorder illegal parking detection device, characterized in that: include: A data acquisition module, used for continuously recording video frame data around the vehicle through a driving recorder when it is determined that the vehicle is in a stationary state; A feature extraction module, used to preprocess the video frame data and extract key image features from the preprocessed video frame data; A parking violation detection and recognition module, used for inputting the key image features into a parking violation detection model based on machine vision pre-training to perform parking violation detection and recognition; The illegal parking warning information generation module is used to automatically trigger an illegal parking warning and generate illegal parking warning information when it is detected that the vehicle has illegally parked.

9. An electronic device, characterized in that: The device comprises a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the driving recorder illegal parking detection method described in any one of claims 1-7 according to the instructions in the program code.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store program code, and the program code is used to execute the driving recorder illegal parking detection method described in any one of claims 1-7.

Images